Bipul Poudel
all writing

Reading series · 5 posts

Neural Network From Scratch Using Numpy only

Build a neural network step by step using only Python and NumPy.

  1. 01
    Oct 05, 2026 · 12 min read

    A Single Neuron From Scratch With NumPy: Weights, Bias, and Activation Functions

    The first post in Neural Network From Scratch Using Numpy Only. Build one neuron, walk through the forward pass by hand, and compare step, sigmoid, and ReLU activations.

  2. 02
    Oct 05, 2026 · 16 min read

    Teaching a Single Neuron to Learn: Backpropagation and Gradient Descent in NumPy

    Part two of Neural Network From Scratch Using Numpy Only. Add a loss function, derive gradients with the chain rule, and watch a neuron with random weights learn the right ones.

  3. 03
    Oct 05, 2026 · 14 min read

    Why One Neuron Isn't Enough: The XOR Problem and the Case for Hidden Layers

    Part three of Neural Network From Scratch Using Numpy Only. See why a single neuron can only draw a straight line, watch it fail on XOR, and learn why hidden layers and non-linear activations change everything.

  4. 04
    Oct 14, 2026 · 8 min read

    From One Neuron to a Network: A Multilayer Forward Pass With NumPy

    Build a small neural network from layers of neurons, trace how weights and biases transform an input, and implement the forward pass using only NumPy.

  5. 05
    Oct 14, 2026 · 10 min read

    Training a Multilayer Neural Network With NumPy: Backpropagation From Scratch

    Extend a NumPy forward pass into a trainable multilayer network. Generate synthetic targets, derive hidden-layer gradients, and use backpropagation with stochastic gradient descent.