On Statistical Bias In Active Learning: How and When to Fix It | OpenReview

On Statistical Bias In Active Learning: How and When to Fix It | OpenReview

Active learning is a powerful tool when labelling data is expensive, but it introduces a bias because the training data no longer follows the population distribution. We formalize this bias and...

3 mentions: @seb_far@tom_rainforth@ottamm_190
Date: 2021/05/04 04:48

Referring Tweets

@ottamm_190 プールベース能動学習における統計的バイアスを取り除くため、不偏推定量を二つ提案。逐次的に情報量の多いデータにラベル付けすると、訓練データが母集団分布から抽出されたものと乖離。このサンプリングバイアスに言及した論文は少ない。線形回帰の例が分かりやすい。 t.co/Cf1tN7v8q7 t.co/vQvBPfocrV
@seb_far Did you know active learning introduces bias? In #ICLR2021 spotlight we show how to fix this and also when not to! With @yaringal @tom_rainforth Poster 9amPST/5pmUK today. t.co/gWQhTZ4Kar Watch live in Oral Session 2 with Q&A: 2045 UK. Paper: t.co/bqipB2lB7a
@tom_rainforth I'm looking forward to presenting 4 papers at @iclr_conf this week with my coauthors: - On Statistical Bias in Active Learning: How and When to Fix It t.co/zx9m9jo4KK Poster Mon 5pm BST, Talk 8.45pm BST with @seb_far @yaringal t.co/PDJTEQljgj

Related Entries

Read more [最新版] JSAI2018 チュートリアル「"深層学習時代の" ゼロから始める自然言語処理」
0 users, 0 mentions 2018/06/10 19:00
Read more Normal Distribution Demonstration (Bean Machine script for Physion) - YouTube
0 users, 0 mentions 2020/08/21 02:21
Read more [2012.02338] An optimal quantum sampling regression algorithm for variational eigensolving in the lo...
0 users, 1 mentions 2020/12/14 14:21
Read more [1902.09700] Learning to Sample Hard Instances for Graph Algorithms
0 users, 1 mentions 2021/01/01 11:21
Read more TGS Salt Identification Challenge | Kaggle
1 users, 1 mentions 2021/03/31 14:22