API Documentation
datasets
This module contains the DataSample class, MultiModalArray, MultiModalSparseArray, MultiModalSparseInfo and MultiModalData, class The DataSample class encapsulates a sample ‘s components nbL and nbEx numbers, MultiModalArray class inherit from numpy ndarray and contains a 2d data ndarray with the shape (n_samples, n_view_i * n_features_i)
0  1  2  3 

xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
xxxxxxxx  xxxx  xxxx  xxxx 
MultiModalSparseArray inherit from scipy sparce matrix with the shape (n_samples, n_view_i * n_features_i)

class
multimodal.datasets.data_sample.
DataSample
(data=None, **kwargs) A DataSample instance
Example: >>> from multimodal.datasets.base import load_dict >>> from multimodal.tests.datasets.get_dataset_path import get_dataset_path >>> from multimodal.datasets.data_sample import DataSample >>> file = 'input_x_dic.pkl' >>> data = load_dict(get_dataset_path(file)) >>> print(data.__class__) <class 'dict'> >>> s = DataSample(data) >>> type(s.data) <class 'multimodal.datasets.data_sample.MultiModalArray'>
 Input:
Parameters: data : dict
kwargs : others arguments
Attributes

clear
() → None. Remove all items from D.

copy
() → a shallow copy of D

data
MultiModalArray

fromkeys
($type, iterable, value=None, /) Returns a new dict with keys from iterable and values equal to value.

get
(k[, d]) → D[k] if k in D, else d. d defaults to None.

items
() → a setlike object providing a view on D's items

keys
() → a setlike object providing a view on D's keys

pop
(k[, d]) → v, remove specified key and return the corresponding value. If key is not found, d is returned if given, otherwise KeyError is raised

popitem
() → (k, v), remove and return some (key, value) pair as a 2tuple; but raise KeyError if D is empty.

setdefault
(k[, d]) → D.get(k,d), also set D[k]=d if k not in D

update
([E, ]**F) → None. Update D from dict/iterable E and F. If E is present and has a .keys() method, then does: for k in E: D[k] = E[k] If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v In either case, this is followed by: for k in F: D[k] = F[k]

values
() → an object providing a view on D's values

class
multimodal.datasets.data_sample.
MultiModalArray
MultiModalArray inherit from numpy ndarray
Parameters: data : can be
 dictionary of multiview array with shape = (n_samples, n_features) for multiview
for each view.
 {0: array([[]],
1: array([[]], …}
 numpy array like with shape = (n_samples, n_features) for multiview
for each view.
 [[[…]],
[[…]], …]
 {array like} with (n_samples, nviews * n_features) with ‘views_ind’ diferent to ‘None’
for Multiview input samples.
views_ind : arraylike (default= None ) if None
[0, n_features//2, n_features]) is constructed (2 views) Paramater specifying how to extract the data views from X:
 views_ind is a 1D array of sorted integers, the entries
indicate the limits of the slices used to extract the views,
where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.
Attributes
views_ind (list of views’ indice (may be None)) n_views (int number of views) shapes_int: list of int numbers of feature for each views :Example: >>> from multimodal.datasets.base import load_dict >>> from multimodal.tests.datasets.get_dataset_path import get_dataset_path >>> from multimodal.datasets.data_sample import DataSample >>> file = ‘input_x_dic.pkl’ >>> data = load_dict(get_dataset_path(file)) >>> print(data.__class__) <class ‘dict’> >>> multiviews = MultiModalArray(data) >>> multiviews.shape (120, 240) >>> multiviews.shapes_int [120, 120] >>> multiviews.n_views 2 
T
Same as self.transpose(), except that self is returned if self.ndim < 2.
Examples
>>> x = np.array([[1.,2.],[3.,4.]]) >>> x array([[ 1., 2.], [ 3., 4.]]) >>> x.T array([[ 1., 3.], [ 2., 4.]]) >>> x = np.array([1.,2.,3.,4.]) >>> x array([ 1., 2., 3., 4.]) >>> x.T array([ 1., 2., 3., 4.])

all
(axis=None, out=None, keepdims=False) Returns True if all elements evaluate to True.
Refer to numpy.all for full documentation.
See also
numpy.all
 equivalent function

any
(axis=None, out=None, keepdims=False) Returns True if any of the elements of a evaluate to True.
Refer to numpy.any for full documentation.
See also
numpy.any
 equivalent function

argmax
(axis=None, out=None) Return indices of the maximum values along the given axis.
Refer to numpy.argmax for full documentation.
See also
numpy.argmax
 equivalent function

argmin
(axis=None, out=None) Return indices of the minimum values along the given axis of a.
Refer to numpy.argmin for detailed documentation.
See also
numpy.argmin
 equivalent function

argpartition
(kth, axis=1, kind='introselect', order=None) Returns the indices that would partition this array.
Refer to numpy.argpartition for full documentation.
New in version 1.8.0.
See also
numpy.argpartition
 equivalent function

argsort
(axis=1, kind='quicksort', order=None) Returns the indices that would sort this array.
Refer to numpy.argsort for full documentation.
See also
numpy.argsort
 equivalent function

astype
(dtype, order='K', casting='unsafe', subok=True, copy=True) Copy of the array, cast to a specified type.
Parameters: dtype : str or dtype
Typecode or datatype to which the array is cast.
order : {‘C’, ‘F’, ‘A’, ‘K’}, optional
Controls the memory layout order of the result. ‘C’ means C order, ‘F’ means Fortran order, ‘A’ means ‘F’ order if all the arrays are Fortran contiguous, ‘C’ order otherwise, and ‘K’ means as close to the order the array elements appear in memory as possible. Default is ‘K’.
casting : {‘no’, ‘equiv’, ‘safe’, ‘same_kind’, ‘unsafe’}, optional
Controls what kind of data casting may occur. Defaults to ‘unsafe’ for backwards compatibility.
 ‘no’ means the data types should not be cast at all.
 ‘equiv’ means only byteorder changes are allowed.
 ‘safe’ means only casts which can preserve values are allowed.
 ‘same_kind’ means only safe casts or casts within a kind, like float64 to float32, are allowed.
 ‘unsafe’ means any data conversions may be done.
subok : bool, optional
If True, then subclasses will be passedthrough (default), otherwise the returned array will be forced to be a baseclass array.
copy : bool, optional
By default, astype always returns a newly allocated array. If this is set to false, and the dtype, order, and subok requirements are satisfied, the input array is returned instead of a copy.
Returns: arr_t : ndarray
Unless copy is False and the other conditions for returning the input array are satisfied (see description for copy input parameter), arr_t is a new array of the same shape as the input array, with dtype, order given by dtype, order.
Raises: ComplexWarning
When casting from complex to float or int. To avoid this, one should use
a.real.astype(t)
.Notes
Starting in NumPy 1.9, astype method now returns an error if the string dtype to cast to is not long enough in ‘safe’ casting mode to hold the max value of integer/float array that is being casted. Previously the casting was allowed even if the result was truncated.
Examples
>>> x = np.array([1, 2, 2.5]) >>> x array([ 1. , 2. , 2.5])
>>> x.astype(int) array([1, 2, 2])

base
Base object if memory is from some other object.
Examples
The base of an array that owns its memory is None:
>>> x = np.array([1,2,3,4]) >>> x.base is None True
Slicing creates a view, whose memory is shared with x:
>>> y = x[2:] >>> y.base is x True

byteswap
(inplace) Swap the bytes of the array elements
Toggle between lowendian and bigendian data representation by returning a byteswapped array, optionally swapped inplace.
Parameters: inplace : bool, optional
If
True
, swap bytes inplace, default isFalse
.Returns: out : ndarray
The byteswapped array. If inplace is
True
, this is a view to self.Examples
>>> A = np.array([1, 256, 8755], dtype=np.int16) >>> map(hex, A) ['0x1', '0x100', '0x2233'] >>> A.byteswap(True) array([ 256, 1, 13090], dtype=int16) >>> map(hex, A) ['0x100', '0x1', '0x3322']
Arrays of strings are not swapped
>>> A = np.array(['ceg', 'fac']) >>> A.byteswap() array(['ceg', 'fac'], dtype='S3')

choose
(choices, out=None, mode='raise') Use an index array to construct a new array from a set of choices.
Refer to numpy.choose for full documentation.
See also
numpy.choose
 equivalent function

clip
(min=None, max=None, out=None) Return an array whose values are limited to
[min, max]
. One of max or min must be given.Refer to numpy.clip for full documentation.
See also
numpy.clip
 equivalent function

compress
(condition, axis=None, out=None) Return selected slices of this array along given axis.
Refer to numpy.compress for full documentation.
See also
numpy.compress
 equivalent function

conj
() Complexconjugate all elements.
Refer to numpy.conjugate for full documentation.
See also
numpy.conjugate
 equivalent function

conjugate
() Return the complex conjugate, elementwise.
Refer to numpy.conjugate for full documentation.
See also
numpy.conjugate
 equivalent function

copy
(order='C') Return a copy of the array.
Parameters: order : {‘C’, ‘F’, ‘A’, ‘K’}, optional
Controls the memory layout of the copy. ‘C’ means Corder, ‘F’ means Forder, ‘A’ means ‘F’ if a is Fortran contiguous, ‘C’ otherwise. ‘K’ means match the layout of a as closely as possible. (Note that this function and :func:numpy.copy are very similar, but have different default values for their order= arguments.)
See also
numpy.copy
,numpy.copyto
Examples
>>> x = np.array([[1,2,3],[4,5,6]], order='F')
>>> y = x.copy()
>>> x.fill(0)
>>> x array([[0, 0, 0], [0, 0, 0]])
>>> y array([[1, 2, 3], [4, 5, 6]])
>>> y.flags['C_CONTIGUOUS'] True

ctypes
An object to simplify the interaction of the array with the ctypes module.
This attribute creates an object that makes it easier to use arrays when calling shared libraries with the ctypes module. The returned object has, among others, data, shape, and strides attributes (see Notes below) which themselves return ctypes objects that can be used as arguments to a shared library.
Parameters: None
Returns: c : Python object
Possessing attributes data, shape, strides, etc.
See also
numpy.ctypeslib
Notes
Below are the public attributes of this object which were documented in “Guide to NumPy” (we have omitted undocumented public attributes, as well as documented private attributes):
 data: A pointer to the memory area of the array as a Python integer. This memory area may contain data that is not aligned, or not in correct byteorder. The memory area may not even be writeable. The array flags and datatype of this array should be respected when passing this attribute to arbitrary Ccode to avoid trouble that can include Python crashing. User Beware! The value of this attribute is exactly the same as self._array_interface_[‘data’][0].
 shape (c_intp*self.ndim): A ctypes array of length self.ndim where the basetype is the Cinteger corresponding to dtype(‘p’) on this platform. This basetype could be c_int, c_long, or c_longlong depending on the platform. The c_intp type is defined accordingly in numpy.ctypeslib. The ctypes array contains the shape of the underlying array.
 strides (c_intp*self.ndim): A ctypes array of length self.ndim where the basetype is the same as for the shape attribute. This ctypes array contains the strides information from the underlying array. This strides information is important for showing how many bytes must be jumped to get to the next element in the array.
 data_as(obj): Return the data pointer cast to a particular ctypes object. For example, calling self._as_parameter_ is equivalent to self.data_as(ctypes.c_void_p). Perhaps you want to use the data as a pointer to a ctypes array of floatingpoint data: self.data_as(ctypes.POINTER(ctypes.c_double)).
 shape_as(obj): Return the shape tuple as an array of some other ctypes type. For example: self.shape_as(ctypes.c_short).
 strides_as(obj): Return the strides tuple as an array of some other ctypes type. For example: self.strides_as(ctypes.c_longlong).
Be careful using the ctypes attribute  especially on temporary arrays or arrays constructed on the fly. For example, calling
(a+b).ctypes.data_as(ctypes.c_void_p)
returns a pointer to memory that is invalid because the array created as (a+b) is deallocated before the next Python statement. You can avoid this problem using eitherc=a+b
orct=(a+b).ctypes
. In the latter case, ct will hold a reference to the array until ct is deleted or reassigned.If the ctypes module is not available, then the ctypes attribute of array objects still returns something useful, but ctypes objects are not returned and errors may be raised instead. In particular, the object will still have the as parameter attribute which will return an integer equal to the data attribute.
Examples
>>> import ctypes >>> x array([[0, 1], [2, 3]]) >>> x.ctypes.data 30439712 >>> x.ctypes.data_as(ctypes.POINTER(ctypes.c_long)) <ctypes.LP_c_long object at 0x01F01300> >>> x.ctypes.data_as(ctypes.POINTER(ctypes.c_long)).contents c_long(0) >>> x.ctypes.data_as(ctypes.POINTER(ctypes.c_longlong)).contents c_longlong(4294967296L) >>> x.ctypes.shape <numpy.core._internal.c_long_Array_2 object at 0x01FFD580> >>> x.ctypes.shape_as(ctypes.c_long) <numpy.core._internal.c_long_Array_2 object at 0x01FCE620> >>> x.ctypes.strides <numpy.core._internal.c_long_Array_2 object at 0x01FCE620> >>> x.ctypes.strides_as(ctypes.c_longlong) <numpy.core._internal.c_longlong_Array_2 object at 0x01F01300>

cumprod
(axis=None, dtype=None, out=None) Return the cumulative product of the elements along the given axis.
Refer to numpy.cumprod for full documentation.
See also
numpy.cumprod
 equivalent function

cumsum
(axis=None, dtype=None, out=None) Return the cumulative sum of the elements along the given axis.
Refer to numpy.cumsum for full documentation.
See also
numpy.cumsum
 equivalent function

data
Python buffer object pointing to the start of the array’s data.

diagonal
(offset=0, axis1=0, axis2=1) Return specified diagonals. In NumPy 1.9 the returned array is a readonly view instead of a copy as in previous NumPy versions. In a future version the readonly restriction will be removed.
Refer to
numpy.diagonal()
for full documentation.See also
numpy.diagonal
 equivalent function

dot
(b, out=None) Dot product of two arrays.
Refer to numpy.dot for full documentation.
See also
numpy.dot
 equivalent function
Examples
>>> a = np.eye(2) >>> b = np.ones((2, 2)) * 2 >>> a.dot(b) array([[ 2., 2.], [ 2., 2.]])
This array method can be conveniently chained:
>>> a.dot(b).dot(b) array([[ 8., 8.], [ 8., 8.]])

dtype
Datatype of the array’s elements.
Parameters: None Returns: d : numpy dtype object See also
numpy.dtype
Examples
>>> x array([[0, 1], [2, 3]]) >>> x.dtype dtype('int32') >>> type(x.dtype) <type 'numpy.dtype'>

dump
(file) Dump a pickle of the array to the specified file. The array can be read back with pickle.load or numpy.load.
Parameters: file : str
A string naming the dump file.

dumps
() Returns the pickle of the array as a string. pickle.loads or numpy.loads will convert the string back to an array.
Parameters: None

fill
(value) Fill the array with a scalar value.
Parameters: value : scalar
All elements of a will be assigned this value.
Examples
>>> a = np.array([1, 2]) >>> a.fill(0) >>> a array([0, 0]) >>> a = np.empty(2) >>> a.fill(1) >>> a array([ 1., 1.])

flags
Information about the memory layout of the array.
Notes
The flags object can be accessed dictionarylike (as in
a.flags['WRITEABLE']
), or by using lowercased attribute names (as ina.flags.writeable
). Short flag names are only supported in dictionary access.Only the UPDATEIFCOPY, WRITEABLE, and ALIGNED flags can be changed by the user, via direct assignment to the attribute or dictionary entry, or by calling ndarray.setflags.
The array flags cannot be set arbitrarily:
 UPDATEIFCOPY can only be set
False
.  ALIGNED can only be set
True
if the data is truly aligned.  WRITEABLE can only be set
True
if the array owns its own memory or the ultimate owner of the memory exposes a writeable buffer interface or is a string.
Arrays can be both Cstyle and Fortranstyle contiguous simultaneously. This is clear for 1dimensional arrays, but can also be true for higher dimensional arrays.
Even for contiguous arrays a stride for a given dimension
arr.strides[dim]
may be arbitrary ifarr.shape[dim] == 1
or the array has no elements. It does not generally hold thatself.strides[1] == self.itemsize
for Cstyle contiguous arrays orself.strides[0] == self.itemsize
for Fortranstyle contiguous arrays is true.Attributes
C_CONTIGUOUS (C) The data is in a single, Cstyle contiguous segment. F_CONTIGUOUS (F) The data is in a single, Fortranstyle contiguous segment. OWNDATA (O) The array owns the memory it uses or borrows it from another object. WRITEABLE (W) The data area can be written to. Setting this to False locks the data, making it readonly. A view (slice, etc.) inherits WRITEABLE from its base array at creation time, but a view of a writeable array may be subsequently locked while the base array remains writeable. (The opposite is not true, in that a view of a locked array may not be made writeable. However, currently, locking a base object does not lock any views that already reference it, so under that circumstance it is possible to alter the contents of a locked array via a previously created writeable view onto it.) Attempting to change a nonwriteable array raises a RuntimeError exception. ALIGNED (A) The data and all elements are aligned appropriately for the hardware. UPDATEIFCOPY (U) This array is a copy of some other array. When this array is deallocated, the base array will be updated with the contents of this array. FNC F_CONTIGUOUS and not C_CONTIGUOUS. FORC F_CONTIGUOUS or C_CONTIGUOUS (onesegment test). BEHAVED (B) ALIGNED and WRITEABLE. CARRAY (CA) BEHAVED and C_CONTIGUOUS. FARRAY (FA) BEHAVED and F_CONTIGUOUS and not C_CONTIGUOUS.  UPDATEIFCOPY can only be set

flat
A 1D iterator over the array.
This is a numpy.flatiter instance, which acts similarly to, but is not a subclass of, Python’s builtin iterator object.
Examples
>>> x = np.arange(1, 7).reshape(2, 3) >>> x array([[1, 2, 3], [4, 5, 6]]) >>> x.flat[3] 4 >>> x.T array([[1, 4], [2, 5], [3, 6]]) >>> x.T.flat[3] 5 >>> type(x.flat) <type 'numpy.flatiter'>
An assignment example:
>>> x.flat = 3; x array([[3, 3, 3], [3, 3, 3]]) >>> x.flat[[1,4]] = 1; x array([[3, 1, 3], [3, 1, 3]])

flatten
(order='C') Return a copy of the array collapsed into one dimension.
Parameters: order : {‘C’, ‘F’, ‘A’, ‘K’}, optional
‘C’ means to flatten in rowmajor (Cstyle) order. ‘F’ means to flatten in columnmajor (Fortran style) order. ‘A’ means to flatten in columnmajor order if a is Fortran contiguous in memory, rowmajor order otherwise. ‘K’ means to flatten a in the order the elements occur in memory. The default is ‘C’.
Returns: y : ndarray
A copy of the input array, flattened to one dimension.
Examples
>>> a = np.array([[1,2], [3,4]]) >>> a.flatten() array([1, 2, 3, 4]) >>> a.flatten('F') array([1, 3, 2, 4])

getfield
(dtype, offset=0) Returns a field of the given array as a certain type.
A field is a view of the array data with a given datatype. The values in the view are determined by the given type and the offset into the current array in bytes. The offset needs to be such that the view dtype fits in the array dtype; for example an array of dtype complex128 has 16byte elements. If taking a view with a 32bit integer (4 bytes), the offset needs to be between 0 and 12 bytes.
Parameters: dtype : str or dtype
The data type of the view. The dtype size of the view can not be larger than that of the array itself.
offset : int
Number of bytes to skip before beginning the element view.
Examples
>>> x = np.diag([1.+1.j]*2) >>> x[1, 1] = 2 + 4.j >>> x array([[ 1.+1.j, 0.+0.j], [ 0.+0.j, 2.+4.j]]) >>> x.getfield(np.float64) array([[ 1., 0.], [ 0., 2.]])
By choosing an offset of 8 bytes we can select the complex part of the array for our view:
>>> x.getfield(np.float64, offset=8) array([[ 1., 0.], [ 0., 4.]])

imag
The imaginary part of the array.
Examples
>>> x = np.sqrt([1+0j, 0+1j]) >>> x.imag array([ 0. , 0.70710678]) >>> x.imag.dtype dtype('float64')

item
(*args) Copy an element of an array to a standard Python scalar and return it.
Parameters: *args : Arguments (variable number and type)
 none: in this case, the method only works for arrays with one element (a.size == 1), which element is copied into a standard Python scalar object and returned.
 int_type: this argument is interpreted as a flat index into the array, specifying which element to copy and return.
 tuple of int_types: functions as does a single int_type argument, except that the argument is interpreted as an ndindex into the array.
Returns: z : Standard Python scalar object
A copy of the specified element of the array as a suitable Python scalar
Notes
When the data type of a is longdouble or clongdouble, item() returns a scalar array object because there is no available Python scalar that would not lose information. Void arrays return a buffer object for item(), unless fields are defined, in which case a tuple is returned.
item is very similar to a[args], except, instead of an array scalar, a standard Python scalar is returned. This can be useful for speeding up access to elements of the array and doing arithmetic on elements of the array using Python’s optimized math.
Examples
>>> x = np.random.randint(9, size=(3, 3)) >>> x array([[3, 1, 7], [2, 8, 3], [8, 5, 3]]) >>> x.item(3) 2 >>> x.item(7) 5 >>> x.item((0, 1)) 1 >>> x.item((2, 2)) 3

itemset
(*args) Insert scalar into an array (scalar is cast to array’s dtype, if possible)
There must be at least 1 argument, and define the last argument as item. Then,
a.itemset(*args)
is equivalent to but faster thana[args] = item
. The item should be a scalar value and args must select a single item in the array a.Parameters: *args : Arguments
If one argument: a scalar, only used in case a is of size 1. If two arguments: the last argument is the value to be set and must be a scalar, the first argument specifies a single array element location. It is either an int or a tuple.
Notes
Compared to indexing syntax, itemset provides some speed increase for placing a scalar into a particular location in an ndarray, if you must do this. However, generally this is discouraged: among other problems, it complicates the appearance of the code. Also, when using itemset (and item) inside a loop, be sure to assign the methods to a local variable to avoid the attribute lookup at each loop iteration.
Examples
>>> x = np.random.randint(9, size=(3, 3)) >>> x array([[3, 1, 7], [2, 8, 3], [8, 5, 3]]) >>> x.itemset(4, 0) >>> x.itemset((2, 2), 9) >>> x array([[3, 1, 7], [2, 0, 3], [8, 5, 9]])

itemsize
Length of one array element in bytes.
Examples
>>> x = np.array([1,2,3], dtype=np.float64) >>> x.itemsize 8 >>> x = np.array([1,2,3], dtype=np.complex128) >>> x.itemsize 16

max
(axis=None, out=None) Return the maximum along a given axis.
Refer to numpy.amax for full documentation.
See also
numpy.amax
 equivalent function

mean
(axis=None, dtype=None, out=None, keepdims=False) Returns the average of the array elements along given axis.
Refer to numpy.mean for full documentation.
See also
numpy.mean
 equivalent function

min
(axis=None, out=None, keepdims=False) Return the minimum along a given axis.
Refer to numpy.amin for full documentation.
See also
numpy.amin
 equivalent function

nbytes
Total bytes consumed by the elements of the array.
Notes
Does not include memory consumed by nonelement attributes of the array object.
Examples
>>> x = np.zeros((3,5,2), dtype=np.complex128) >>> x.nbytes 480 >>> np.prod(x.shape) * x.itemsize 480

ndim
Number of array dimensions.
Examples
>>> x = np.array([1, 2, 3]) >>> x.ndim 1 >>> y = np.zeros((2, 3, 4)) >>> y.ndim 3

newbyteorder
(new_order='S') Return the array with the same data viewed with a different byte order.
Equivalent to:
arr.view(arr.dtype.newbytorder(new_order))
Changes are also made in all fields and subarrays of the array data type.
Parameters: new_order : string, optional
Byte order to force; a value from the byte order specifications below. new_order codes can be any of:
 ‘S’  swap dtype from current to opposite endian
 {‘<’, ‘L’}  little endian
 {‘>’, ‘B’}  big endian
 {‘=’, ‘N’}  native order
 {‘’, ‘I’}  ignore (no change to byte order)
The default value (‘S’) results in swapping the current byte order. The code does a caseinsensitive check on the first letter of new_order for the alternatives above. For example, any of ‘B’ or ‘b’ or ‘biggish’ are valid to specify bigendian.
Returns: new_arr : array
New array object with the dtype reflecting given change to the byte order.

nonzero
() Return the indices of the elements that are nonzero.
Refer to numpy.nonzero for full documentation.
See also
numpy.nonzero
 equivalent function

partition
(kth, axis=1, kind='introselect', order=None) Rearranges the elements in the array in such a way that value of the element in kth position is in the position it would be in a sorted array. All elements smaller than the kth element are moved before this element and all equal or greater are moved behind it. The ordering of the elements in the two partitions is undefined.
New in version 1.8.0.
Parameters: kth : int or sequence of ints
Element index to partition by. The kth element value will be in its final sorted position and all smaller elements will be moved before it and all equal or greater elements behind it. The order all elements in the partitions is undefined. If provided with a sequence of kth it will partition all elements indexed by kth of them into their sorted position at once.
axis : int, optional
Axis along which to sort. Default is 1, which means sort along the last axis.
kind : {‘introselect’}, optional
Selection algorithm. Default is ‘introselect’.
order : str or list of str, optional
When a is an array with fields defined, this argument specifies which fields to compare first, second, etc. A single field can be specified as a string, and not all fields need be specified, but unspecified fields will still be used, in the order in which they come up in the dtype, to break ties.
See also
numpy.partition
 Return a parititioned copy of an array.
argpartition
 Indirect partition.
sort
 Full sort.
Notes
See
np.partition
for notes on the different algorithms.Examples
>>> a = np.array([3, 4, 2, 1]) >>> a.partition(3) >>> a array([2, 1, 3, 4])
>>> a.partition((1, 3)) array([1, 2, 3, 4])

prod
(axis=None, dtype=None, out=None, keepdims=False) Return the product of the array elements over the given axis
Refer to numpy.prod for full documentation.
See also
numpy.prod
 equivalent function

ptp
(axis=None, out=None) Peak to peak (maximum  minimum) value along a given axis.
Refer to numpy.ptp for full documentation.
See also
numpy.ptp
 equivalent function

put
(indices, values, mode='raise') Set
a.flat[n] = values[n]
for all n in indices.Refer to numpy.put for full documentation.
See also
numpy.put
 equivalent function

ravel
([order]) Return a flattened array.
Refer to numpy.ravel for full documentation.
See also
numpy.ravel
 equivalent function
ndarray.flat
 a flat iterator on the array.

real
The real part of the array.
See also
numpy.real
 equivalent function
Examples
>>> x = np.sqrt([1+0j, 0+1j]) >>> x.real array([ 1. , 0.70710678]) >>> x.real.dtype dtype('float64')

repeat
(repeats, axis=None) Repeat elements of an array.
Refer to numpy.repeat for full documentation.
See also
numpy.repeat
 equivalent function

reshape
(shape, order='C') Returns an array containing the same data with a new shape.
Refer to numpy.reshape for full documentation.
See also
numpy.reshape
 equivalent function

resize
(new_shape, refcheck=True) Change shape and size of array inplace.
Parameters: new_shape : tuple of ints, or n ints
Shape of resized array.
refcheck : bool, optional
If False, reference count will not be checked. Default is True.
Returns: None
Raises: ValueError
If a does not own its own data or references or views to it exist, and the data memory must be changed. PyPy only: will always raise if the data memory must be changed, since there is no reliable way to determine if references or views to it exist.
SystemError
If the order keyword argument is specified. This behaviour is a bug in NumPy.
See also
resize
 Return a new array with the specified shape.
Notes
This reallocates space for the data area if necessary.
Only contiguous arrays (data elements consecutive in memory) can be resized.
The purpose of the reference count check is to make sure you do not use this array as a buffer for another Python object and then reallocate the memory. However, reference counts can increase in other ways so if you are sure that you have not shared the memory for this array with another Python object, then you may safely set refcheck to False.
Examples
Shrinking an array: array is flattened (in the order that the data are stored in memory), resized, and reshaped:
>>> a = np.array([[0, 1], [2, 3]], order='C') >>> a.resize((2, 1)) >>> a array([[0], [1]])
>>> a = np.array([[0, 1], [2, 3]], order='F') >>> a.resize((2, 1)) >>> a array([[0], [2]])
Enlarging an array: as above, but missing entries are filled with zeros:
>>> b = np.array([[0, 1], [2, 3]]) >>> b.resize(2, 3) # new_shape parameter doesn't have to be a tuple >>> b array([[0, 1, 2], [3, 0, 0]])
Referencing an array prevents resizing…
>>> c = a >>> a.resize((1, 1)) Traceback (most recent call last): ... ValueError: cannot resize an array that has been referenced ...
Unless refcheck is False:
>>> a.resize((1, 1), refcheck=False) >>> a array([[0]]) >>> c array([[0]])

round
(decimals=0, out=None) Return a with each element rounded to the given number of decimals.
Refer to numpy.around for full documentation.
See also
numpy.around
 equivalent function

searchsorted
(v, side='left', sorter=None) Find indices where elements of v should be inserted in a to maintain order.
For full documentation, see numpy.searchsorted
See also
numpy.searchsorted
 equivalent function

setfield
(val, dtype, offset=0) Put a value into a specified place in a field defined by a datatype.
Place val into a’s field defined by dtype and beginning offset bytes into the field.
Parameters: val : object
Value to be placed in field.
dtype : dtype object
Datatype of the field in which to place val.
offset : int, optional
The number of bytes into the field at which to place val.
Returns: None
See also
Examples
>>> x = np.eye(3) >>> x.getfield(np.float64) array([[ 1., 0., 0.], [ 0., 1., 0.], [ 0., 0., 1.]]) >>> x.setfield(3, np.int32) >>> x.getfield(np.int32) array([[3, 3, 3], [3, 3, 3], [3, 3, 3]]) >>> x array([[ 1.00000000e+000, 1.48219694e323, 1.48219694e323], [ 1.48219694e323, 1.00000000e+000, 1.48219694e323], [ 1.48219694e323, 1.48219694e323, 1.00000000e+000]]) >>> x.setfield(np.eye(3), np.int32) >>> x array([[ 1., 0., 0.], [ 0., 1., 0.], [ 0., 0., 1.]])

setflags
(write=None, align=None, uic=None) Set array flags WRITEABLE, ALIGNED, and UPDATEIFCOPY, respectively.
These Booleanvalued flags affect how numpy interprets the memory area used by a (see Notes below). The ALIGNED flag can only be set to True if the data is actually aligned according to the type. The UPDATEIFCOPY flag can never be set to True. The flag WRITEABLE can only be set to True if the array owns its own memory, or the ultimate owner of the memory exposes a writeable buffer interface, or is a string. (The exception for string is made so that unpickling can be done without copying memory.)
Parameters: write : bool, optional
Describes whether or not a can be written to.
align : bool, optional
Describes whether or not a is aligned properly for its type.
uic : bool, optional
Describes whether or not a is a copy of another “base” array.
Notes
Array flags provide information about how the memory area used for the array is to be interpreted. There are 6 Boolean flags in use, only three of which can be changed by the user: UPDATEIFCOPY, WRITEABLE, and ALIGNED.
WRITEABLE (W) the data area can be written to;
ALIGNED (A) the data and strides are aligned appropriately for the hardware (as determined by the compiler);
UPDATEIFCOPY (U) this array is a copy of some other array (referenced by .base). When this array is deallocated, the base array will be updated with the contents of this array.
All flags can be accessed using their first (upper case) letter as well as the full name.
Examples
>>> y array([[3, 1, 7], [2, 0, 0], [8, 5, 9]]) >>> y.flags C_CONTIGUOUS : True F_CONTIGUOUS : False OWNDATA : True WRITEABLE : True ALIGNED : True UPDATEIFCOPY : False >>> y.setflags(write=0, align=0) >>> y.flags C_CONTIGUOUS : True F_CONTIGUOUS : False OWNDATA : True WRITEABLE : False ALIGNED : False UPDATEIFCOPY : False >>> y.setflags(uic=1) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: cannot set UPDATEIFCOPY flag to True

shape
Tuple of array dimensions.
Notes
May be used to “reshape” the array, as long as this would not require a change in the total number of elements
Examples
>>> x = np.array([1, 2, 3, 4]) >>> x.shape (4,) >>> y = np.zeros((2, 3, 4)) >>> y.shape (2, 3, 4) >>> y.shape = (3, 8) >>> y array([[ 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 0.]]) >>> y.shape = (3, 6) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: total size of new array must be unchanged

size
Number of elements in the array.
Equivalent to
np.prod(a.shape)
, i.e., the product of the array’s dimensions.Examples
>>> x = np.zeros((3, 5, 2), dtype=np.complex128) >>> x.size 30 >>> np.prod(x.shape) 30

sort
(axis=1, kind='quicksort', order=None) Sort an array, inplace.
Parameters: axis : int, optional
Axis along which to sort. Default is 1, which means sort along the last axis.
kind : {‘quicksort’, ‘mergesort’, ‘heapsort’}, optional
Sorting algorithm. Default is ‘quicksort’.
order : str or list of str, optional
When a is an array with fields defined, this argument specifies which fields to compare first, second, etc. A single field can be specified as a string, and not all fields need be specified, but unspecified fields will still be used, in the order in which they come up in the dtype, to break ties.
See also
numpy.sort
 Return a sorted copy of an array.
argsort
 Indirect sort.
lexsort
 Indirect stable sort on multiple keys.
searchsorted
 Find elements in sorted array.
partition
 Partial sort.
Notes
See
sort
for notes on the different sorting algorithms.Examples
>>> a = np.array([[1,4], [3,1]]) >>> a.sort(axis=1) >>> a array([[1, 4], [1, 3]]) >>> a.sort(axis=0) >>> a array([[1, 3], [1, 4]])
Use the order keyword to specify a field to use when sorting a structured array:
>>> a = np.array([('a', 2), ('c', 1)], dtype=[('x', 'S1'), ('y', int)]) >>> a.sort(order='y') >>> a array([('c', 1), ('a', 2)], dtype=[('x', 'S1'), ('y', '<i4')])

squeeze
(axis=None) Remove singledimensional entries from the shape of a.
Refer to numpy.squeeze for full documentation.
See also
numpy.squeeze
 equivalent function

std
(axis=None, dtype=None, out=None, ddof=0, keepdims=False) Returns the standard deviation of the array elements along given axis.
Refer to numpy.std for full documentation.
See also
numpy.std
 equivalent function

strides
Tuple of bytes to step in each dimension when traversing an array.
The byte offset of element
(i[0], i[1], ..., i[n])
in an array a is:offset = sum(np.array(i) * a.strides)
A more detailed explanation of strides can be found in the “ndarray.rst” file in the NumPy reference guide.
See also
numpy.lib.stride_tricks.as_strided
Notes
Imagine an array of 32bit integers (each 4 bytes):
x = np.array([[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]], dtype=np.int32)
This array is stored in memory as 40 bytes, one after the other (known as a contiguous block of memory). The strides of an array tell us how many bytes we have to skip in memory to move to the next position along a certain axis. For example, we have to skip 4 bytes (1 value) to move to the next column, but 20 bytes (5 values) to get to the same position in the next row. As such, the strides for the array x will be
(20, 4)
.Examples
>>> y = np.reshape(np.arange(2*3*4), (2,3,4)) >>> y array([[[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]], [[12, 13, 14, 15], [16, 17, 18, 19], [20, 21, 22, 23]]]) >>> y.strides (48, 16, 4) >>> y[1,1,1] 17 >>> offset=sum(y.strides * np.array((1,1,1))) >>> offset/y.itemsize 17
>>> x = np.reshape(np.arange(5*6*7*8), (5,6,7,8)).transpose(2,3,1,0) >>> x.strides (32, 4, 224, 1344) >>> i = np.array([3,5,2,2]) >>> offset = sum(i * x.strides) >>> x[3,5,2,2] 813 >>> offset / x.itemsize 813

sum
(axis=None, dtype=None, out=None, keepdims=False) Return the sum of the array elements over the given axis.
Refer to numpy.sum for full documentation.
See also
numpy.sum
 equivalent function

swapaxes
(axis1, axis2) Return a view of the array with axis1 and axis2 interchanged.
Refer to numpy.swapaxes for full documentation.
See also
numpy.swapaxes
 equivalent function

take
(indices, axis=None, out=None, mode='raise') Return an array formed from the elements of a at the given indices.
Refer to numpy.take for full documentation.
See also
numpy.take
 equivalent function

tobytes
(order='C') Construct Python bytes containing the raw data bytes in the array.
Constructs Python bytes showing a copy of the raw contents of data memory. The bytes object can be produced in either ‘C’ or ‘Fortran’, or ‘Any’ order (the default is ‘C’order). ‘Any’ order means Corder unless the F_CONTIGUOUS flag in the array is set, in which case it means ‘Fortran’ order.
New in version 1.9.0.
Parameters: order : {‘C’, ‘F’, None}, optional
Order of the data for multidimensional arrays: C, Fortran, or the same as for the original array.
Returns: s : bytes
Python bytes exhibiting a copy of a’s raw data.
Examples
>>> x = np.array([[0, 1], [2, 3]]) >>> x.tobytes() b'\x00\x00\x00\x00\x01\x00\x00\x00\x02\x00\x00\x00\x03\x00\x00\x00' >>> x.tobytes('C') == x.tobytes() True >>> x.tobytes('F') b'\x00\x00\x00\x00\x02\x00\x00\x00\x01\x00\x00\x00\x03\x00\x00\x00'

tofile
(fid, sep="", format="%s") Write array to a file as text or binary (default).
Data is always written in ‘C’ order, independent of the order of a. The data produced by this method can be recovered using the function fromfile().
Parameters: fid : file or str
An open file object, or a string containing a filename.
sep : str
Separator between array items for text output. If “” (empty), a binary file is written, equivalent to
file.write(a.tobytes())
.format : str
Format string for text file output. Each entry in the array is formatted to text by first converting it to the closest Python type, and then using “format” % item.
Notes
This is a convenience function for quick storage of array data. Information on endianness and precision is lost, so this method is not a good choice for files intended to archive data or transport data between machines with different endianness. Some of these problems can be overcome by outputting the data as text files, at the expense of speed and file size.

tolist
() Return the array as a (possibly nested) list.
Return a copy of the array data as a (nested) Python list. Data items are converted to the nearest compatible Python type.
Parameters: none
Returns: y : list
The possibly nested list of array elements.
Notes
The array may be recreated,
a = np.array(a.tolist())
.Examples
>>> a = np.array([1, 2]) >>> a.tolist() [1, 2] >>> a = np.array([[1, 2], [3, 4]]) >>> list(a) [array([1, 2]), array([3, 4])] >>> a.tolist() [[1, 2], [3, 4]]

tostring
(order='C') Construct Python bytes containing the raw data bytes in the array.
Constructs Python bytes showing a copy of the raw contents of data memory. The bytes object can be produced in either ‘C’ or ‘Fortran’, or ‘Any’ order (the default is ‘C’order). ‘Any’ order means Corder unless the F_CONTIGUOUS flag in the array is set, in which case it means ‘Fortran’ order.
This function is a compatibility alias for tobytes. Despite its name it returns bytes not strings.
Parameters: order : {‘C’, ‘F’, None}, optional
Order of the data for multidimensional arrays: C, Fortran, or the same as for the original array.
Returns: s : bytes
Python bytes exhibiting a copy of a’s raw data.
Examples
>>> x = np.array([[0, 1], [2, 3]]) >>> x.tobytes() b'\x00\x00\x00\x00\x01\x00\x00\x00\x02\x00\x00\x00\x03\x00\x00\x00' >>> x.tobytes('C') == x.tobytes() True >>> x.tobytes('F') b'\x00\x00\x00\x00\x02\x00\x00\x00\x01\x00\x00\x00\x03\x00\x00\x00'

trace
(offset=0, axis1=0, axis2=1, dtype=None, out=None) Return the sum along diagonals of the array.
Refer to numpy.trace for full documentation.
See also
numpy.trace
 equivalent function

transpose
(*axes) Returns a view of the array with axes transposed.
For a 1D array, this has no effect. (To change between column and row vectors, first cast the 1D array into a matrix object.) For a 2D array, this is the usual matrix transpose. For an nD array, if axes are given, their order indicates how the axes are permuted (see Examples). If axes are not provided and
a.shape = (i[0], i[1], ... i[n2], i[n1])
, thena.transpose().shape = (i[n1], i[n2], ... i[1], i[0])
.Parameters: axes : None, tuple of ints, or n ints
 None or no argument: reverses the order of the axes.
 tuple of ints: i in the jth place in the tuple means a’s ith axis becomes a.transpose()’s jth axis.
 n ints: same as an ntuple of the same ints (this form is intended simply as a “convenience” alternative to the tuple form)
Returns: out : ndarray
View of a, with axes suitably permuted.
See also
ndarray.T
 Array property returning the array transposed.
Examples
>>> a = np.array([[1, 2], [3, 4]]) >>> a array([[1, 2], [3, 4]]) >>> a.transpose() array([[1, 3], [2, 4]]) >>> a.transpose((1, 0)) array([[1, 3], [2, 4]]) >>> a.transpose(1, 0) array([[1, 3], [2, 4]])

var
(axis=None, dtype=None, out=None, ddof=0, keepdims=False) Returns the variance of the array elements, along given axis.
Refer to numpy.var for full documentation.
See also
numpy.var
 equivalent function

view
(dtype=None, type=None) New view of array with the same data.
Parameters: dtype : datatype or ndarray subclass, optional
Datatype descriptor of the returned view, e.g., float32 or int16. The default, None, results in the view having the same datatype as a. This argument can also be specified as an ndarray subclass, which then specifies the type of the returned object (this is equivalent to setting the
type
parameter).type : Python type, optional
Type of the returned view, e.g., ndarray or matrix. Again, the default None results in type preservation.
Notes
a.view()
is used two different ways:a.view(some_dtype)
ora.view(dtype=some_dtype)
constructs a view of the array’s memory with a different datatype. This can cause a reinterpretation of the bytes of memory.a.view(ndarray_subclass)
ora.view(type=ndarray_subclass)
just returns an instance of ndarray_subclass that looks at the same array (same shape, dtype, etc.) This does not cause a reinterpretation of the memory.For
a.view(some_dtype)
, ifsome_dtype
has a different number of bytes per entry than the previous dtype (for example, converting a regular array to a structured array), then the behavior of the view cannot be predicted just from the superficial appearance ofa
(shown byprint(a)
). It also depends on exactly howa
is stored in memory. Therefore ifa
is Cordered versus fortranordered, versus defined as a slice or transpose, etc., the view may give different results.Examples
>>> x = np.array([(1, 2)], dtype=[('a', np.int8), ('b', np.int8)])
Viewing array data using a different type and dtype:
>>> y = x.view(dtype=np.int16, type=np.matrix) >>> y matrix([[513]], dtype=int16) >>> print(type(y)) <class 'numpy.matrixlib.defmatrix.matrix'>
Creating a view on a structured array so it can be used in calculations
>>> x = np.array([(1, 2),(3,4)], dtype=[('a', np.int8), ('b', np.int8)]) >>> xv = x.view(dtype=np.int8).reshape(1,2) >>> xv array([[1, 2], [3, 4]], dtype=int8) >>> xv.mean(0) array([ 2., 3.])
Making changes to the view changes the underlying array
>>> xv[0,1] = 20 >>> print(x) [(1, 20) (3, 4)]
Using a view to convert an array to a recarray:
>>> z = x.view(np.recarray) >>> z.a array([1], dtype=int8)
Views share data:
>>> x[0] = (9, 10) >>> z[0] (9, 10)
Views that change the dtype size (bytes per entry) should normally be avoided on arrays defined by slices, transposes, fortranordering, etc.:
>>> x = np.array([[1,2,3],[4,5,6]], dtype=np.int16) >>> y = x[:, 0:2] >>> y array([[1, 2], [4, 5]], dtype=int16) >>> y.view(dtype=[('width', np.int16), ('length', np.int16)]) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: new type not compatible with array. >>> z = y.copy() >>> z.view(dtype=[('width', np.int16), ('length', np.int16)]) array([[(1, 2)], [(4, 5)]], dtype=[('width', '<i2'), ('length', '<i2')])

class
multimodal.datasets.data_sample.
MultiModalSparseArray
(*arg, **kwargs) MultiModalArray inherit from numpy ndarray
Parameters: data : can be
 dictionary of multiview array with shape = (n_samples, n_features) for multiview
for each view.
 {0: array([[]],
1: array([[]], …}
 numpy array like with shape = (n_samples, n_features) for multiview
for each view.
 [[[…]],
[[…]], …]
 {array like} with (n_samples, nviews * n_features) with ‘views_ind’ diferent to ‘None’
for Multiview input samples.
 views_ind : arraylike (default= None ) if None
[0, n_features//2, n_features]) is constructed (2 views) Paramater specifying how to extract the data views from X:
 views_ind is a 1D array of sorted integers, the entries
indicate the limits of the slices used to extract the views,
where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.
 views_ind is a 1D array of sorted integers, the entries
indicate the limits of the slices used to extract the views,
where view
Attributes
views_ind (list of views’ indice (may be None)) n_views : int number of views shapes_int: list of int numbers of feature for each views keys : name of key, where data come from a dictionary :Example: >>> from multimodal.datasets.base import load_dict >>> from multimodal.tests.datasets.get_dataset_path import get_dataset_path >>> from multimodal.datasets.data_sample import DataSample >>> file = ‘input_x_dic.pkl’ >>> data = load_dict(get_dataset_path(file)) 
arcsin
() Elementwise arcsin.
See numpy.arcsin for more information.

arcsinh
() Elementwise arcsinh.
See numpy.arcsinh for more information.

arctan
() Elementwise arctan.
See numpy.arctan for more information.

arctanh
() Elementwise arctanh.
See numpy.arctanh for more information.

argmax
(axis=None, out=None) Return indices of minimum elements along an axis.
Implicit zero elements are also taken into account. If there are several maximum values, the index of the first occurrence is returned.
Parameters: axis : {2, 1, 0, 1, None}, optional
Axis along which the argmax is computed. If None (default), index of the maximum element in the flatten data is returned.
out : None, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value, as this argument is not used.
Returns: ind : np.matrix or int
Indices of maximum elements. If matrix, its size along axis is 1.

argmin
(axis=None, out=None) Return indices of minimum elements along an axis.
Implicit zero elements are also taken into account. If there are several minimum values, the index of the first occurrence is returned.
Parameters: axis : {2, 1, 0, 1, None}, optional
Axis along which the argmin is computed. If None (default), index of the minimum element in the flatten data is returned.
out : None, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value, as this argument is not used.
Returns: ind : np.matrix or int
Indices of minimum elements. If matrix, its size along axis is 1.

asformat
(format) Return this matrix in a given sparse format
Parameters: format : {string, None}
 desired sparse matrix format
 None for no format conversion
 “csr” for csr_matrix format
 “csc” for csc_matrix format
 “lil” for lil_matrix format
 “dok” for dok_matrix format and so on

asfptype
() Upcast matrix to a floating point format (if necessary)

astype
(t) Cast the matrix elements to a specified type.
The data will be copied.
Parameters: t : string or numpy dtype
Typecode or datatype to which to cast the data.

ceil
() Elementwise ceil.
See numpy.ceil for more information.

check_format
(full_check=True) check whether the matrix format is valid
Parameters: full_check : bool, optional
If True, rigorous check, O(N) operations. Otherwise basic check, O(1) operations (default True).

conj
() Elementwise complex conjugation.
If the matrix is of noncomplex data type, then this method does nothing and the data is not copied.

conjugate
() Elementwise complex conjugation.
If the matrix is of noncomplex data type, then this method does nothing and the data is not copied.

copy
() Returns a copy of this matrix.
No data/indices will be shared between the returned value and current matrix.

count_nonzero
() Number of nonzero entries, equivalent to
np.count_nonzero(a.toarray())
Unlike getnnz() and the nnz property, which return the number of stored entries (the length of the data attribute), this method counts the actual number of nonzero entries in data.

deg2rad
() Elementwise deg2rad.
See numpy.deg2rad for more information.

diagonal
() Returns the main diagonal of the matrix

dot
(other) Ordinary dot product
Examples
>>> import numpy as np >>> from scipy.sparse import csr_matrix >>> A = csr_matrix([[1, 2, 0], [0, 0, 3], [4, 0, 5]]) >>> v = np.array([1, 0, 1]) >>> A.dot(v) array([ 1, 3, 1], dtype=int64)

eliminate_zeros
() Remove zero entries from the matrix
This is an in place operation

expm1
() Elementwise expm1.
See numpy.expm1 for more information.

floor
() Elementwise floor.
See numpy.floor for more information.

getH
() Return the Hermitian transpose of this matrix.
See also
np.matrix.getH
 NumPy’s implementation of getH for matrices

get_shape
() Get shape of a matrix.

getcol
(i) Returns a copy of column i of the matrix, as a (m x 1) CSR matrix (column vector).

getformat
() Format of a matrix representation as a string.

getmaxprint
() Maximum number of elements to display when printed.

getnnz
(axis=None) Number of stored values, including explicit zeros.
Parameters: axis : None, 0, or 1
Select between the number of values across the whole matrix, in each column, or in each row.
See also
count_nonzero
 Number of nonzero entries

getrow
(i) Returns a copy of row i of the matrix, as a (1 x n) CSR matrix (row vector).

has_canonical_format
Determine whether the matrix has sorted indices and no duplicates
 Returns
 True: if the above applies
 False: otherwise
has_canonical_format implies has_sorted_indices, so if the latter flag is False, so will the former be; if the former is found True, the latter flag is also set.

has_sorted_indices
Determine whether the matrix has sorted indices
 Returns
 True: if the indices of the matrix are in sorted order
 False: otherwise

log1p
() Elementwise log1p.
See numpy.log1p for more information.

max
(axis=None, out=None) Return the maximum of the matrix or maximum along an axis. This takes all elements into account, not just the nonzero ones.
Parameters: axis : {2, 1, 0, 1, None} optional
Axis along which the sum is computed. The default is to compute the maximum over all the matrix elements, returning a scalar (i.e. axis = None).
out : None, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value, as this argument is not used.
Returns: amax : coo_matrix or scalar
Maximum of a. If axis is None, the result is a scalar value. If axis is given, the result is a sparse.coo_matrix of dimension
a.ndim  1
.See also
min
 The minimum value of a sparse matrix along a given axis.
np.matrix.max
 NumPy’s implementation of ‘max’ for matrices

maximum
(other) Elementwise maximum between this and another matrix.

mean
(axis=None, dtype=None, out=None) Compute the arithmetic mean along the specified axis.
Returns the average of the matrix elements. The average is taken over all elements in the matrix by default, otherwise over the specified axis. float64 intermediate and return values are used for integer inputs.
Parameters: axis : {2, 1, 0, 1, None} optional
Axis along which the mean is computed. The default is to compute the mean of all elements in the matrix (i.e. axis = None).
dtype : datatype, optional
Type to use in computing the mean. For integer inputs, the default is float64; for floating point inputs, it is the same as the input dtype.
out : np.matrix, optional
Alternative output matrix in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.
Returns: m : np.matrix
See also
np.matrix.mean
 NumPy’s implementation of ‘mean’ for matrices

min
(axis=None, out=None) Return the minimum of the matrix or maximum along an axis. This takes all elements into account, not just the nonzero ones.
Parameters: axis : {2, 1, 0, 1, None} optional
Axis along which the sum is computed. The default is to compute the minimum over all the matrix elements, returning a scalar (i.e. axis = None).
out : None, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value, as this argument is not used.
Returns: amin : coo_matrix or scalar
Minimum of a. If axis is None, the result is a scalar value. If axis is given, the result is a sparse.coo_matrix of dimension
a.ndim  1
.See also
max
 The maximum value of a sparse matrix along a given axis.
np.matrix.min
 NumPy’s implementation of ‘min’ for matrices

minimum
(other) Elementwise minimum between this and another matrix.

multiply
(other) Pointwise multiplication by another matrix, vector, or scalar.

nnz
Number of stored values, including explicit zeros.
See also
count_nonzero
 Number of nonzero entries

nonzero
() nonzero indices
Returns a tuple of arrays (row,col) containing the indices of the nonzero elements of the matrix.
Examples
>>> from scipy.sparse import csr_matrix >>> A = csr_matrix([[1,2,0],[0,0,3],[4,0,5]]) >>> A.nonzero() (array([0, 0, 1, 2, 2]), array([0, 1, 2, 0, 2]))

power
(n, dtype=None) This function performs elementwise power.
Parameters: n : n is a scalar
dtype : If dtype is not specified, the current dtype will be preserved.

prune
() Remove empty space after all nonzero elements.

rad2deg
() Elementwise rad2deg.
See numpy.rad2deg for more information.

reshape
(shape, order='C') Gives a new shape to a sparse matrix without changing its data.
Parameters: shape : length2 tuple of ints
The new shape should be compatible with the original shape.
order : ‘C’, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value, as this argument is not used.
Returns: reshaped_matrix : self with the new dimensions of shape
See also
np.matrix.reshape
 NumPy’s implementation of ‘reshape’ for matrices

rint
() Elementwise rint.
See numpy.rint for more information.

set_shape
(shape) See reshape.

setdiag
(values, k=0) Set diagonal or offdiagonal elements of the array.
Parameters: values : array_like
New values of the diagonal elements.
Values may have any length. If the diagonal is longer than values, then the remaining diagonal entries will not be set. If values if longer than the diagonal, then the remaining values are ignored.
If a scalar value is given, all of the diagonal is set to it.
k : int, optional
Which offdiagonal to set, corresponding to elements a[i,i+k]. Default: 0 (the main diagonal).

shape
Get shape of a matrix.

sign
() Elementwise sign.
See numpy.sign for more information.

sin
() Elementwise sin.
See numpy.sin for more information.

sinh
() Elementwise sinh.
See numpy.sinh for more information.

sort_indices
() Sort the indices of this matrix in place

sorted_indices
() Return a copy of this matrix with sorted indices

sqrt
() Elementwise sqrt.
See numpy.sqrt for more information.

sum
(axis=None, dtype=None, out=None) Sum the matrix elements over a given axis.
Parameters: axis : {2, 1, 0, 1, None} optional
Axis along which the sum is computed. The default is to compute the sum of all the matrix elements, returning a scalar (i.e. axis = None).
dtype : dtype, optional
The type of the returned matrix and of the accumulator in which the elements are summed. The dtype of a is used by default unless a has an integer dtype of less precision than the default platform integer. In that case, if a is signed then the platform integer is used while if a is unsigned then an unsigned integer of the same precision as the platform integer is used.
out : np.matrix, optional
Alternative output matrix in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.
Returns: sum_along_axis : np.matrix
A matrix with the same shape as self, with the specified axis removed.
See also
np.matrix.sum
 NumPy’s implementation of ‘sum’ for matrices

sum_duplicates
() Eliminate duplicate matrix entries by adding them together
The is an in place operation

tan
() Elementwise tan.
See numpy.tan for more information.

tanh
() Elementwise tanh.
See numpy.tanh for more information.

toarray
(order=None, out=None) See the docstring for spmatrix.toarray.

tobsr
(blocksize=None, copy=True) Convert this matrix to Block Sparse Row format.
With copy=False, the data/indices may be shared between this matrix and the resultant bsr_matrix.
When blocksize=(R, C) is provided, it will be used for construction of the bsr_matrix.

tocoo
(copy=True) Convert this matrix to COOrdinate format.
With copy=False, the data/indices may be shared between this matrix and the resultant coo_matrix.

tocsc
(copy=False) Convert this matrix to Compressed Sparse Column format.
With copy=False, the data/indices may be shared between this matrix and the resultant csc_matrix.

tocsr
(copy=False) Convert this matrix to Compressed Sparse Row format.
With copy=False, the data/indices may be shared between this matrix and the resultant csr_matrix.

todense
(order=None, out=None) Return a dense matrix representation of this matrix.
Parameters: order : {‘C’, ‘F’}, optional
Whether to store multidimensional data in C (rowmajor) or Fortran (columnmajor) order in memory. The default is ‘None’, indicating the NumPy default of Cordered. Cannot be specified in conjunction with the out argument.
out : ndarray, 2dimensional, optional
If specified, uses this array (or numpy.matrix) as the output buffer instead of allocating a new array to return. The provided array must have the same shape and dtype as the sparse matrix on which you are calling the method.
Returns: arr : numpy.matrix, 2dimensional
A NumPy matrix object with the same shape and containing the same data represented by the sparse matrix, with the requested memory order. If out was passed and was an array (rather than a numpy.matrix), it will be filled with the appropriate values and returned wrapped in a numpy.matrix object that shares the same memory.

todia
(copy=False) Convert this matrix to sparse DIAgonal format.
With copy=False, the data/indices may be shared between this matrix and the resultant dia_matrix.

todok
(copy=False) Convert this matrix to Dictionary Of Keys format.
With copy=False, the data/indices may be shared between this matrix and the resultant dok_matrix.

tolil
(copy=False) Convert this matrix to LInked List format.
With copy=False, the data/indices may be shared between this matrix and the resultant lil_matrix.

transpose
(axes=None, copy=False) Reverses the dimensions of the sparse matrix.
Parameters: axes : None, optional
This argument is in the signature solely for NumPy compatibility reasons. Do not pass in anything except for the default value.
copy : bool, optional
Indicates whether or not attributes of self should be copied whenever possible. The degree to which attributes are copied varies depending on the type of sparse matrix being used.
Returns: p : self with the dimensions reversed.
See also
np.matrix.transpose
 NumPy’s implementation of ‘transpose’ for matrices

trunc
() Elementwise trunc.
See numpy.trunc for more information.
Boosting
multimodal.boosting.mumbo
Multimodal Boosting
This module contains a MultiModal Boosting (MuMBo)
estimator for classification implemented in the MumboClassifier
class.

class
multimodal.boosting.mumbo.
MumboClassifier
(base_estimator=None, n_estimators=50, random_state=None, best_view_mode='edge') It then iterates the process on the same dataset but where the weights of incorrectly classified instances are adjusted such that subsequent classifiers focus more on difficult cases. A MuMBo classifier.
A MuMBo classifier is a metaestimator that implements a multimodal (or multiview) boosting algorithm:
It fits a set of classifiers on the original dataset splitted into several views and retains the classifier obtained for the best view.
This class implements the MuMBo algorithm [R11].
Parameters: base_estimator : object, optional (default=DecisionTreeClassifier)
Base estimator from which the boosted ensemble is built. Support for sample weighting is required, as well as proper classes_ and n_classes_ attributes. The default is a DecisionTreeClassifier with parameter
max_depth=1
.n_estimators : integer, optional (default=50)
Maximum number of estimators at which boosting is terminated.
random_state : int, RandomState instance or None, optional (default=None)
If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random.
best_view_mode : {“edge”, “error”}, optional (default=”edge”)
Mode used to select the best view at each iteration:
 if
best_view_mode == "edge"
, the best view is the view maximizing the edge value (variable δ (delta) in [R11]),  if
best_view_mode == "error"
, the best view is the view minimizing the classification error.
See also
sklearn.ensemble.AdaBoostClassifier
,sklearn.ensemble.GradientBoostingClassifier
,sklearn.tree.DecisionTreeClassifier
References
[R11] (1, 2, 3) Sokol Koço, “Tackling the uneven views problem with cooperation based ensemble learning methods”, PhD Thesis, AixMarseille Université, 2013, http://www.theses.fr/en/2013AIXM4101. Examples
>>> from multimodal.boosting.mumbo import MumboClassifier >>> from sklearn.datasets import load_iris >>> X, y = load_iris(return_X_y=True) >>> views_ind = [0, 2, 4] # view 0: sepal data, view 1: petal data >>> clf = MumboClassifier(random_state=0) >>> clf.fit(X, y, views_ind) MumboClassifier(random_state=0) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [1] >>> views_ind = [[0, 2], [1, 3]] # view 0: length data, view 1: width data >>> clf = MumboClassifier(random_state=0) >>> clf.fit(X, y, views_ind) MumboClassifier(random_state=0) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [1]
>>> from sklearn.tree import DecisionTreeClassifier >>> base_estimator = DecisionTreeClassifier(max_depth=2) >>> clf = MumboClassifier(base_estimator=base_estimator, random_state=0) >>> clf.fit(X, y, views_ind) MumboClassifier(base_estimator=DecisionTreeClassifier(max_depth=2), random_state=0) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [1]
Attributes
estimators_ (list of classifiers) Collection of fitted subestimators. classes_ (numpy.ndarray, shape = (n_classes,)) Classes labels. n_classes_ (int) Number of classes. estimator_weights_ (numpy.ndarray of floats, shape = (len(estimators_),)) Weights for each estimator in the boosted ensemble. estimator_errors_ (array of floats) Empirical loss for each iteration. best_views_ (numpy.ndarray of integers, shape = (len(estimators_),)) Indices of the best view for each estimator in the boosted ensemble. 
decision_function
(X) Compute the decision function of X.
Parameters: X : { arraylike, sparse matrix},
shape = (n_samples, n_views * n_features) Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR. maybe also MultimodalData
Returns: dec_fun : numpy.ndarray, shape = (n_samples, k)
Decision function of the input samples. The order of outputs is the same of that of the classes_ attribute. Binary classification is a special cases with
k == 1
, otherwisek == n_classes
. For binary classification, values <=0 mean classification in the first class inclasses_
and values >0 mean classification in the second class inclasses_
.

fit
(X, y, views_ind=None) Build a multimodal boosted classifier from the training set (X, y).
Parameters: X : dict dictionary with all views
or MultiModalData , MultiModalArray, MultiModalSparseArray or {arraylike, sparse matrix}, shape = (n_samples, n_features) Training multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
y : arraylike, shape = (n_samples,)
Target values (class labels).
views_ind : arraylike (default=[0, n_features//2, n_features])
Paramater specifying how to extract the data views from X:
If views_ind is a 1D array of sorted integers, the entries indicate the limits of the slices used to extract the views, where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.With this convention each view is therefore a view (in the NumPy sense) of X and no copy of the data is done.
If views_ind is an array of arrays of integers, then each array of integers
views_ind[n]
specifies the indices of the viewn
, which is then given byX[:, views_ind[n]]
.With this convention each view creates therefore a partial copy of the data in X. This convention is thus more flexible but less efficient than the previous one.
Returns: self : object
Returns self.

get_params
(deep=True) Get parameters for this estimator.
Parameters: deep : bool, default=True
If True, will return the parameters for this estimator and contained subobjects that are estimators.
Returns: params : dict
Parameter names mapped to their values.

predict
(X) Predict classes for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in the ensemble.
Parameters: X : {arraylike, sparse matrix}, shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: y : numpy.ndarray, shape = (n_samples,)
Predicted classes.

score
(X, y) Return the mean accuracy on the given test data and labels.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview test samples. Sparse matrix can be CSC, CSR
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : float
Mean accuracy of self.predict(X) wrt. y.

set_params
(**params) Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as
Pipeline
). The latter have parameters of the form<component>__<parameter>
so that it’s possible to update each component of a nested object.Parameters: **params : dict
Estimator parameters.
Returns: self : estimator instance
Estimator instance.

staged_decision_function
(X) Compute decision function of X for each boosting iteration.
This method allows monitoring (i.e. determine error on testing set) after each boosting iteration.
Parameters: X : {arraylike, sparse matrix}, shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR. maybe also MultimodalData
Returns: dec_fun : generator of numpy.ndarrays, shape = (n_samples, k)
Decision function of the input samples. The order of outputs is the same of that of the classes_ attribute. Binary classification is a special cases with
k == 1
, otherwisek==n_classes
. For binary classification, values <=0 mean classification in the first class inclasses_
and values >0 mean classification in the second class inclasses_
.

staged_predict
(X) Return staged predictions for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in the ensemble.
This generator method yields the ensemble prediction after each iteration of boosting and therefore allows monitoring, such as to determine the prediction on a test set after each boost.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: y : generator of numpy.ndarrays, shape = (n_samples,)
Predicted classes.

staged_score
(X, y) Return staged mean accuracy on the given test data and labels.
This generator method yields the ensemble score after each iteration of boosting and therefore allows monitoring, such as to determine the score on a test set after each boost.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview test samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : generator of floats
Mean accuracy of self.staged_predict(X) wrt. y.
 if
multimodal.boosting.combo
This module contains a MultiConfusion MMatrix Bosting (CoMBo)
estimator for classification implemented in the MuComboClassifier
class.

class
multimodal.boosting.combo.
MuComboClassifier
(base_estimator=None, n_estimators=50, random_state=None) It then iterates the process on the same dataset but where the weights of incorrectly classified instances are adjusted such that subsequent classifiers focus more on difficult cases. A MuCoMBo classifier.
A MuMBo classifier is a metaestimator that implements a multimodal (or multiview) boosting algorithm:
It fits a set of classifiers on the original dataset splitted into several views and retains the classifier obtained for the best view.
This class implements the MuMBo algorithm [R34].
Parameters: base_estimator : object, optional (default=DecisionTreeClassifier)
Base estimator from which the boosted ensemble is built. Support for sample weighting is required, as well as proper classes_ and n_classes_ attributes. The default is a DecisionTreeClassifier with parameter
max_depth=1
.n_estimators : integer, optional (default=50)
Maximum number of estimators at which boosting is terminated.
random_state : int, RandomState instance or None, optional (default=None)
If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random.
See also
sklearn.ensemble.AdaBoostClassifier
,sklearn.ensemble.GradientBoostingClassifier
,sklearn.tree.DecisionTreeClassifier
References
[R34] (1, 2) Koc{c}o, Sokol and Capponi, C{‘e}cile A Boosting Approach to Multiview Classification with Cooperation, 2011,Proceedings of the 2011 European Conference on Machine Learning and Knowledge Discovery in Databases  Volume Part II, 209–228 SpringerVerlag https://link.springer.com/chapter/10.1007/9783642237836_1 [R44] Sokol Koço, “Tackling the uneven views problem with cooperation based ensemble learning methods”, PhD Thesis, AixMarseille Université, 2013, http://www.theses.fr/en/2013AIXM4101. Examples
>>> from multimodal.boosting.combo import MuComboClassifier >>> from sklearn.datasets import load_iris >>> X, y = load_iris(return_X_y=True) >>> views_ind = [0, 2, 4] # view 0: sepal data, view 1: petal data >>> clf = MuComboClassifier(random_state=0) >>> clf.fit(X, y, views_ind) MuComboClassifier(random_state=0) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [0] >>> views_ind = [[0, 2], [1, 3]] # view 0: length data, view 1: width data >>> clf = MuComboClassifier(random_state=0) >>> clf.fit(X, y, views_ind) MuComboClassifier(random_state=0) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [0]
>>> from sklearn.tree import DecisionTreeClassifier >>> base_estimator = DecisionTreeClassifier(max_depth=2) >>> clf = MuComboClassifier(base_estimator=base_estimator, random_state=1) >>> clf.fit(X, y, views_ind) MuComboClassifier(base_estimator=DecisionTreeClassifier(max_depth=2), random_state=1) >>> print(clf.predict([[ 5., 3., 1., 1.]])) [0]
Attributes
estimators_ (list of classifiers) Collection of fitted subestimators. classes_ (numpy.ndarray, shape = (n_classes,)) Classes labels. n_classes_ (int) Number of classes. n_views_ (int) Number of views estimator_weights_ (numpy.ndarray of floats, shape = (len(estimators_),)) Weights for each estimator in the boosted ensemble. estimator_errors_ (array of floats) Empirical loss for each iteration. best_views_ (numpy.ndarray of integers, shape = (len(estimators_),)) Indices of the best view for each estimator in the boosted ensemble. n_yi_ (numpy ndarray of int contains number of train sample for each classe shape (n_classes,)) 
decision_function
(X) Compute the decision function of X.
Parameters: X : {arraylike, sparse matrix}, shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: dec_fun : numpy.ndarray, shape = (n_view, n_samples, k)
Decision function of the input samples. The order of outputs is the same of that of the classes_ attribute. Binary classification is a special cases with
k == 1
, otherwisek == n_classes
. For binary classification, values <=0 mean classification in the first class inclasses_
and values >0 mean classification in the second class inclasses_
.

fit
(X, y, views_ind=None) Build a multimodal boosted classifier from the training set (X, y).
Parameters: X : dict dictionary with all views
or MultiModalData , MultiModalArray, MultiModalSparseArray or {arraylike, sparse matrix}, shape = (n_samples, n_features) Training multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
y : arraylike, shape = (n_samples,)
Target values (class labels).
views_ind : arraylike (default=[0, n_features//2, n_features])
Paramater specifying how to extract the data views from X:
If views_ind is a 1D array of sorted integers, the entries indicate the limits of the slices used to extract the views, where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.With this convention each view is therefore a view (in the NumPy sense) of X and no copy of the data is done.
If views_ind is an array of arrays of integers, then each array of integers
views_ind[n]
specifies the indices of the viewn
, which is then given byX[:, views_ind[n]]
.With this convention each view creates therefore a partial copy of the data in X. This convention is thus more flexible but less efficient than the previous one.
Returns: self : object
Returns self.
Raises: ValueError estimator must support sample_weight
ValueError where `X` and `view_ind` are not compatibles

get_params
(deep=True) Get parameters for this estimator.
Parameters: deep : bool, default=True
If True, will return the parameters for this estimator and contained subobjects that are estimators.
Returns: params : dict
Parameter names mapped to their values.

predict
(X) Predict classes for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in the ensemble.
Parameters: X : {arraylike, sparse matrix}, shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: y : numpy.ndarray, shape = (n_samples,)
Predicted classes.
Raises: ValueError ‘X’ input matrix must be have the same total number of features
of ‘X’ fit data

score
(X, y) Return the mean accuracy on the given test data and labels.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview test samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : float
Mean accuracy of self.predict(X) wrt. y.

set_params
(**params) Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as
Pipeline
). The latter have parameters of the form<component>__<parameter>
so that it’s possible to update each component of a nested object.Parameters: **params : dict
Estimator parameters.
Returns: self : estimator instance
Estimator instance.

staged_decision_function
(X) Compute decision function of X for each boosting iteration.
This method allows monitoring (i.e. determine error on testing set) after each boosting iteration.
Parameters: X : {arraylike, sparse matrix}, shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: dec_fun : generator of numpy.ndarrays, shape = (n_samples, k)
Decision function of the input samples. The order of outputs is the same of that of the classes_ attribute. Binary classification is a special cases with
k == 1
, otherwisek==n_classes
. For binary classification, values <=0 mean classification in the first class inclasses_
and values >0 mean classification in the second class inclasses_
.

staged_predict
(X) Return staged predictions for X.
The predicted class of an input sample is computed as the weighted mean prediction of the classifiers in the ensemble.
This generator method yields the ensemble prediction after each iteration of boosting and therefore allows monitoring, such as to determine the prediction on a test set after each boost.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview input samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
Returns: y : generator of numpy.ndarrays, shape = (n_samples,)
Predicted classes.

staged_score
(X, y) Return staged mean accuracy on the given test data and labels.
This generator method yields the ensemble score after each iteration of boosting and therefore allows monitoring, such as to determine the score on a test set after each boost.
Parameters: X : {arraylike, sparse matrix} of shape = (n_samples, n_features)
Multiview test samples. Sparse matrix can be CSC, CSR, COO, DOK, or LIL. COO, DOK and LIL are converted to CSR.
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : generator of floats
Mean accuracy of self.staged_predict(X) wrt. y.

Kernels
multimodal.kernels.mvml

class
multimodal.kernels.mvml.
MVML
(lmbda=0.1, eta=1, nystrom_param=1.0, kernel='linear', kernel_params=None, learn_A=1, learn_w=0, precision=0.0001, n_loops=6) The MVML Classifier
Parameters: lmbda : float regression_params lmbda (default = 0.1) for basic regularization
eta : float regression_params eta (default = 1), first for basic regularization,
regularization of A (not necessary if A is not learned)
kernel : list of str (default: “precomputed”) if kernel is as input of fit function set kernel to
“precomputed” list or str indicate the metrics used for each kernels list of pairwise kernel function name (default : “precomputed”) if kernel is as input of fit function set kernel to “precomputed” example : [‘rbf’, ‘additive_chi2’, ‘linear’ ] for function defined in as PAIRWISE_KERNEL_FUNCTIONS
kernel_params : list of str default
list of dict of corresponding kernels params KERNEL_PARAMS
nystrom_param: value between 0 and 1 indicating level of nyström approximation; 1 = no approximation
learn_A : integer (default 1) choose if A is learned or not: 1  yes (default);
2  yes, sparse; 3  no (MVML_Cov); 4  no (MVML_I)
learn_w : integer (default 0) where learn w is needed
precision : float (default
n_loops : (default 6) number of iterions
Examples
>>> from multimodal.kernels.mvml import MVML >>> from sklearn.datasets import load_iris >>> X, y = load_iris(return_X_y=True) >>> y[y>0] = 1 >>> views_ind = [0, 2, 4] # view 0: sepal data, view 1: petal data >>> clf = MVML() >>> clf.get_params() {'eta': 1, 'kernel': 'linear', 'kernel_params': None, 'learn_A': 1, 'learn_w': 0, 'lmbda': 0.1, 'n_loops': 6, 'nystrom_param': 1.0, 'precision': 0.0001} >>> clf.fit(X, y, views_ind) MVML() >>> print(clf.predict([[ 5., 3., 1., 1.]])) 0
Attributes
lmbda (float regression_params lmbda (default = 0.1)) eta (float regression_params eta (default = 1)) regression_params (array/list of regression parameters) kernel (list or str indicate the metrics used for each kernels) list of pairwise kernel function name (default : “precomputed”) example : [‘rbf’, ‘additive_chi2’, ‘linear’ ] for function defined in as PAIRWISE_KERNEL_FUNCTIONS example kernel=[‘rbf’, ‘rbf’], for the first two views kernel_params: list of dict of corresponding kernels params KERNEL_PARAMS learn_A ( 1 where Learn matrix A is needded) learn_w (integer where learn w is needed) precision (float (default) n_loops (number of itterions) n_approx (number of samples in approximation, equals n if no approx.) classes_ (array like unique label for classes) warning_message (dictionary with warning messages) X_ ( metriclearning.datasets.data_sample.Metriclearn_array
array of input sample)K_ ( metriclearning.datasets.data_sample.Metriclearn_array
array of processed kernels)y_ (arraylike, shape = (n_samples,)) Target values (class labels). regression_ (if the classifier is used as regression (default) 
decision_function
(X) Compute the decision function of X.
Parameters: X : { arraylike, sparse matrix},
shape = (n_samples, n_views * n_features) Multiview input samples. maybe also MultimodalData
Returns: dec_fun : numpy.ndarray, shape = (n_samples, )
Decision function of the input samples. For binary classification, values <=0 mean classification in the first class in
classes_
and values >0 mean classification in the second class inclasses_
.

fit
(X, y=None, views_ind=None) Fit the MVML classifier
Parameters: X :  Metriclearn_array {arraylike, sparse matrix}, shape = (n_samples, n_features)
Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
or  Dictionary of {array like} with shape = (n_samples, n_features) for multiview
for each view.
 Array of {array like} with shape = (n_samples, n_features) for multiview for each view.
 {array like} with (n_samples, nviews * n_features) with ‘views_ind’ diferent to ‘None’
y : arraylike, shape = (n_samples,)
Target values (class labels). array of length n_samples containing the classification/regression labels for training data
views_ind : arraylike (default=[0, n_features//2, n_features])
Paramater specifying how to extract the data views from X:
views_ind is a 1D array of sorted integers, the entries indicate the limits of the slices used to extract the views, where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.With this convention each view is therefore a view (in the NumPy sense) of X and no copy of the data is done.
Returns: self : object
Returns self.

get_params
(deep=True) Get parameters for this estimator.
Parameters: deep : bool, default=True
If True, will return the parameters for this estimator and contained subobjects that are estimators.
Returns: params : dict
Parameter names mapped to their values.

predict
(X) Parameters: X : different formats are supported
 Metriclearn_array {arraylike, sparse matrix}, shape = (n_samples, n_features) Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
 Dictionary of {array like} with shape = (n_samples, n_features) for multiview for each view.
 Array of {array like} with shape = (n_samples, n_features) for multiview for each view.
 {array like} with (n_samples, nviews * n_features) with ‘views_ind’ diferent to ‘None’
Returns: y : numpy.ndarray, shape = (n_samples,)
Predicted classes.

score
(X, y) Return the mean accuracy on the given test data and labels.
Parameters: X : {arraylike} of shape = (n_samples, n_features)
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : float
Mean accuracy of self.predict(X) wrt. y.

set_params
(**params) Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as
Pipeline
). The latter have parameters of the form<component>__<parameter>
so that it’s possible to update each component of a nested object.Parameters: **params : dict
Estimator parameters.
Returns: self : estimator instance
Estimator instance.

multimodal.kernels.lpMKL

class
multimodal.kernels.lpMKL.
MKL
(lmbda, nystrom_param=1.0, kernel='linear', kernel_params=None, use_approx=True, precision=0.0001, n_loops=50) MKL Classifier for multiview learning
Parameters: lmbda : float coeficient for combined kernels
nystrom_param : float (default
value between 0 and 1 indicating level of nyström approximation; 1 = no approximation
kernel : list of str (default: “precomputed”) if kernel is as input of fit function set kernel to
“precomputed” list or str indicate the metrics used for each kernels list of pairwise kernel function name (default : “precomputed”) if kernel is as input of fit function set kernel to “precomputed” example : [‘rbf’, ‘additive_chi2’, ‘linear’ ] for function defined in as PAIRWISE_KERNEL_FUNCTIONS
kernel_params : list of str default
list of dict of corresponding kernels params KERNEL_PARAMS
use_approx : (default
n_loops : (default 50) number of iterions
Attributes
lmbda (float coeficient for combined kernels) m_param (float (default) value between 0 and 1 indicating level of nyström approximation; 1 = no approximation kernel (list or str indicate the metrics used for each kernels) list of pairwise kernel function name (default : “precomputed”) example : [‘rbf’, ‘additive_chi2’, ‘linear’ ] for function defined in as PAIRWISE_KERNEL_FUNCTIONS example kernel=[‘rbf’, ‘rbf’], for the first two views kernel_params: list of dict of corresponding kernels params KERNEL_PARAMS precision (float (default) n_loops (number of iterions) classes_ (array like unique label for classes) X_ ( metriclearning.datasets.data_sample.Metriclearn_array
array of input sample)K_ ( metriclearning.datasets.data_sample.Metriclearn_array
array of processed kernels)y_ (arraylike, shape = (n_samples,)) Target values (class labels). C (learning solution that is learned in MKL) weights (learned weight for combining the solutions of views, learned in) 
decision_function
(X) Compute the decision function of X.
Parameters: X : dict dictionary with all views {array like} with shape = (n_samples, n_features) for multiview
for each view. or MultiModalData , MultiModalArray or {arraylike,}, shape = (n_samples, n_features) Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
Returns: dec_fun : numpy.ndarray, shape = (n_samples, )
Decision function of the input samples. For binary classification, values <=0 mean classification in the first class in
classes_
and values >0 mean classification in the second class inclasses_
.

fit
(X, y=None, views_ind=None) Parameters: X : different formats are supported
 Metriclearn_array {arraylike, sparse matrix}, shape = (n_samples, n_features) Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
 Dictionary of {array like} with shape = (n_samples, n_features) for multiview for each view.
 Array of {array like} with shape = (n_samples, n_features) for multiview for each view.
 {array like} with (n_samples, nviews * n_features) with ‘views_ind’ diferent to ‘None’
y : arraylike, shape = (n_samples,)
Target values (class labels). array of length n_samples containing the classification/regression labels for training data
views_ind : arraylike (default=[0, n_features//2, n_features])
Paramater specifying how to extract the data views from X:
views_ind is a 1D array of sorted integers, the entries indicate the limits of the slices used to extract the views, where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.With this convention each view is therefore a view (in the NumPy sense) of X and no copy of the data is done.
Returns: self : object
Returns self.

get_params
(deep=True) Get parameters for this estimator.
Parameters: deep : bool, default=True
If True, will return the parameters for this estimator and contained subobjects that are estimators.
Returns: params : dict
Parameter names mapped to their values.

learn_lpMKL
() function of lpMKL learning
Returns: return tuple (C, weights)

lpMKL_predict
(X, C, weights) Parameters: X : arraylike test kernels precomputed array like
C : corresponding to Confusion learned matrix
weights : learned weights
Returns: y : numpy.ndarray, shape = (n_samples,)
Predicted classes.

predict
(X) Parameters: X : dict dictionary with all views {array like} with shape = (n_samples, n_features) for multiview
for each view. or MultiModalData , MultiModalArray or {arraylike,}, shape = (n_samples, n_features) Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
views_ind : arraylike (default=[0, n_features//2, n_features])
Paramater specifying how to extract the data views from X:
views_ind is a 1D array of sorted integers, the entries indicate the limits of the slices used to extract the views, where view
n
is given byX[:, views_ind[n]:views_ind[n+1]]
.With this convention each view is therefore a view (in the NumPy sense) of X and no copy of the data is done.
Returns: y : numpy.ndarray, shape = (n_samples,)
Predicted classes.

score
(X, y) Return the mean accuracy on the given test data and labels.
Parameters: X : dict dictionary with all views {array like} with shape = (n_samples, n_features) for multiview
for each view. or MultiModalData , MultiModalArray or {arraylike,}, shape = (n_samples, n_features) Training multiview input samples. can be also Kernel where attibute ‘kernel’ is set to precompute “precomputed”
y : arraylike, shape = (n_samples,)
True labels for X.
Returns: score : float
Mean accuracy of self.predict(X) wrt. y.

set_params
(**params) Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as
Pipeline
). The latter have parameters of the form<component>__<parameter>
so that it’s possible to update each component of a nested object.Parameters: **params : dict
Estimator parameters.
Returns: self : estimator instance
Estimator instance.

multimodal.kernels.mkernel

class
multimodal.kernels.mkernel.
MKernel
Abstract class MKL and MVML should inherit from for methods of transform kernel to/from data.
Attributes
W_sqrootinv_dict (dict of nyström approximation kernel) in the case of nystrom approximation the a dictonary of reduced kernel is calculated kernel_params (list of dict of corresponding kernels) params KERNEL_PARAMS