Category Ranking
98%
Total Visits
921
Avg Visit Duration
2 minutes
Citations
20
Article Abstract
LOBSTER (LOss-Based SensiTivity rEgulaRization) is a method for training neural networks having a sparse topology. Let the sensitivity of a network parameter be the variation of the loss function with respect to the variation of the parameter. Parameters with low sensitivity, i.e. having little impact on the loss when perturbed, are shrunk and then pruned to sparsify the network. Our method allows to train a network from scratch, i.e. without preliminary learning or rewinding. Experiments on multiple architectures and datasets show competitive compression ratios with minimal computational overhead.
Download full-text PDF |
Source |
|---|---|
| http://dx.doi.org/10.1016/j.neunet.2021.11.029 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Algorithm Selection for Estimating Causal Effects: An example using the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers to Be.
Epidemiology
August 2025
Department of Epidemiology, Emory University.
Background: The Super Learner is an ensemble learning method that has been widely used with doubly robust causal effect estimators. It is recommended to deploy the Super Learner with a diverse library of algorithms. To our knowledge, however, the magnitude of the improvements gained by including many algorithms has not yet been systematically evaluated in common epidemiologic research settings.
View Article and Find Full Text PDFFedNolowe: A normalized loss-based weighted aggregation strategy for robust federated learning in heterogeneous environments.
PLoS One
August 2025
Saigon University (SGU), Ho Chi Minh City, Vietnam.
Federated Learning supports collaborative model training across distributed clients while keeping sensitive data decentralized. Still, non-independent and identically distributed data pose challenges like unstable convergence and client drift. We propose Federated Normalized Loss-based Weighted Aggregation (FedNolowe) (Code is available at https://github.
View Article and Find Full Text PDFAnterior insular cortex glutamate-glutamine (Glx) levels predict general psychopathology via heightened error sensitivity.
Front Neurosci
July 2025
Graduate School of Medical Science and Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea.
Introduction: The anterior insular cortex (AIC) integrates interoceptive, cognitive-emotional, and error-monitoring signals, and is consistently hyperactive in anxiety and depression. Converging evidence links elevated glutamate + glutamine (Glx) in fronto-insular regions to stress reactivity; however, it is unknown whether AIC Glx relates to a transdiagnostic general psychopathology factor (G-score) or to the tendency to overweight prediction errors during learning. We therefore combined functional MRS (fMRS) with reinforcement-learning modeling to test whether (i) baseline AIC Glx predicts the G-score derived from bifactor analysis of PHQ-9, GAD-7, and STAI-X1, and (ii) task-evoked Glx changes track individual differences in error sensitivity during gain- and loss-based learning.
View Article and Find Full Text PDFEvaluation of the predictors of tooth loss using artificial intelligence-based machine learning approach: A retrospective study.
J Indian Soc Periodontol
June 2025
Department of Periodontics, Government Dental College and Hospital, Chennai, Tamil Nadu, India.
Background: A common and persistent inflammatory condition impacting the supportive structures of teeth, periodontal disease presents notable challenges in dental healthcare. It leads to various clinical issues, including the loss of clinical attachment, increased pocket depth, and tooth mobility. The global prevalence of periodontitis is substantial, with an estimated 20%-50% of the world's population affected, particularly in developing countries.
View Article and Find Full Text PDFUnderstanding how the extinction of mutualistic partners affects species and their functional role in the community is a foundational question in ecology. Such changes impact the architecture of interaction networks, but it remains unclear how they affect interaction outcomes, such as the nutrient resource provisioning for animals. Here, we fitted a community-level frugivory model with a comprehensive dataset of plant-bird interactions in six communities from the Yungas Forest in Argentina.
View Article and Find Full Text PDF