Home /
Research
Showing 151 - 156 / 904
Information Maximization for Few-Shot Learning
MalikBoudiaf, ImtiazZiko, JérômeRony....
Published date-12/01/2020
Few-ShotLearning, Meta-Learning
We introduce Transductive Infomation Maximization (TIM) for few-shot learning. Our method maximizes the mutual information between the query features and their label predictions for a given few-shot task, in conjunction …
Evaluating Attribution for Graph Neural Networks
BenjaminSanchez-Lengeling, JenniferWei, BrianLee....
Published date-12/01/2020
Interpretability of machine learning models is critical to scientific understanding, AI safety, as well as debugging. Attribution is one approach to interpretability, which highlights input dimensions that are influential to …
Robust compressed sensing using generative models
AjilJalal, LiuLiu, AlexandrosG.Dimakis....
Published date-12/01/2020
We consider estimating a high dimensional signal in $\R^n$ using a sublinear number of linear measurements. In analogy to classical compressed sensing, here we assume a generative model as a …
Soft Contrastive Learning for Visual Localization
JanineThoma, DandaPaniPaudel, LucV.Gool....
Published date-12/01/2020
ContrastiveLearning, ImageRetrieval, VisualLocalization
Localization by image retrieval is inexpensive and scalable due to simple mapping and matching techniques. Such localization, however, depends upon the quality of image features often obtained using Contrastive learning …
Barking up the right tree: an approach to search over molecule synthesis DAGs
JohnBradshaw, BrooksPaige, MattJ.Kusner....
Published date-12/01/2020
When designing new molecules with particular properties, it is not only important what to make but crucially how to make it. These instructions form a synthesis directed acyclic graph (DAG), …
The Advantage of Conditional Meta-Learning for Biased Regularization and Fine Tuning
GiuliaDenevi, MassimilianoPontil, CarloCiliberto....
Published date-12/01/2020
Meta-Learning
Biased regularization and fine tuning are two recent meta-learning approaches. They have been shown to be effective to tackle distributions of tasks, in which the tasks’ target vectors are all …