relu-network

Machine Learning Function Approximation: This code implements the fully-connected Deep Neural Network (DNN) architectures considered in the paper "The gap between theory and practice in function approximation with deep neural networks" available at https://arxiv.org/abs/2001.07523

Multi-Head ReLU Implicit Neural Representation Networks

Code for optimizing MILP problems with embedded ReLU networks

Implementing a 4 layer neural to identify digits from a 28x28 grid using just python and numpy

This project utilizes neural networks to recognize handwritten digits (0-9) through multiclass classification, employing ReLU activation and Softmax function for accurate predictions.

Text Generation

This repository delves into the role of activation functions in perceptron-based classification models. It features a comprehensive Jupyter notebook demonstrating different activation functions, their mathematical foundations, and their impact on model performance.

Traininig a simple neural network with ReLU activation function to classify the data.

Proof 5: Why neural networks prefer low frequencies — a ReLU frequency tradeoff theorem. 3/3 theorems empirically verified.

Proof 7: Why the Cantor function (Devil's Staircase) is structurally hard for ReLU networks. All theorems verified.

Computer Vision Model for Age Estimation from Facial Images

Neural Network from scratch, LoRA implementation, and SMS spam cleaning with MinHash deduplication

Graph by matplotlib