FCAI Software

Building AI can be easy –
with the right tools

AI research enters practice in the form of software. FCAI delivers open-source tools for everyone, so that they can make use of the new ideas easily. This page shows examples of software ready for general use, as well as implementations of recent cutting edge research that are useful for other AI researchers.

Highlighted software

We contribute to, or lead in the development of, several established software for probabilistic programming, likelihood-free inference, and deep learning. These tools are ready for a broad use in academia and industry alike.

Software for general use

We contribute to a broad range of software libraries that can be used for developing AI solutions.

• BADS: Fast Bayesian optimization for fitting computational models

• Bambi: Python library for Bayesian generalized linear multilevel models

• bayesforecast: Automatic forecasting and Bayesian modeling for time series with Stan

• BOSS: Bayesian optimization for materials science

• brms: R library for Bayesian generalized (non-)linear multivariate multilevel models

• lgpr: R package for Longitudinal Gaussian Process Regression (build on top of Stan)

• loo: Model validation in Stan ecosystem

• posterior: Cross-platform tools for working with outputs of MCMC methods

• posteriordb: Database of models, data, and reference results for testing and benchmarking inference methods and their implementations

• preliz: Prior elicitation for probabilistic programming

• priorsense: Tools for prior diagnostics and sensitivity analysis for Stan models

• projpred: Feature selection for generalized linear models

• twinify: Privacy-preserving generation of synthetic data

• User-in-the-box: Modelling and simulating biomechanical user models in HCI tasks

• VAI-Lab: Virtual laboratory platform for AI-assisted research and design

• χiplot: Web-based visualisation platform for multidimensional data

Research software

We are currently building several tools that are still primarily meant for other AI researchers, but many of which will later mature for general use. See the web pages of individual research groups for more examples.

• Amortized inference with user simulations: Computationally efficient tools for explaining user behavior

• d3p: Differentially private probabilistic programming

• Deep Automodulators: Generative modeling with automodulators

• DGC-Net: Geometric methods for pixel correspondence estimation

• E-LPIPS: Robust image similarity evaluation

• EntityBot: An assistant for recommending information entities in everyday tasks

• Fourier Accountant: Tight privacy accounting for differential privacy

• GANSpace: Interpretable control for GAN-based image generation

• Ice-BeeM: Contrastive estimation for nonlinear ICA

• HSC-Net: Single-image localization of 3D objects

• MRPT: Fast approximate nearest neighbor search

• Multi-FinGAN: Multi-fingered robotic grasping based on single RGB image

• PIONEER: Pioneer networks for generative modeling

• SLISEMAP: Dimensionality reduction with local explanations

• ViTa-SLAM: Biologically inspired visuo-tactile SLAM