class22

class22 - Active Learning 9.520 Class 22, 03 May 2006...

Info iconThis preview shows pages 1–11. Sign up to view the full content.

View Full Document Right Arrow Icon
Active Learning 9.520 Class 22, 03 May 2006 Claire Monteleoni MIT CSAIL
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Outline Motivation Historical framework: query learning Current framework: selective sampling Some recent results Open problems
Background image of page 2
Active learning motivation Machine learning applications, e.g. Medical diagnosis Document/webpage classification Speech recognition Unlabeled data is abundant, but labels are expensive. Active learning is a useful model here. Allows for intelligent choices of which examples to label. Label-complexity : the number of labeled examples required to learn via active learning Æ can be much lower than the PAC sample complexity!
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Supervised learning Given access to labeled data (drawn iid from an unknown underlying distribution P), want to learn a classifier chosen from hypothesis class H, with misclassification rate < ε . Sample complexity characterized by d = VC dimension of H. If data is separable, need roughly d/ ε labeled samples. Slide credit: Sanjoy Dasgupta
Background image of page 4
Active learning In many situations unlabeled data is easy to come by, but there is a charge for each label. What is the minimum number of labels needed to achieve the target error rate? Slide credit: S. Dasgupta
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Active learning variants There are several models of active learning: Query learning (a.k.a. Membership queries) Selective sampling Active model selection Experiment design Various evaluation frameworks : Regret minimization Minimize label-complexity to reach fixed error rate Label-efficiency (fixed label budget) We focus on classification , though regression AL exists too.
Background image of page 6
Membership queries Earliest model of active learning in theory work [Angluin 1992] X = space of possible inputs, like {0,1} n H = class of hypotheses Target concept h * H to be identified exactly. You can ask for the label of any point in X: no unlabeled data. H 0 = H For t = 1,2,… pick a point x X and query its label h * (x) let H t = all hypotheses in H t-1 consistent with (x, h * (x)) What is the minimum number of “membership queries” needed to reduce H to just {h * }? Slide credit: S. Dasgupta
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
X = {0,1} n H = AND-of-positive-literals, like x 1 x 3 x 10 S = { } (set of AND positions) For i = 1 to n: ask for the label of (1,…,1,0,1,…,1) [0 at position i] if negative: S = S {i} Total: n queries General idea: synthesize highly informative points. Each query cuts the version space -- the set of consistent hypotheses - -in ±half . Slide credit: S. Dasgupta Membership queries: example
Background image of page 8
Problem Many results in this framework, even for complicated hypothesis classes. [Baum and Lang, 1991] tried fitting a neural net to handwritten characters. Synthetic instances created were incomprehensible to humans! [Lewis and Gale, 1992] tried training text classifiers. “an artificial text created by a learning algorithm is unlikely to be a legitimate natural language expression, and probably would be uninterpretable by a human teacher.” Slide credit: S. Dasgupta
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Selective sampling [Cohn, Atlas & Ladner, 1992] Selective sampling: Given: pool (or stream ) of unlabeled examples, x, drawn i.i.d.
Background image of page 10
Image of page 11
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 11/11/2011 for the course BIO 9.07 taught by Professor Ruthrosenholtz during the Spring '04 term at MIT.

Page1 / 51

class22 - Active Learning 9.520 Class 22, 03 May 2006...

This preview shows document pages 1 - 11. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online