TinyNet is a tiny neural network developed and vectorized with NumPy, and designed in a modular fashion. The aim is to provide students with an understanding of basic building blocks of neural networks and their coordination.
Features:
- Batch, mini-batch and stochastic GD
- Layers:
- Dense
- Convolutional (2D)
- Pooling (2D)
- Regularizers:
- L2
- Dropout
- Optimizers:
- Momentum
- Adam
- Batch normalization
- Gradient Checking
Notebooks: