AI Experiments

Ongoing posts and code documenting the process of building a question answering model.

Explore how to use LIME and SHAP for interpretability.

Learn about using interpretability techniques on deep learning models.

Refractor predicts churn probabilities for telecom customers and shows which customer attributes contribute to those predictions.

Learn how to use deep learning approaches (including sequence models, VAEs, and GANS) for anomaly detection.

A semantic search engine that takes some input text and returns relevant famous quotes.

Learn about how we used transfer learning and a pretrained BERT model to build our sentiment analysis prototype.

A notebook showing how to train a complaint classifier with Snorkel. Using data from the Consumer Financial Protection Bureau.

Handtrack.js is a library for prototyping realtime hand detection (bounding box), directly in the browser.

An interactive UMAP visualization of the MNIST data set.

A toy example about logistic regression and different active learning strategies.

A probabilistic programming prototype that predicts future real estate prices across New York City boroughs and neighborhoods.

Brief uses neural networks to score and highlight the most interesting sentences within any article.

An interactive notebook about using three.js to render tens of thousands of points.

Encartopedia visualizes Wikipedia topic clusters and plots your journey through them.

An interactive visualization that uses T-SNE to cluster movies together based on user ratings.

Luhn's method, from 1958, provides a foundation for understanding modern auto-summarization techniques.