MLD3: Machine Learning for Data-Driven Decisions, University of Michigan
Code repository for work by the MLD3 lab.
Ann Arbor, MI
Pinned Repositories
AMAISE
The source code for AMAISE: A Machine Learning Approach to Index-Free Sequence Enrichment and the accession codes for the data used to train and test AMAISE.
CounterfactualAnnot-SemiOPE
[NeurIPS 2023] Counterfactual-Augmented Importance Sampling for Semi-Offline Policy Evaluation. https://arxiv.org/abs/2310.17146
Deep-Learning-Applied-to-Chest-X-rays-Exploiting-and-Preventing-Shortcuts
[MLHC 2020] Deep Learning Applied to Chest X-Rays: Exploiting and Preventing Shortcuts (Jabbour, Fouhey, Kazerooni, Sjoding, Wiens). https://arxiv.org/abs/2009.10132
Deep-Residual-Time-Series-Forecasting
Implementation of architecture for 2020 OhioT1D competition submission. Includes weights from pre-training runs with Tidepool data set. Baseline architecture is N-BEATS, modifications include RNN/shared output blocks, additional Losses. https://folk.idi.ntnu.no/kerstinb/kdh/KDH_ECAI_2020_Proceedings.pdf
FIDDLE
FlexIble Data-Driven pipeLinE – a preprocessing pipeline that transforms structured EHR data into feature vectors to be used with ML algorithms. https://doi.org/10.1093/jamia/ocaa139
FIDDLE-experiments
Experiments applying FIDDLE on MIMIC-III and eICU. https://doi.org/10.1093/jamia/ocaa139
M-CURES
"Early identification of patients admitted to hospital for covid-19 at risk of clinical deterioration: model development and multisite external validation study" (Kamran, Tang, et al.), BMJ 2022.
OfflineRL_FactoredActions
[NeurIPS 2022] Leveraging Factored Action Spaces for Efficient Offline RL in Healthcare.
RL-Set-Valued-Policy
[ICML 2020] Clinician-in-the-Loop Decision Making: Reinforcement Learning with Near-Optimal Set-Valued Policies. https://arxiv.org/abs/2007.12678, https://icml.cc/virtual/2020/poster/5797
RL4BG
Public code release for "Deep Reinforcement Learning for Closed-Loop Blood Glucose Control" (Ian Fox et al.), MLHC 2020. https://arxiv.org/abs/2009.09051
MLD3: Machine Learning for Data-Driven Decisions, University of Michigan's Repositories
MLD3/FIDDLE
FlexIble Data-Driven pipeLinE – a preprocessing pipeline that transforms structured EHR data into feature vectors to be used with ML algorithms. https://doi.org/10.1093/jamia/ocaa139
MLD3/FIDDLE-experiments
Experiments applying FIDDLE on MIMIC-III and eICU. https://doi.org/10.1093/jamia/ocaa139
MLD3/M-CURES
"Early identification of patients admitted to hospital for covid-19 at risk of clinical deterioration: model development and multisite external validation study" (Kamran, Tang, et al.), BMJ 2022.
MLD3/OfflineRL_FactoredActions
[NeurIPS 2022] Leveraging Factored Action Spaces for Efficient Offline RL in Healthcare.
MLD3/Deep-Learning-Applied-to-Chest-X-rays-Exploiting-and-Preventing-Shortcuts
[MLHC 2020] Deep Learning Applied to Chest X-Rays: Exploiting and Preventing Shortcuts (Jabbour, Fouhey, Kazerooni, Sjoding, Wiens). https://arxiv.org/abs/2009.10132
MLD3/OfflineRL_ModelSelection
[MLHC 2021] Model Selection for Offline RL: Practical Considerations for Healthcare Settings. https://arxiv.org/abs/2107.11003
MLD3/ARDS_PLOS_ONE_2019
Machine Learning for Patient Risk Stratification for Acute Respiratory Distress Syndrome (Zeiberg & Prahlad et al.), PLOS ONE, March 2019. https://doi.org/10.1371/journal.pone.0214465
MLD3/Calibrated-Survival-Analysis
Code Release for "Estimating Calibrated Individualized Survival Curves with Deep Learning" (Kamran & Wiens), AAAI 2021. https://www.aaai.org/AAAI21Papers/AAAI-8472.KamranF.pdf
MLD3/Combining-chest-X-rays-and-EHR-data-ARF
MLD3/Hierarchical_Survival_Analysis
MLD3/DCEM
[ICML 2024] From Biased Selective Labels to Pseudo-Labels: An Expectation-Maximization Framework for Learning from Biased Decisions
MLD3/JCO_CCI_aGVHD_prediction
Predicting Acute Graft-versus-Host Disease Using Machine Learning and Longitudinal Vital Sign Data from Electronic Health Records (Tang et al.), JCO Clinical Cancer Informatics 2020. https://doi.org/10.1200/CCI.19.00105
MLD3/sparse-informative-variables
MLD3/AMAISE
The source code for AMAISE: A Machine Learning Approach to Index-Free Sequence Enrichment and the accession codes for the data used to train and test AMAISE.
MLD3/CounterfactualAnnot-SemiOPE
[NeurIPS 2023] Counterfactual-Augmented Importance Sampling for Semi-Offline Policy Evaluation. https://arxiv.org/abs/2310.17146
MLD3/DEPICT
MLD3/disparate_censorship
MLD3/MILwAPI
Code and Additional Information for "Multiple Instance Learning with Absolute Position Information"
MLD3/PROP-RL
Optimizing Loop Diuretic Treatment in Hospitalized Patients: A Case Study in Practical Application of Offline Reinforcement Learning to Healthcare
MLD3/mld3.github.io
Lab website
MLD3/AD_from_BP
Predicting Alzheimer's disease onset using blood pressure trajectories
MLD3/ADTRCI_AD_from_EHR
Code for the paper "Cohort discovery and risk stratification for AD: an EHR-based approach" in Alzheimer's and Dementia: TRCI
MLD3/AJS_Opioids_Use_Prediction
Predicting Postoperative Opioid Use with Machine Learning and Insurance Claims in Opioid-Naïve Patients (Hur, Tang, ..., Waljee, Wiens). The American Journal of Surgery, 2021. https://doi.org/10.1016/j.amjsurg.2021.03.058
MLD3/ConceptCredibleModel
MLD3/credible_learning
[KDD 2018] Learning Credible Models
MLD3/denoising-autoencoders-for-learning-from-noisy-patient-reported-data
MLD3/gaming_detection
[NeurIPS 2024] Who’s Gaming the System? A Causally-Motivated Approach for Detecting Strategic Adaptation
MLD3/Instance_Dependent_Label_Noise
Code for "Leveraging an Alignment Set in Tackling Instance-Dependent Label Noise" in CHIL 2023
MLD3/Learning-to-Rank-for-Optimal-Treatment-Allocation-Under-Resource-Constraints
MLD3/SAMultipleNoisyLabels
Survival Analysis with Multiple Noisy Labels