rctn / sparsecoding

Reference sparse coding implementations for efficient learning and inference.
BSD 3-Clause "New" or "Revised" License
17 stars 4 forks source link

BUG: learn_dictionary uses only one batch update per epoch #64

Closed belsten closed 6 months ago

belsten commented 7 months ago

The learn_dictionary method only samples one random batch from the dataset per epoch, as opposed to using all the data in the dataset. In other words, one update was called a whole epoch. Noticed when training was much faster than expected.