Classifier adaptation at prediction time

By Amélie Royer, Christoph H. Lampert. 2015

68 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning, Meta & Continual Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Representation & Self-Supervised Learning

In Meta & Continual Learning

Citations in the corpus

iCaRL: Incremental Classifier and Representation Learning (2016, methodology): “Royer and Lampert [25] adapt classifiers to a time-varying data stream but their method cannot handle newly appearing classes.”
vCLIMB: A Novel Video Class Incremental Learning Benchmark (2022, background): “Under these conditions, deep neural networks suffer from what is known as catastrophic forgetting [13], where the fine-tuning on novel tasks significantly reduces the performance of the model in a previously learned task, and drift [29], where unseen training data does not fit the previously estimat”
Incremental Learning in Online Scenario (2020, background): “In additional to addressing the catastrophic forgetting problem, we also need to consider concept drift for already learned classes due to the fact that data distribution in real life application may change over time in unforeseen ways [23].”
Memory Aware Synapses: Learning what (not) to forget (2017, background): “While previous works [26,31] adapt the learning system at prediction time in a transductive setting, our goal here is to build a continual system that can adapt the importance of the weights to what the system needs to remember.”
The Extreme Value Machine (2015, methodology): “Along these lines, Royer and Lampert [21] describe a method for classifier adaptation that is effective when inherent dependencies are present in the test data.”
Incremental Learning in Deep Convolutional Neural Networks Using Partial Network Sharing (2017, methodology): “Available: https://arxiv.org/abs/1802.05800 [28] A. Royer and C. H. Lampert, ‘‘Classifier adaptation at prediction time,’’ in Proc.”
A Comprehensive Survey on Test-Time Adaptation Under Distribution Shifts (2023, unknown): “Without requiring the whole dataset at test time, another line of TTPA methods [16], [566] focus on the case of online adaptation at prediction time (i.”
Continual Learning in Neural Networks (2019, background): “While previous works [116, 129] adapt the learning system at prediction time in a transductive setting, our goal here is to build a continual system that can adapt the importance of the weights to what the system needs to remember.”
Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift (2020, methodology): “Several methods for adaptation at test time have been developed speciﬁcally for dealing with label shift [54, 38, 62].”
Deep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives (2022, background): “Source-free domain adaptation is also closely related to test time adaptation, in that we encounter a single or a few test samples that are different from source domain data (Royer and Lampert, 2015; Fredericks et al., 2014; Wang et al., 2020; Liu et al., 2021l).”