Neural Networks From Scratch- Anusha Kodavuru

Neural Networks From Scratch

A curated collection of Kaggle notebooks covering Neural Networks from first principles to practical implementations in Deep Learning using PyTorch and TensorFlow.

Building Micrograd Build an automatic differentiation engine from scratch by deriving gradients, constructing a computational graph, and implementing backpropagation inspired by Micrograd.
Build a Neural Network Build your first Artificial Neural Network in Keras and experiment with hidden layers, neurons, activation functions, optimizers, and epochs to improve model performance.
Perceptron Trick Understand how the Perceptron learning rule updates weights to classify linearly separable data using a synthetic classification dataset.
Perceptron as a Plane Visualize a perceptron as a geometric decision boundary and understand how weights and bias define separating hyperplanes.
Backpropagation (Regression) Derive and implement the complete backpropagation algorithm for a neural network performing regression, including forward propagation and manual gradient computation.
Backpropagation (Classification) Apply backpropagation to a neural network classifier and understand how gradients propagate through the network for binary classification.
Digit Classification Train an Artificial Neural Network on the MNIST handwritten digit dataset and build a complete multiclass image classification pipeline.
Graduate Admission Prediction Develop an ANN regression model to predict graduate admission chances, including preprocessing, feature engineering, and model evaluation.