File: vapnik-chervonenkis-game.html
What does it mean to shatter a dataset? This game teaches the concept behind Vapnik-Chervonenkis dimension through interactive point classification. Learn how different separators (lines, circles, rectangles) can classify datasets and discover when a hypothesis class can shatter any configuration of points.
Key Concepts: VC-dimension, shattering, hypothesis classes, classification
File: rademacher-calculator.html
What is Rademacher complexity? Play around with different degree polynomials and random data to explore this fundamental concept in statistical learning theory. This interactive calculator performs numerical Rademacher complexity calculations with real-time visualizations.
Key Concepts: Rademacher complexity, generalization bounds, polynomial fitting
File: overfitting-explorer.html
Explore the fundamental tradeoff between bias and variance through interactive polynomial fitting. Adjust model complexity, data size, and noise levels to see overfitting happen in real-time. Includes bias-variance decomposition analysis and learning curves to understand when models generalize well vs. when they memorize training data.
Key Concepts: Overfitting, underfitting, bias-variance tradeoff, model complexity, generalization
File: padic-gibbs-explorer.html
Discover how regression works in non-Archimedean settings through interactive p-adic linear regression. Unlike traditional regression that minimizes Euclidean distance, p-adic regression uses the p-adic norm where “closeness” has a completely different meaning. Watch as the Gibbs sampler explores parameter space using Boltzmann acceptance probabilities, with real-time convergence analysis and trend visualization. This is the implementation from https://arxiv.org/abs/2503.23488 p-Adic Polynomial Regression as Alternative to Neural Network for Approximating p-Adic Functions of Many Variables by Alexander Zubrev.
Key Concepts: p-adic numbers, non-Archimedean analysis, Gibbs sampling, Boltzmann distribution, alternative geometries
File: cn2_learning_game.html
Play as the CN2 rule learner to craft if-then rules using beam search. Balance coverage and precision to build a classifier that predicts when to play tennis.
Key Concepts: rule-based learning, beam search, coverage vs precision, CN2 algorithm
File: network-analysis-playground.html
Explore Week 11 network analysis ideas through five small canvas activities: force-directed layouts, centrality measures, community detection, bipartite projection, and resilience under node removal. The activities use small classroom-sized graphs so students can connect each visual change to a specific graph concept.
Key Concepts: network analysis, spring layouts, centrality, communities, bipartite projection, network resilience
File: p3-recommendation-engine.html
Teach path-based collaborative filtering with an animated
user-movie graph built from the MovieLens
ml-latest-small dataset. Add movies you like,
watch nearby MovieLens users attach to the graph, then run
repeated P3 random walks from you to a movie, to a user, and
back to a recommended movie. Recommendation counts update
live while each three-hop path is highlighted on the
canvas.
Key Concepts: recommender systems, P3 random walks, collaborative filtering, bipartite graphs, MovieLens
Explore more interactive mathematical and computational tools: