| Age | Commit message (Collapse) | Author | |
|---|---|---|---|
| 2023-12-16 | Define vector outer product (nnMatrixMulOuter), which removes the need to ↵ | 3gg | |
| transpose layer inputs during training. | |||
| 2023-12-16 | Address BorrowMut TODO. There does not appear to be a good way to enforce ↵ | 3gg | |
| const-ness here. | |||
| 2023-12-16 | Decouple activations from linear layer. | 3gg | |
| 2023-11-23 | Documentation. | 3gg | |
| 2023-11-23 | Formatting. | 3gg | |
| 2022-05-15 | Optimize matrix multiplication to be more cache-friendly. | 3gg | |
| Mnist-1000: 15s -> 7.74s. | |||
| 2022-05-15 | Optimize away weights_T during training. | 3gg | |
| Mnist-1000: 18s -> 15s. | |||
| 2022-05-11 | Initial commit. | jeanne | |
 |
index : neuralnet | |
| Unnamed repository; edit this file 'description' to name the repository. |
| aboutsummaryrefslogtreecommitdiff |