Time series project MVA. Implementation of the paper Online Dictionary Learning for Sparse Coding ( https://www.di.ens.fr/sierra/pdfs/icml09.pdf)
The paper introduces an online optimization procedure for the problem of dictionary learning designed to dynamically adapt to changing data distributions. Leveraging stochastic gradient descent for incremental updates and incorporating a sparsity-inducing L1 penalty, the proposed algorithm enhances the efficiency and adaptability of sparse coding.
Experiment have been done on both image and audio parts.
-
The data for images experiments was taken here : https://www.kaggle.com/datasets/puneet6060/intel-image-classification?resource=download&select=seg_test (Also available in the image_dataset folder)
-
In the experiments Folder you will find the notebook of the image experiment and also signal experiments
-
In the src folder you will find the source code for the experiment
-
In the Dataset folder, you will find the data used in signal experiments