Machine Learning Playground
Logistic regression from scratch, an MNIST neural network and a collaborative-filtering recommender in one interactive app.
- My role
- Solo project
- Team
- Solo project
- Year
- 2025
The problem
After Andrew Ng's Machine Learning Specialization, I wanted to prove I could implement the core ideas myself, not just pass the quizzes.
What I built
- Logistic regression from scratch in NumPy, predicting heart-disease risk on the Framingham Heart Study dataset.
- An MNIST digit classifier in TensorFlow/Keras with training progress and sample predictions.
- A recommender system using user and movie embeddings in a neural network (collaborative filtering).
Results
- The MNIST classifier reached 97.33% test accuracy after 10 epochs.
Gallery
