Results 61 to 70 of about 3,000 (250)

Measuring Catastrophic Forgetting in Neural Networks

Proceedings of the AAAI Conference on Artificial Intelligence, 2018
Deep neural networks are used in many state-of-the-art systems for machine perception. Once a network is trained to do a specific task, e.g., bird classification, it cannot easily be trained to do new tasks, e.g., incrementally learning to recognize additional bird species or learning an entirely different task such as flower ...
Ronald Kemker, Marc McClure, Angelina Abitino, Tyler L. Hayes, Christopher Kanan   +4 more
openaire   +2 more sources

Advances and Perspectives in Graphene‐Based Quantum Dots Enabled Neuromorphic Devices

Advanced Science, EarlyView.
Graphene‐based QDs are zero‐dimensional carbon nanomaterials with pronounced quantum confinement and tunable electronic structures. Herein, we summarize their synthesis strategies and functionalization methods, and highlight their functional roles and operating mechanisms in devices, as well as recent advances in neuromorphic electronics. We anticipate
Yulin Zhen, Wei Zeng, Zherui Zhao, JiYu Zhao, Ziyan Huang, Zhuo Chen, Kui Zhou, Guanglong Ding, Su‐Ting Han, Ye Zhou   +9 more
wiley   +1 more source

Reconfigurable Selector‐Only Memory (SOM) for Scalable Neuromorphic Computing

Advanced Science, EarlyView.
ABSTRACT Highly scalable reconfigurable neuromorphic devices are critical for addressing continual‐learning challenges in artificial intelligence. However, the scalability of existing reconfigurable devices is severely constrained by limited operating margins and insufficient process maturity.
Jin‐Yu Wen, Chuan‐Qi Yi, Ya‐Ru Zhang, Bin‐Hao Wang, Zi‐Xuan Liu, Chun‐Yu Zhou, Hao Tong, Xiang‐Shui Miao   +7 more
wiley   +1 more source

Continual learning and catastrophic forgetting

Preprint of a book chapter; 21 pages, 4 ...
Gido M. van de Ven, Nicholas Soures, Dhireesha Kudithipudi   +2 more
openaire   +3 more sources

Deep Learning Network‐Tailored Microenvironment Matching of 4D Bioprinting Bioactive Scaffolds for Bone Regeneration

Advanced Science, EarlyView.
A DLN dataset was built to analyze MABS composition versus in vitro/in vivo osteogenesis and angiogenesis. An MLP neural network, taking BG morphological parameters as input, extracts bioactive features from these datasets. A rabbit tibial defect model then validates 4D‐printed MABS for adaptability and bone regeneration in critical defects.
Xiongjie Liang, Yuechi Zhang, Weifeng Hu, Shiyan Lv, Fan Jia, Yan Zhang, Feng Wu, Changjiang Yang, Guanqi Zhen, Jinglong Yan, Xu Cui, Wei Zhao, Guanghua Chen   +12 more
wiley   +1 more source

A Closed‐Loop Framework for Inverse Design: Dynamic Training and Intelligent Optimization for Heterostructured Materials

Advanced Science, EarlyView.
To accelerate the inverse design of heterostructured metal matrix composites, a closed‐loop scientific machine learning framework integrates continual learning prediction with NSGA‐II‐PMCP optimization. The framework maps microstructural descriptors to strength, toughness, and modulus, expands high‐quality Pareto solutions, and guides experimentally ...
Zhiyan Zhong, Xueru Zheng, Xiao Zhou, Zhengheng Tao, Wan Han, Jingyi Pan, Dian Wu, Ning Gao, Lei Liu, Zhongyang Wang, Fanchao Meng, Tongxiang Fan   +11 more
wiley   +1 more source

Mitigating Catastrophic Forgetting in Pest Detection Through Adaptive Response Distillation

Agriculture
Pest detection in agriculture faces the challenge of adapting to new pest species while preserving the ability to recognize previously learned ones. Traditional model fine-tuning approaches often result in catastrophic forgetting, where the acquisition ...
Hongjun Zhang, Zhendong Yin, Dasen Li, Yanlong Zhao   +3 more
doaj   +1 more source

An Appraisal of Incremental Learning Methods

Entropy, 2020
As a special case of machine learning, incremental learning can acquire useful knowledge from incoming data continuously while it does not need to access the original data.
Yong Luo, Liancheng Yin, Wenchao Bai, Keming Mao   +3 more
doaj   +1 more source

Machine Learning Interatomic Potentials for Energy Materials: Architectures, Training Strategies, and Applications

Advanced Energy Materials, EarlyView.
Machine learning interatomic potentials bridge quantum accuracy and computational efficiency for materials discovery. Architectures from Gaussian process regression to equivariant graph neural networks, training strategies including active learning and foundation models, and applications in solid‐state electrolytes, batteries, electrocatalysts ...
In Kee Park, Saeed Pourasad, Jinhong Mun, Arunkumar Chitteth Rajan, Tae Hyeon Park, Anand Rohit, Mohammad Zafari, Tan‐Lien Pham, Eunjae Jung, Bonseung Koo, Miran Ha, Gi Beom Sim, Hyun Gyu Park, Won Woong Choi, Kwanghyeon Jo, Soohaeng Yoo Willow, David ChangMo Yang, Chang Woo Myung, Geunsik Lee, Kwang S. Kim   +19 more
wiley   +1 more source

Zero-shot incremental learning using spatial-frequency feature representations

Scientific Reports
Zero-shot incremental learning aims to enable a model to generalize to new classes without forgetting previously learned classes. However, the semantic gap between old and new sample classes can lead to catastrophic forgetting.
Jie Ren, Yang Zhao, Weichuan Zhang, Changming Sun   +3 more
doaj   +1 more source