Neural Network FAQ

Learning and using neural networks in an intuitive, visual way: interactive tutorials with charts and animations, a visualized neural network lab, mission-like challenges for key concepts, visualized classic deep learning models, and a drag-to-build model editor.

The download is free. Pro features are available through an optional subscription; the tutorials, lab, missions and editor are all part of the free experience.

iOS 15.6 or later on the App Store, Android on Google Play, and macOS — three platforms from one codebase.

Activations, regularizations, loss functions, classifications, Embedding, CNN1d, CNN2d, RNN and more, shown with charts and animations — plus classic deep learning models explained through modular diagrams with detailed documents, reference papers and source code.

Yes. The visual model editor lets you build a model by dragging your fingers, with real-time calculation and display of each step's output and real-time error reporting for parameter, input type and value compatibility problems. You can export the model as PyTorch-style code.

Simplified datasets in 2D and 3D, for both regression and classification, with full visualization. The Values panel shows every number of a sample layer by layer — inputs, each neuron's b/z/a, every edge weight and its product, output and loss — and you can copy it or export it as CSV.

Ten languages: English, French, German, Italian, Japanese, Korean, Russian, Spanish, Simplified Chinese and Traditional Chinese.