Code
This page contains links to code for some recent projects from the Yu group (see specific papers for other code repositories). Going forward, code will be added to the Yu-group github.Multipurpose data-science libraries
- VeridicalFlow🚰: a library for building stable, trustworthy data-science pipelines based on the PCS framework
- simChef 🍳: An R package to facilitate PCS simulation studies
- imodels 🔎: a python package for fitting interpretable models
Stable modeling
- iRF: iterative Random Forests (Python package, R package) (PCS-guided)
- siRF (GitHub code repository)
- lo-siRF (GitHub code repository)
- escv selection of regularization for lasso for sparser and more interpretable models. (paper) (first work related to PCS)
- stanmf: stability driven non-negative matrix factorization (PCS-guided)
- Causal inference: X-learner: CATE prediction, staDISC: discovering stable subgroups (PCS-guided)
Interpretation / Visualization
- Mechanistic interpretation of neural networks 🔧: SUFO: embedding space interpretations, CD-T: circuit discovery in transformers
- Input feature attribution of neural networks 🧠: ACD: hierarchical interpretations, TRIM: interpreting transformations, CDEP: penalizing explanations , SPEX: scaling feature interaction explanations for LLMs
- Adaptive wavelets 🌊: AWD: adaptive wavelets and wavelet distillation
- Visualization 🔥: Superheat: An r package for generating beautiful and customizable heatmaps
- Epistasis regulates genetic control of cardiac hypertrophy: preprint, GitHub code repository, PCS documentation (PCS-guided)
- epiTree — Learning epistatic polygenic phenotypes with boolean interactions (PCS inference)
- Extensive and accessible COVID-19 data + forecasting at the county-level + hospital-level.
- staDRIP: drug-response prediction (PCS-guided)