Features
Self-supervised contrastive learning for time-series
Supervised embedding using auxiliary behavioral labels
Hybrid hypothesis- and discovery-driven modes
Consistency metrics for comparing latent spaces
k-nearest neighbor decoding from embeddings
Supports calcium imaging and electrophysiology data
Integration with DeepLabCut for pose embeddings
scikit-learn-compatible API
Built-in plotting with matplotlib and plotly
Time-series attribution maps (AISTATS 2025)
Multi-session and multi-animal data support
GPU-accelerated training via PyTorch
Docker container for reproducible analysis
Natural language query input
Step-by-step solutions (Pro)
Algorithmic computation engine
Curated knowledgebase across domains
Visual plots and graphics
Unit and measure conversions
Physics and chemistry problem solver
Statistical analysis tools
Money and finance computations
Date and time calculations
Personal health and nutrition data
Mobile app for iOS and Android
Wolfram Problem Generator
