gzyIDE / rust_perceptron

Rust practice

perceptron for MNIST Classification

Single layer perceptron to classify hand-written digits, whose size is 28x28 pix.

layers

Input: 28x28 (f64)
Weight: 28x28 (f64)
Output: 10x1 (f64)

Classification accuracy

Trained with 10000 training data, the perceptron achieved around 80% Top-1 accuracy.