Validation and inference
Advanced tools for analysing, inferring, and validating models against various datasets, scales, and species.
EBRAINS offers cutting-edge tools for scientists, including a model validation web service, Python libraries based on Elephant for statistical analysis of electrophysiological data, Frites for information-theoretic analysis and network-level statistics, and TVB-Inverse for Monte Carlo simulations in a Bayesian framework.
Elephant
Frites - Framework for information theoretical analysis of electrophysiological data and statistics
TVB-Inverse
This tool offers powerful methods that enable principled solutions to complex inverse problems and statistical inference at meso to macro scales, using invasive and non-invasive recordings. It leverages Bayesian inference to uncover the posterior distribution of TVB's model parameters at the whole-brain level, such as the spatial map of excitability or the degree of degradation in a personalised connectome.
TVB-inverse combines fast simulations (such as JAX, JIT, and C++) of virtual brains with state-of-the-art Monte Carlo sampling and probabilistic AI/ML algorithms, enabling reliable, efficient, and flexible causal inference at the whole-brain scale in both CPU and GPU. With TVB-inverse, scientists can efficiently perform causal inference when the experiments are difficult or impossible, making it an essential tool for researchers in neuroscience with clinical applications.
A set of Jupyter notebooks takes you step-by-step through the entire workflow:
- Full Bayesian inference with adaptive Hamiltonian Monte Carlo sampling (NUTS)
- Approximate Bayesian inference with automatic variational inference (ADVI)
- Amortised posterior estimation with simulation-based inference (SBI)
- Maximum posterior estimation with optimisation techniques (L-BFGS)
- Dynamical system identification with variational autoencoders (VAEs)
Create an account
EBRAINS is open and free. Sign up now for complete access to our tools and services.