AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

On the Minimal Supervision for Training Any Binary Classifier from Only Unlabeled Data

Nan Lu, Gang Niu, Aditya Krishna Menon, Masashi Sugiyama

ICLR 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 49 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Train using two unlabeled samples with different class priors and the UU unbiased risk estimatorbinary classification · MNIST, Fashion-MNIST, SVHN, and CIFAR-10classification errorImproved
Use UU instead of pSVM, BER, or BER-FCbinary classification from two unlabeled sets · pendigits, covertype, MNIST, spambase, letter, USPSclassification errorMixed
Move the two training class priors closer togetherUU learning and CCN on MNIST · MNISTclassification errorHurt
Use UU rather than CCN as the class-prior gap narrowslearning from two unlabeled sets · MNISTclassification errorImproved
Perturb the training class-prior values used by UUUU · MNIST and CIFAR-10classification error: 2.31~(0.14) -> 2.35~(0.14)No clear effect
Use sigmoid loss instead of logistic lossUU · MNISTclassification errorNo clear effect
Use BER or BER-FC, which optimize balanced error rather than classification errortwo-unlabeled-set learning · datasets with class prior near 0 or 1classification errorHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy