See the calculation, change it, then observe the consequence.

ModelOrigen is an educational, browser-first application for machine learning and deep learning. The name combines Model with Origen (origin): the product is built to help learners trace where a model's prediction comes from by connecting visible output with the calculations, parameters, intermediate states, predictions, and evaluation evidence that produced it.

Built by

Origen Technologies builds and maintains this application.

What the application covers

The current visualizer includes classical models such as Linear Regression, Logistic Regression, K-Means, KNN, Decision Trees, SVM, Random Forest, Gradient Boosting, XGBoost, AdaBoost, and Extra Trees, plus focused deep-learning labs for MLP neural networks, CNNs, RNNs, LSTMs, GRUs, Transformer self-attention, autoencoders, and a tiny educational GAN.

How computation works

Dataset preparation and model training are designed to run in the browser. Expensive model work uses Web Workers so the production host can remain a static-file server rather than a cloud ML backend.

Scope

ModelOrigen is for educational exploration. It is not AutoML, a hosted training service, model deployment infrastructure, or a substitute for production-scale statistical validation.