I found that directly using libabries such as tensorflow, pytorch, and sklearn made it difficult to fundementally understand the math and inner workings of neural networks. Therefore, I decided to build one from scratch only using numpy in order to gain a more intuitive understanding of neural networks.
Planning on including:
- AdaM Optimization (RMS Prop & momentum)
- Mini Batch gradient descent
- Regularization (L2 & Dropout)
- Learning Rate Decay