TrustyAI Explainability Toolkit

Supporting models and data to doi 10.1021/acs.jcim.1c01163

OrganismCore transforms reasoning into executable artifacts built on the Universal Reasoning Substrate (URS). Its purpose is to accelerate discovery by making reasoning itself a programmable, transmissible, and model-agnostic object.

This repository contains a code of Unfold and Conquer Attribution Guidance, which is presented in a conference of Association for the Advancement of Artificial Intelligence 2023.

Ths repo has the list of Interesting Literature in the domain of XAI

Re-Implementation of COVID-NET in PyTorch

Hacking a Neural Network to understand what concepts the network learns in order to solve a logic task.

Skin cancer classification using Transfer Learning and explainable AI

Methods to interpret machine learning models/black box which can help us understand how it’s making decisions.

See the world through the eyes of AI

An AI-powered clinical assistant using Retrieval-Augmented Generation (RAG) on the MIMIC-IV DiReCT dataset. It retrieves relevant patient cases and generates diagnostic reasoning using LLMs. Built with Streamlit, Transformers, FAISS, and SentenceTransformers.

Detects Pneumonia using Chest X rays through Deep Learning models

GUI-based ransomware attack detection using processor and disk I/O telemetry with CNN2D classification, SHAP explainability, adversarial robustness testing, and 10-fold stratified cross-validation.

Attention-guided convolutional autoencoder for one-class anomaly detection and localization on CIFAR-10, using CBAM and reconstruction-based scoring.

A comprehensive comparative study of 10+ feature selection techniques (including RFE and SHAP) to optimize ML models. Achieved a 73% reduction in feature space while maintaining >96% accuracy, highlighting key trade-offs between performance efficiency and model interpretability for production environments.

experimental setup to hook llms to a logic engine, for proof traces and financial math