Normalization layers are all that sharpness-aware minimization needs
Sharpness-aware minimization (SAM) was proposed to reduce sharpness of minima and
has been shown to enhance generalization performance in various settings. In this work we …
has been shown to enhance generalization performance in various settings. In this work we …
[HTML][HTML] Few-shot and meta-learning methods for image understanding: a survey
K He, N Pu, M Lao, MS Lew - International Journal of Multimedia …, 2023 - Springer
State-of-the-art deep learning systems (eg, ImageNet image classification) typically require
very large training sets to achieve high accuracies. Therefore, one of the grand challenges is …
very large training sets to achieve high accuracies. Therefore, one of the grand challenges is …
Guiding the last layer in federated learning with pre-trained models
G Legate, N Bernier, L Page-Caccia… - Advances in …, 2024 - proceedings.neurips.cc
Federated Learning (FL) is an emerging paradigm that allows a model to be trained across a
number of participants without sharing data. Recent works have begun to consider the …
number of participants without sharing data. Recent works have begun to consider the …
Simulated annealing in early layers leads to better generalization
Recently, a number of iterative learning methods have been introduced to improve
generalization. These typically rely on training for longer periods of time in exchange for …
generalization. These typically rely on training for longer periods of time in exchange for …
Visual domain bridge: A source-free domain adaptation for cross-domain few-shot learning
M Yazdanpanah, P Moradi - Proceedings of the IEEE/CVF …, 2022 - openaccess.thecvf.com
Due to the covariate shift, deep neural networks performance always degrades when
applied to novel domains. In order to mitigate this problem, domain adaptation techniques …
applied to novel domains. In order to mitigate this problem, domain adaptation techniques …
Disentangled feature representation for few-shot image classification
Learning the generalizable feature representation is critical to few-shot image classification.
While recent works exploited task-specific feature embedding using meta-tasks for few-shot …
While recent works exploited task-specific feature embedding using meta-tasks for few-shot …
Semantic-aware graph matching mechanism for multi-label image recognition
Multi-label image recognition aims to predict a set of labels that present in an image. The
key to deal with such problem is to mine the associations between image contents and …
key to deal with such problem is to mine the associations between image contents and …
Weakly correlated distillation for remote sensing object recognition
W Zhao, X Lv, H Wang, Y Liu, Y He… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
Remote sensing object labels require high specialization, resulting in a limited number of
labeled samples. Without large labeled samples to support training, general remote sensing …
labeled samples. Without large labeled samples to support training, general remote sensing …
A survey on cross-domain few-shot image classification
Due to the limited availability of labelled data in many real-world scenarios, we have to
resort to data from other domains to improve models' performance, which prompts the …
resort to data from other domains to improve models' performance, which prompts the …
Affine Collaborative Normalization: A shortcut for adaptation in medical image analysis
The paradigm of “pretraining-then-finetuning”(PT-FT) has been extensively explored to
enhance the performance of clinical applications with limited annotations. A major …
enhance the performance of clinical applications with limited annotations. A major …