Herbert Gish — accepted AI-conference papers

2020

Towards a New Understanding of the Training of Neural Networks with Mislabeled Training Data

Herbert Gish, Jan Silovský, Man-Ling Sung, Man-Hung Siu, William Hartmann, Zhuolin Jiang

ICASSP 2020accepted

We investigate the problem of machine learning with mislabeled training data. We try to make the effects of mislabeled training better understood through analysis of the basic model and equations that characterize the problem. This includes results about the ability of the noisy model to make the sa…

Cited by 0 Source ↗Scholar ↗

2016

Importance sampling of delta-AUC: A basis for active learning for improved keyword search

Kerri Barnes, Matthew Snover, Man-Hung Siu, Herbert Gish

ICASSP 2016accepted

We present an importance sampling based approach to the active learning problem of selecting additional training data to supplement a seed model. Our proposed Δ-AUC selection optimizes AUC improvement in keyword search and is evaluated on the Spanish Fisher corpus. We show that over different traini…

Cited by 0 Source ↗Scholar ↗