← accel-gpu

Simple Neural Network

Feedforward inference: 784 → 128 (ReLU) → 10 (softmax). Random weights for demo. Simulates MNIST-style digit classification.

Running inference…