Project of 4th semester of MIPT.
More information can be found in SDP
Used files: NerNet and Mnist
The lib have three architecture of neural networks:
- Dense-800
- Dense-200 - Dense-100 - Dense-50
- Conv-32 - MaxPool - Dense-256
You should use the method make_and_check of the class NerNet. The parametr path is responsible of architecture of nernet. For example, if path = "Weight/Conv", class uses 3rd acrhitecture, moreover, if directory is empty, nernet will fit itself and save weights. If directory is full, nernet will use these weights.