Share This Author
WARP: Word-level Adversarial ReProgramming
This paper presents an alternative approach based on adversarial reprogramming, which extends earlier work on automatic prompt generation, and outperforms all existing methods with up to 25M trainable parameters on the public leaderboard of the GLUE benchmark.
YerevaNN’s Systems for WMT20 Biomedical Translation Task: The Effect of Fixing Misaligned Sentence Pairs
YerevaNN’s neural machine translation systems and data processing pipelines developed for WMT20 biomedical translation task are described and most of the improvements are explained by the heavy data preprocessing pipeline which attempts to fix poorly aligned sentences in the parallel data.
Robust Classification under Class-Dependent Domain Shift
This paper explores a special type of dataset shift which it is called class-dependent domain shift, characterized by the following features: the input data causally depends on the label, the shift in the data is fully explained by a known variable, the variable which controls the shift can depend on the labels.
Deep Semi-Supervised Image Classification Algorithms: a Survey
Most of the recently proposed deep semi-supervised learning algorithms for image classification and the main trends of research in the field are described and identified.
Improving VAE based molecular representations for compound property prediction
This work proposes a simple method to improve chemical property prediction performance of machine learning models by incorporating additional information on correlated molecular descriptors in the representations learned by variational autoencoders and proves the method on three property prediction tasks.
Failure Modes of Domain Generalization Algorithms
- T. Galstyan, Hrayr Harutyunyan, H. Khachatrian, G. V. Steeg, A. Galstyan
- Computer Science
- 26 November 2021
This paper proposes an evaluation framework for domain generalization algorithms that allows decomposition of the error into components capturing distinct aspects of generalization, and shows that the largest contributor to the generalization error varies across methods, datasets, regularization strengths and even training lengths.