PROJECT NOTES / AI & MACHINE LEARNING
MNIST Model Benchmark
Compare model quality alongside computation cost.
A digit-recognition benchmark exploring KNN, PCA with KNN, SVM, Random Forest, and a convolutional neural network.
AI & MACHINE LEARNING↗
01MNIST
02Preprocess
03Train models
04Evaluate
The problem
An accuracy score alone obscures differences in training and prediction cost across algorithms.
How it comes together
The study compares multiple classical and deep-learning approaches on MNIST, with analysis of accuracy, training time, and prediction time.
03 / UNDER THE HOOD
SIMPLIFIED ARCHITECTUREFollow the flow.
01MNIST
02Preprocess
03Train models
04Evaluate
05Compare
Classical algorithms and a CNN in one benchmark.
Dimensionality reduction through PCA plus KNN.
Training and inference timing comparisons.
Notebook-based analysis and visualizations.
Why this approach?
Compare quality and computation together so the tradeoff remains visible when selecting an approach.
Explore the project’s supporting sources.Project repository
KEEP EXPLORING