Neural Network From Scratch

• 

  NeuralNetX

Inspiration : To truly understand how neural networks learn, beyond using high-level ML libraries, by building everything from first principles.

What it does : Implements a fully working neural network from scratch that learns the XOR problem using forward propagation and backpropagation.

How we built it : Built using core Python only, manually coding neurons, activation functions, loss calculation, gradient computation, and weight updates.

Challenges we ran into : Implementing backpropagation correctly without vectorized math and debugging gradient flow across layers.

Accomplishments that we're proud of : Successfully trained a neural network without any ML libraries and demonstrated real non-linear learning.

What we learned : Gained a deep understanding of how neural networks actually learn, how gradients work, and how small design choices affect training.

What's next for Neural Network From Scratch : Add configurable architectures, experiment with ReLU activation, and extend the model to real-world datasets.

Built With

• python
• vscode