Results 81 to 90 of about 33,100 (246)
Online Structured Laplace Approximations For Overcoming Catastrophic Forgetting [PDF]
, 2018 We introduce the Kronecker factored online Laplace approximation for overcoming catastrophic forgetting in neural networks. The method is grounded in a Bayesian online learning framework, where we recursively approximate the posterior after every task ...
Barber, David +2 more
core +1 more source
Lifelong Neural Predictive Coding: Learning Cumulatively Online without Forgetting
, 2021 In lifelong learning systems, especially those based on artificial neural networks, one of the biggest obstacles is the severe inability to retain old knowledge as new information is encountered. This phenomenon is known as catastrophic forgetting.
Giles, C. Lee +3 more
core
Benchmarking Large Language Models for Polymer Property Predictions
Macromolecular Rapid Communications, EarlyView.Large language models (LLMs) are fine‐tuned on polymer thermal property datasets to directly predict glass transition, melting, and decomposition temperatures from SMILES inputs. Compared to state‐of‐the‐art models such as Polymer Genome, polyGNN, and polyBERT, LLMs achieve competitive yet lower accuracy.
Sonakshi Gupta +3 more
wiley +1 more source
Nature Communications, 2022 Many existing deep learning algorithms for particle picking are not predictable on unseen datasets. Here the authors report an exemplar-based continual learning approach, EPicker, enabling accumulation of new knowledge of cryoEM particle picking without ...
Xinyu Zhang +4 more
doaj +1 more source
Deep Reinforcement Learning‐Based Control for Real‐Time Hybrid Simulation of Civil Structures
International Journal of Robust and Nonlinear Control, EarlyView.ABSTRACT Real‐time Hybrid Simulation (RTHS) is a cyber‐physical technique that studies the dynamic behavior of a system by combining physical and numerical components that are coupled through a boundary condition enforcer. In structural engineering, the numerical components are subjected to environmental loads that become dynamic displacements of the ...
Andrés Felipe Niño +6 more
wiley +1 more source
Sliding Doors: Frame Uptake and Rejection by Learners in a Museum‐Based Climate Learning Experience
Science Education, EarlyView.ABSTRACT Science education efforts that support public understanding of modern climate change are critically needed. However, implementing climate‐related learning experiences can be challenging, as public audiences tend to experience a wide range of understandings of and emotions around the issue. In light of these challenges, many scholars have posed
Lynne Zummo +7 more
wiley +1 more source
Synaptic metaplasticity in binarized neural networks
Nature Communications, 2021 Deep neural networks usually rapidly forget the previously learned tasks while training new ones. Laborieux et al. propose a method for training binarized neural networks inspired by neuronal metaplasticity that allows to avoid catastrophic forgetting ...
Axel Laborieux +3 more
doaj +1 more source
Symbolic Interaction, EarlyView.This study of first‐year primary school draws on Goffman's concept of “collective behavior” to examine how order is established and disrupted through the mutual adjustment of all participants' actions. We employed a multi‐method longitudinal design, using semi‐standardized observations and qualitative interviews with teachers and children at three ...
Doris Bühler‐Niederberger +2 more
wiley +1 more source
TESOL Quarterly, EarlyView.Abstract This article reports on a qualitative study of the way instructors and students understand and respond to traumatizing events in a Sri Lankan university. It shows how the attitudes and practices in the society at large are carried over to classrooms even though local institutions do not have a programmatic trauma‐informed pedagogy.
Suresh Canagarajah +1 more
wiley +1 more source
Is the COBRA‐OS 4 French Aortic Occlusion Device Feasible for Partial REBOA?
World Journal of Surgery, EarlyView.Noncompressible torso hemorrhage is a leading cause of preventable death. Resuscitative endovascular balloon occlusion of the aorta (REBOA) is used to control hemorrhage but full aortic occlusion is limited by distal ischemia. Partial REBOA (pREBOA) allows some distal perfusion to prolong occlusion, but precise balloon titration is difficult with ...
Adam Power +4 more
wiley +1 more source