Deep learning library from scratch (numpy)
see 1D CNN example in repository
Supports
- 1D Convolution (Conv1D)
- Max & Average Pooling (MaxPool1D, AvgPool1D)
- Fully Connected Layers (FCNN)
- Activations: ReLU, LeakyReLU, Sigmoid, Swish
- Loss Functions: Squared Error Loss (MSE), Binary Cross-Entropy (planned)
- Forward and Backward Pass: gradients computed by chain rule
- Flatten Layer: Reshapes feature maps for FCNN
- Utility Function: Padding, logging