Sparse Coding

An implementation of Sparse Coding with Dictionary Learning. Given a dataset, this will decompose the dataset into a sparse combination of a few dictionary elements, where the dictionary is learned during computation.

:- use_module('path/to/.../src/methods/sparse_coding/sparse_coding.pl').

%% usage example
TrainData = [5.1,3.5,1.4, 4.9,3.0,1.4, 4.7,3.2,1.3, 4.6,3.1,1.5],
TestData = [3,2,0, 5,1,4, 0,0,4, 3,3,5, 0,5,5, 2,5,5],
sparse_coding_initModelWithTrain(TrainData, 3, 15, 0.1, 0.0, 100, 0.01, 0.000001),
sparse_coding_encode(TestData, 3, CodesList, CodesRows).

Available Predicates

links/resources

sparse_coding_initModelWithTrain/8

Initializes sparse_coding model and trains it.

%% predicate definition 
sparse_coding_initModelWithTrain(MatList, MatRows, Atoms, Lambda1, Lambda2, MaxIterations, ObjTolerance, NewtonTolerance) :-
        Atoms >= 0,
        Lambda1 > 0,
        Lambda2 >= 0,
        MaxIterations >= 0,
        ObjTolerance > 0,
        NewtonTolerance > 0,
        convert_list_to_float_array(MatList, MatRows, array(Xsize, Xrownum, X)),
        initModelWithTrainI(X, Xsize, Xrownum, Atoms, Lambda1, Lambda2, MaxIterations, ObjTolerance, NewtonTolerance).

%% foreign c++ predicate definition 
foreign(initModelWithTrain, c, initModelWithTrainI( +pointer(float_array), +integer, +integer,
                                                    +integer, +float32, +float32, +integer, +float32, +float32)).

Parameters

Name	Type	Description	Default
data	+matrix	Matrix of training data.	-
atoms	+integer	Number of atoms in the dictionary.	15
lambda1	+float	Sparse coding l1-norm regularization parameter.	0.0
lambda2	+float	Sparse coding l2-norm regularization parameter.	0.0
maxIterations	+integer	Maximum number of iterations for sparse coding (0 indicates no limit).	0
objTolerance	+float	Tolerance for convergence of the objective function.	0.01
newtonTolerance	+float	Tolerance for convergence of Newton method.	1e-6

sparse_coding_initModelNoTrain/6

Initializes sparse_coding model but will not train the model, and a subsequent call to Train will be required before the model can encode points with Encode.

%% predicate definition 
sparse_coding_initModelNoTrain(Atoms, Lambda1, Lambda2, MaxIterations, ObjTolerance, NewtonTolerance) :-
        Atoms >= 0,
        Lambda1 > 0,
        Lambda2 >= 0,
        MaxIterations >= 0,
        ObjTolerance > 0,
        NewtonTolerance > 0,
        initModelNoTrainI(Atoms, Lambda1, Lambda2, MaxIterations, ObjTolerance, NewtonTolerance).

%% foreign c++ predicate definition 
foreign(initModelNoTrain, c, initModelNoTrainI(+integer, +float32, +float32, +integer, +float32, +float32)).

Parameters

Name	Type	Description	Default
atoms	+integer	Number of atoms in the dictionary.	15
lambda1	+float	Sparse coding l1-norm regularization parameter.	0.0
lambda2	+float	Sparse coding l2-norm regularization parameter.	0.0
maxIterations	+integer	Maximum number of iterations for sparse coding (0 indicates no limit).	0
objTolerance	+float	Tolerance for convergence of the objective function.	0.01
newtonTolerance	+float	Tolerance for convergence of Newton method.	1e-6

sparse_coding_encode/4

Sparse code each point in the given dataset via LARS, using the current dictionary and store the encoded data in the codes matrix.

%% predicate definition 
sparse_coding_encode(DataList, DataRows, CodesList, YCols) :-
        convert_list_to_float_array(DataList, DataRows, array(Xsize, Xrownum, X)),
        encodeI(X, Xsize, Xrownum, Y, YCols, YRows),
        convert_float_array_to_2d_list(Y, YCols, YRows, CodesList).

%% foreign c++ predicate definition 
foreign(encode, c, encodeI( +pointer(float_array), +integer, +integer, 
                            -pointer(float_array), -integer, -integer)).

Parameters

Name	Type	Description	Default
data	+matrix	Input data matrix to be encoded.	-
codes	-matrix	Output codes matrix.	-

sparse_coding_train/3

Train the sparse coding model on the given dataset.

%% predicate definition 
sparse_coding_train(DataList, DataRows, ReturnValue) :-
        convert_list_to_float_array(DataList, DataRows, array(Xsize, Xrownum, X)),
        trainI(X, Xsize, Xrownum, ReturnValue).

%% foreign c++ predicate definition 
foreign(train, c, trainI( +pointer(float_array), +integer, +integer, 
                          [-float32])).

Parameters

Name	Type	Description	Default
data	+matrix	Matrix of training data.	-
objValue	-float	The final objective value.	-

Connected Links/Resources

If you want a more detailed explanation, then go to the python documentation. There is most of the time a good explanation on how the methods work and what the parameters do.

added some of the links from the python documentation

Comments

Please register or sign in to add a comment.