CDANs
Constraint-based causal discovery with optimized conditioning sets, designed to handle autocorrelation, non-stationarity, and time-varying causal structure in multivariate time series. The package implements the methods described in our MLHC 2023 paper with a refined API and an emphasis on resource efficiency suitable for bedside monitoring and edge deployment.
pip install cdans
Reference: Ferdous, Hasan, and Gani. CDANs: Temporal Causal Discovery from Autocorrelated and Non-Stationary Time Series Data. MLHC 2023.
Stable release on PyPI
TimeGraph
A synthetic benchmark suite for time series causal discovery. The suite simulates realistic temporal complexities, including controlled causal structures, calibrated autocorrelation, non-stationarity, multiple noise distributions, and seasonal patterns. The intent is to allow honest comparison across methods and to expose the conditions under which each method succeeds or fails.
Reference: Ferdous, Hossain, and Gani. TimeGraph: Synthetic Benchmark Datasets for Robust Time-Series Causal Discovery. KDD 2025.
Published at KDD 2025
ClassyGlass
A benchmark dataset for activity and mobility analysis using smart eyewear. The dataset provides annotated multimodal sensor data and is intended to support the evaluation of causal and predictive methods in human activity recognition and health monitoring.
Reference: Mahmud, Emmert, Ferdous, et al. ClassyGlass: A Benchmark Dataset for Activity and Mobility Analysis using Smart Eyewear. Submitted to KDD 2026 Dataset Track.
Under review