Profile cover photo
Profile photo
Vitaly Feldman

Post has shared content
Call for Papers: Second Workshop on Adaptive Data Analysis
As part of NIPS 2016, we will be running the second annual workshop on adaptive data analysis. Last year's workshop was a big hit. As a new addition this year, we are soliciting submitted contributions in addition to invited speakers. The call for papers is...

Post has shared content

Post has attachment
Some people get to see this every day

Post has attachment
Will trade quality reviews and a great audience on June 23-26, 2016 in NYC for your strong submissions on machine learning theory. Submit by Feb 12!

Post has attachment
Looking forward to a tutorial on "Large-Scale Distributed Systems for Training Neural Networks" tomorrow at NIPS to be given by +Chandra Chekuri (as per page 8 of the booklet Chandra, whether or not you are going to give the tutorial, you should prepare for the avalanche of consulting requests to follow.

Post has shared content
A hiring spree at our IBM Research theory group this year: we are looking for a staff member, a postdoc and a few interns

The theory group at IBM Almaden Research is recruiting!

We anticipate positions for:
-- a research staff member;
-- a post-doctoral researcher;
-- summer research interns (don't know how many yet).

For more information, please see

Post has shared content
Excited to announce the first workshop on understanding of overfitting/generalization in adaptive data analysis (a.k.a. WADAPT). If you are in Montreal next week you are welcome to join. At the very least you'll get to enjoy the sight of machine learners, statisticians and theoretical computer scientists trying to speak the same language (and then we all know you'll need a break from hearing about deep learning).
+Adam Smith +Moritz Hardt +Aaron Roth
Ever get the nagging feeling that overfitting sneaks into every corner of modern data science? That statisticians don't know how to analyze adaptive procedures well? That "p-hacking" happens unintentionally but frequently when data are reused across studies?

You are not alone.  

Workshop on Adaptive Data Analysis
December 11, 2015
@NIPS 2015 in Montréal, Québec, Canada.

Adaptive data analysis is the increasingly common practice by which insights gathered from data are used to inform further analysis of the same data sets. This is common practice both in machine learning, and in scientific research, in which data-sets are shared and re-used across multiple studies. Unfortunately, most of the statistical inference theory used in empirical sciences to control false discovery rates, and in machine learning to avoid overfitting, assumes a fixed class of hypotheses to test, or family of functions to optimize over, selected independently of the data. If the set of analyses run is itself a function of the data, much of this theory becomes invalid, and indeed, has been blamed as one of the causes of the crisis of reproducibility in empirical science.

Recently, there have been several exciting proposals for how to avoid overfitting and guarantee statistical validity even in general adaptive data analysis settings. The problem is important, and ripe for further advances. The goal of this workshop is to bring together members of different communities (from machine learning, statistics, and theoretical computer science) interested in solving this problem, to share recent results, to discuss promising directions for future research, and to foster collaborations. The workshop will consist of several sessions of invited talks with discussions following each session and a discussion panel at the end of the day.

Invited speakers:
  Cynthia Dwork (Microsoft)
  Will Fithian (Berkeley)
  Rina Foygel Barber (U Chicago)
  Andrew Gelman (Columbia)
  Andrea Montanari (Stanford)
  Nati Srebro (TTI Chicago)
  Jonathan Ullman (Northeastern)

Further details and schedule can be found at

Vitaly Feldman (IBM Research)
Moritz Hardt (Google)
Aaron Roth (U Penn)
Adam Smith (Penn State)

Post has shared content

Post has attachment
Great to see others playing with the reusable holdout.

Post has shared content
An excellent book on theory of ML becomes available online for free. Kudos to Cambridge Press for not restricting access to the book.
Cambridge gave us permission to upload a free online copy of our "Understanding Machine Learning" book. Hope you'll enjoy it !
Wait while more posts are being loaded