Session 3: A New Paradigm for Learning Distribution Shift @ColumbiaSEAS
Session 3: A New Paradigm for Learning Distribution Shift  @ColumbiaSEAS
Uploaded June 2025 | Updated November 2026, 12 hours ago
By Adam Klivans, University of Texas at Austin: We revisit the fundamental problem of learning with distribution shift, where a learner is given labeled samples from training distribution D, unlabeled samples from test distribution D′ and is asked to output a classifier with low test error. The standard approach in this setting is to prove a generalization bound in terms of some notion of distance between D and D′. These distances, however, are difficult to compute, and this has been the main stumbling block for efficient algorithm design.We sidestep this issue and define a new model called TDS learning, where a learner runs a test on the training set and is allowed to reject if this test detects distribution shift reative to a fixed output classifier. Moreover, when the test accepts, the output classifier is guaranteed to have low test error. We describe how this approach leads to the first set of efficient algorithms for learning with distribution shift that do not take any assumptions on the test distribution. Our techniques touch on a wide array of topics including pseudorandomness, property testing, and sum of squares proofs.Joint work with Konstantinos Stavropoulos and Arsen Vasilyan
Session 3: A New Paradigm for Learning Distribution ShiftCEEC Symposium: Electrolysis at Scale, from Hydrogen to Critical MaterialsSymposium on AI & Sports | Lightning Talks, Part 1SESSION 5: Columbia CryptoEconomics Workshop 2024Senior Design Expo 2025CCAE 2025: Practical & Theoretical Trends in Research, Development & Design of Electrical DrivesWhy Fusion Was Always 30 Years AwaySports AI Innovation Symposium | Opening KeynoteCCE Day 2: Session 7 - Lightning TalksSESSION 4: Columbia CryptoEconomics Workshop 2024Columbia Space Initiative Reaches New HeightsCCE Plenary Session 2
Columbia Engineering |

Session 3: A New Paradigm for Learning Distribution Shift

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER