Pytorch implementations of Bayes By Backprop, MC Dropout, SGLD, the Local Reparametrization Trick, KF-Laplace, SG-HMC and more

Implementation of the MNIST experiment for Monte Carlo Dropout from http://mlg.eng.cam.ac.uk/yarin/PDFs/NIPS_2015_bayesian_convnets.pdf

This repository reimplemented "MC Dropout" by tensorflow 2.0 Eager Extension.

An experimental Python package for learning Bayesian Neural Network.

(Forked Version) Experiments used in "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning"

A transformative approach to manufacturing optimization, focusing on the textile forming process. This research synergizes domain-specific knowledge with simulation modeling and introduces Bayesian optimization for efficient parameter space exploration.

Demo codes and data for a published work "Uncertainty explanation of artificial intelligence models by SHAP"

Numerical solution and uncertainty quantification of Pennes' bioheat transfer equation in 1-D using deep neural network solver.

We provide two notebooks that enable users to explore and experiment with some BDL techniques as Ensembles, MC Dropout and Laplace Approximation. In this way, they allow you to intuitively visualize the main differences among them in a Simulated Dataset and Boston Dataset.

Attention-Guided TransUNet for multi-class retinal fluid segmentation in OCT with MC Dropout uncertainty quantification

PyTorch implementation of Last-Layer MC Dropout for epistemic uncertainty estimation in Medical AI. Automatically identifies clinical edge cases and label noise in chest X-rays.

Master thesis for the MSc. Artificial Intelligence at the Universiteit van Amsterdam, 2019

Data Drift Analysis and Anomaly detection tools

Kvantifikacija nesigurnosti u modelima umjetne inteligencije: okvir za prediktivno održavanje i analizu rizika

Deepfake detection pipeline with MC Dropout uncertainty estimation, Grad-CAM explainability, and a TrustScore metric for selective, calibrated predictions on Celeb-DF v2.

PyTorch design-space study: 5 sequence models across 3 prediction horizons for sEMG-based variable-impedance teleoperation latency compensation (synthetic data).

📊 Explore Bayesian statistics and econometrics with training materials designed for quantitative analysts and grad students in machine learning.