This noteboook is intended to provide a simple pytorch implementation for dictionary learning (DL) based on stochastic gradient descent. This is in essence a multi-purpose DL method that I put together in the midst of my own work, and I frequently use to carry out simple experiments and tests. This implementation does not follow any particular publication, though it is perhaps closest to the OSDL work on sparse dictionaries (without the double-sparisy component), or to the Online Dictionary Learning algorithm by Mairal (though without their convex surrogate function approach).
Look at Description.ipynb for the general description of the algorithm.