1 / 16

Cause-Effect Pair Challenge

IJCNN 2013 IEEE/INNS. Cause-Effect Pair Challenge. Isabelle Guyon, ChaLearn. …your health?. …climate changes?. … the economy?. Causal discovery. What affects…. Which actions will have beneficial effects?. Available data. A lot of “observational” data.

kmarion
Download Presentation

Cause-Effect Pair Challenge

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. IJCNN 2013 IEEE/INNS Cause-Effect Pair Challenge Isabelle Guyon, ChaLearn clopinet.com/causality

  2. …your health? …climate changes? … the economy? Causal discovery What affects… Which actions will have beneficial effects? clopinet.com/causality

  3. Available data • A lot of “observational” data. Correlation  Causality! • Experiments are often needed, but: • Costly • Unethical • Infeasible clopinet.com/causality

  4. Setup • No feed-back loops. • No time. Samples are drawn randomly and independently. We consider pairs of variables {A, B} for which A B means A = f (B, noise). clopinet.com/causality

  5. Anxiety Peer Pressure Born an Even Day Yellow Fingers Smoking Genetics Allergy Lung Cancer Attention Disorder Coughing Fatigue Car Accident Causal graph example clopinet.com/causality

  6. Anxiety Peer Pressure Born an Even Day Yellow Fingers Smoking Genetics Allergy Lung Cancer Attention Disorder Coughing Fatigue Car Accident Causality assessmentwith experiments clopinet.com/causality

  7. Causality assessmentwithout experiments? • Possible to some extent, using: • Conditional independence tests, e.g. in A -> Z -> B, A <- Z <- B or A <- Z -> B, • A is independent of B given Z • but NOT in A -> Z <- B • But… • Such methods require a lot of data to work well and often rely on simplifying assumptions (e.g. “causal sufficiency”, “faithfulness”, linearity, Gaussian noise) clopinet.com/causality

  8. Cause-effect pair problem A B Smoking Lung Cancer Lung Cancer Fatigue A -> B A <- B A – B A | B Genetics Attention Disorder Lung Cancer Born an Even Day Lung Cancer clopinet.com/causality

  9. Typical method Test whether A -> B is a better explanation than A <- B comparing two models: B = f (A, noise) A = f (B, noise) clopinet.com/causality

  10. Scoring S 0 A -> B A – B or A|B A <- B • Is A a cause of B, B a cause of A, or neither? • Average two AUCs for the separations: • A -> B vs. A – B, A | B, A <- B • A <- B vs. A – B, A | B, A -> B clopinet.com/causality

  11. A ? B A -> B B =Altitude B A A = Temperature clopinet.com/causality

  12. A ? B A <- B B =Wages B A A = Age clopinet.com/causality

  13. A ? B A | B B A clopinet.com/causality

  14. A ? B A - B B A clopinet.com/causality

  15. Conclusion • Imagine…that we could find out: • what causes epidemics • what causes cancer • what causes climate changes • what causes economic changes by analyzing data constantly collected • Bring your solution or your own data! clopinet.com/causality

  16. Credits • Initial impulse: the cause-effect pair task proposed in the causality "pot-luck" challenge by Joris Mooij, Dominik Janzing, and Bernhard Schölkopf. • Protocol review, advisors and beta testers • Hugo Jair Escalante (IANOE, Mexico) 
 • Seth Flaxman (Carnegie Mellon University, USA)
 • Mikael Henaff (New York University, USA) 
 • Dominik Janzing (Max Plank Institute of Biological cybernetics, Germany) 
 • Florin Popescu (Fraunhofer Institute, Berlin, Germany) 
 • Bernhard Schoelkopf (Max Plank Institute of Biological cybernetics, Germany) 
 • Peter Spirtes (Carnegie Mellon University, USA) 
 • Alexander Statnikov (New York University, USA) 
 • Ioannis Tsamardinos (University of Crete, Greece) 
 • Jianxin Yin (University of Pennsylvannia, USA) 
 • Kun Zhang (Max Plank Institute of Biological cybernetics, Germany) • Vincent Lemaire (Orange, France) • Data and code preparation • Isabelle Guyon (ChaLearn, USA) • Alexander Statnikov (New York University, USA) • Mikael Henaff (New York University, USA) clopinet.com/causality

More Related