Active Learning Literature Survey¶
Settles, B. (2009). Active Learning Literature Survey. University of Wisconsin–Madison.
Cited by¶
5 citations across 4 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Theoretical Sampling
- Active learning in machine learning. A model selects which unlabeled examples to query by expected information gain — uncertainty sampling, query-by-committee, expected model change — with the model's current state determining which next examples are most informative.
This sourceSurvey of active learning (uncertainty sampling, query-by-committee, expected model change) as model-driven selection of the next observation; traces its lineage to statistical optimal experimental design, not to grounded theory.
SupportedVerified against the work's full text
Settles's survey names uncertainty sampling, query-by-committee and expected model change as strategies selecting the most informative unlabeled instance the model is least certain about.
“There are several scenarios in which active learners may pose queries, and there are also several different query strategies that have been used to decide which instances are most informative. In this section, I present two illustrative examples in the pool-based active learning setting (in which queries are selected from a large pool of unlabeled instancesU) using an uncertainty sampling query strategy (which …”
- Structurally identical, and apparently arrived at independently: the active-learning literature traces its own ancestry to statistical theories of optimal experimental design rather than to any qualitative-research antecedent, so on the evidence of its standard survey the pair is a convergence on the same control loop rather than a borrowing.
This sourceUniversity of Wisconsin–Madison Computer Sciences Technical Report 1648, 2009 — "This approach is derived from statistical theories of optimal experimental design (Federov, 1972)." Contradicts an earlier claim that the active-learning literature explicitly cites theoretical sampling as a conceptual ancestor: the strings "theoretical sampling", "grounded theory", "Glaser", "Strauss", "sociology", "qualitative" and "ethnograph" do not occur anywhere in the survey, which traces the lineage to optimal experimental design and PAC learning. No other source establishing the asserted citation lineage was found. This key is scoped to the lineage claim
Supported in partVerified against the work's full text
Settles's survey derives one query-synthesis strategy from "statistical theories of optimal experimental design", an OED lineage for one method family — not the claim's literature-wide negative about qualitative antecedents.
“This sort of approach is derived from statistical theories of optimal experi- mental design, or OED (Federov, 1972; Chaloner and Verdinelli, 1995).”
- Active learning in machine learning. A model selects which unlabeled examples to query by expected information gain — uncertainty sampling, query-by-committee, expected model change — with the model's current state determining which next examples are most informative.
Domain-specific¶
- Hempel's Paradox
- Empirically, active learning can reach a target accuracy with far fewer labels than random sampling
This sourceSurvey evidence that active learning reaches a target accuracy with far fewer labels than random sampling.
- Empirically, active learning can reach a target accuracy with far fewer labels than random sampling
- Proactive learning
Mechanisms¶
- Adaptive Refinement Loop
- Chasing informative gaps this way is the coverage analogue of active learning
This sourceDefines active learning as selecting queries or unlabeled instances expected to be most informative to the learner.
- Chasing informative gaps this way is the coverage analogue of active learning
Verification¶
Does it exist? Not checked yet. This entry carries no identifier to resolve. It was extracted from the citation as written in the article, normalized, and deduplicated against the rest of the registry.
Does it back the claim? Read against the text for 2 of 5 citations: 1 supported, 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Links previously used in the corpus¶
Before the registry existed this work was also linked 2 other ways.
- https://burrsettles.com/pub/settles.activelearning.pdf ×1
- https://research.cs.wisc.edu/techreports/2009/TR1648.pdf ×1
Registry ID ref:28647c96d71c · see in the full table