Osman Ali Mian
Osman Ali Mian
IKIM - the Institute for Artificial Intelligence in Medicine
Previously CISPA Helmholtz Center For Information Security
Embedded Files
Email: [fullname] [at] outlook [dot] com
The links below provide access to code implementations for causal structure discovery across a range of practical scenarios. These methods assume the absence of unobserved confounders, meaning that all causal factors influencing each variable in the dataset are themselves included in the dataset.
Single Dataset
Single Dataset
Use GLOBE if you want to find a causal structure from a single, complete dataset.
MULTIPLE Datasets
MULTIPLE Datasets
Use ORION if you have multiple datasets over same variable set, from different sources.
PRIVATE Datasets
PRIVATE Datasets
Use PERI if you have distributed datasets that are private and can not be shared. This code is currently only a proof-of-concept.
Continual Data
Continual Data
If your data arrives in chunks over time and you want to update causal structure from where you left off, use CONTINENT.
Timeseries Data
Timeseries Data
CaDYT can help you discover causal structure from multivariate, potentially irregularly-sampled, timeseries data.
Data Completion
Data Completion
Use LoGIC if your dataset is incomplete and you would like to fill it in while discovering and respecting the underlying causal structure.
BI-VARIATE Data with Heterogeneous noise
BI-VARIATE Data with Heterogeneous noise
If you want to decide causality between two variables without assuming homogeneous distribution of errors, HECI is the right tool.
Your IDEA?
Your IDEA?
Reach out if you have a setting that you would like to explore but is not listed here.
Google Sites
Report abuse