I’m happy to announce SubspaceInference.jl , a fast implementation to generate uncertainties in Neural Networks and Neural ODE’s
Scaling of Bayesian inference techniques to deep neural networks (Bayesian Neural Networks) and Neural ODE’s are challenging due to the high dimensionality of the parameter space. This Julia package provides a faster implementation of Bayesian inference calculation with NN and NODE’s using subspace inference techniques.
Contributing
Contributions to SubspaceInference.jl are most welcome. Please do not hesitate to open a Github issue to share any idea, feedback or suggestion!