Sen(Qian)’s Memo

This website is Donglin Qian (Torin Sen)’s memo, especially about machine learning papers and competitive programming.

Self-supervised

2024-07-27

PU Paper Self-supervised Curriculum Learning Self-Training Self-Paced Learning Knowledge Distillation Memorization Effect

2020-ICML-[Self-PU]Self Boosted and Calibrated Positive-Unlabeled Training

2024-07-01

PU Curriculum Learning Self-Training Sample-Selection Self-supervised Self-Paced Learning Paper Detail Article Case-Control

2023-KDD-[RobustPU]Robust Positive-Unlabeled Learning via Noise Negative Sample Self-correction

典型的なCurriculum Learningを導入したPU。

2024-06-14

Self-supervised Data Augmentation

2020-NIPS-[FixMatch]Simplifying Semi-Supervised Learning with Consistency and Confidence

Data Augmentationして損失に導入する。ここで、強い拡張と弱い拡張の両方を考えてみる。 1. ラベルありデータの損失 2. ラベルなしデータに弱い拡張を施し予測されたラベルをhard labelにする。そのhard labelと同じラベルありデータに強い拡張を施したものの損失を計算する。弱い拡張は左右反転とか平行移動。強い拡張はモザイクとか

2024-06-03

Domain Shift Self-supervised Detail Article 一旦放置

2023-ICML-Bidirectional Adaptation for Robust Semi-Supervised Learning with Inconsistent Data Distributions

2024-01-25

Database Self-supervised NLP Scraping weakly-supervised-learning Paper

2023-Proceedings of the VLDB Endowment-Self-training for Label-efficient information extraction from semi-structured web-pages