Výsledky vyhledávání - "Neural network compression"

Lifelong Learning of Neural Networks: Detecting Novelty and Adapting to New Domains without Forgetting

Autoři: Masana Castrillo, Marc

Thesis Advisors: Weijer, Joost van de, Bagdanov, Andrew D.

Zdroj: TDX (Tesis Doctorals en Xarxa)

Témata: Compressió de xarxes neuronals, Compresión de redes neuronales, Neural network compression, Detecció d'anomalies, Detección de anomalías, Out-of-distribution detection, Aprenentatge continu, Aprendizaje continuo, Continual learning, Tecnologies

Popis souboru: application/pdf

Přístupová URL adresa: http://hdl.handle.net/10803/671591

Dynamic Hierarchical Token Merging for Vision Transformers

Autoři: Haroun, Karim Allenet, Thibault Ben Chehida, Karim a další

Přispěvatelé: Haroun, Karim Allenet, Thibault Ben Chehida, Karim a další

Zdroj: Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications. :677-684

Témata: Vision Transformers, [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO.INFO-CV] Computer Science [cs]/Computer Vision and Pattern Recognition [cs.CV], Neural network compression, Dynamic neural networks, Token merging

Popis souboru: application/pdf

Přístupová URL adresa: https://hal.science/hal-04885469v1
https://doi.org/10.5220/0013284100003912
https://hal.science/hal-04885469v1/document

Leveraging Local Similarity for Token Merging in Vision Transformers

Autoři: Karim Haroun Jean Martinet Karim Ben Chehida a další

Přispěvatelé: Karim Haroun Jean Martinet Karim Ben Chehida a další

Zdroj: Communications in Computer and Information Science ISBN: 9789819666874

Témata: Vision Transformers, [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], Adaptive inference, [INFO.INFO-CV] Computer Science [cs]/Computer Vision and Pattern Recognition [cs.CV], [INFO.INFO-NE] Computer Science [cs]/Neural and Evolutionary Computing [cs.NE], Deep learning, Neural network compression

Popis souboru: application/pdf

Přístupová URL adresa: https://hal.science/hal-04781040v1/document
https://doi.org/10.1007/978-981-96-6688-1_20
https://hal.science/hal-04781040v1

Performance Analysis of Two 8-Bit Floating-Point-based Piecewise Uniform Quantizers for a Laplacian Data Source

Autoři: Jelena R. Nikolic Zoran H. Peric Aleksandra Z. Jovanovic a další

Zdroj: Elektronika ir Elektrotechnika, Vol 31, Iss 1, Pp 56-61 (2025)

Témata: accuracy, floating-point arithmetic, quantization, Electrical engineering. Electronics. Nuclear engineering, neural network compression, TK1-9971

Přístupová URL adresa: https://doaj.org/article/45647fbca5e44ee48e02fb6fff931988

Coupled Tensor Decomposition for Compact Network Representation

Autoři: Van Tien Pham Yassine Zniyed Thanh Phuong Nguyen a další

Přispěvatelé: Van Tien Pham Yassine Zniyed Thanh Phuong Nguyen a další

Zdroj: IEEE Transactions on Neural Networks and Learning Systems. :1-15

Témata: coupled tensor decomposition, canonical polyadic decomposition, low-rank approximations, neural network compression, [STAT.ML] Statistics [stat]/Machine Learning [stat.ML], [SPI.SIGNAL] Engineering Sciences [physics]/Signal and Image processing

Popis souboru: application/pdf

Přístupová URL adresa: https://hal.science/hal-05249052v1
https://hal.science/hal-05249052v2/document
https://hal.science/hal-05249052v2
https://doi.org/10.1109/tnnls.2025.3609797

NMS-KSD: Efficient Knowledge Distillation for Dense Object Detection via Non-Maximum Suppression and Feature Storage

Autoři: Suho Son Byung Cheol Song

Zdroj: IEEE Access, Vol 13, Pp 80723-80736 (2025)

Témata: knowledge distillation, deep neural network compression, Electrical engineering. Electronics. Nuclear engineering, non-maximum suppression, Dense object detection, TK1-9971

Přístupová URL adresa: https://doaj.org/article/2da3f3dbe5b540f596a6a25a7bef4693

Compressing Neural Networks on Limited Computing Resources

Autoři: Seunghyun Lee Dongjun Lee Minju Hyun a další

Zdroj: IEEE Access, Vol 13, Pp 80063-80075 (2025)

Témata: lightweight deep learning models, filter pruning, Electrical engineering. Electronics. Nuclear engineering, knowledge transfer, Deep neural network compression, efficient neural networks, TK1-9971

Přístupová URL adresa: https://doaj.org/article/de6b40cacc844e4c942d46dd7ddeb89f

DARIO: Differentiable Vision Transformer Pruning With Low-Cost Proxies

Autoři: Sun, Haozhe Heuillet, Alexandre Mohr, Felix a další

Přispěvatelé: Sun, Haozhe Heuillet, Alexandre Mohr, Felix a další

Zdroj: IEEE Journal of Selected Topics in Signal Processing. 18:997-1009

Témata: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], neural architecture search, Neural network pruning, deep learning, neural network pruning, Neural network compression, artificial neural networks, neural network compression, Neural architecture search

Popis souboru: application/pdf

Scheduling Machine Learning Compressible Inference Tasks with Limited Energy Budget

Autoři: da Silva Barros, Tiago Ferre, Davide Giroire, Frederic a další

Přispěvatelé: da Silva Barros, Tiago Ferre, Davide Giroire, Frederic a další

Zdroj: Proceedings of the 53rd International Conference on Parallel Processing

Témata: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO.INFO-NI] Computer Science [cs]/Networking and Internet Architecture [cs.NI], 8. Economic growth, energy budget, [INFO.INFO-DS] Computer Science [cs]/Data Structures and Algorithms [cs.DS], scheduling, deadlines, [INFO] Computer Science [cs], 7. Clean energy, neural network compression, [INFO.INFO-RO] Computer Science [cs]/Operations Research [math.OC]

Popis souboru: application/pdf

Přístupová URL adresa: https://hal.science/hal-04676376v1/document
https://hal.science/hal-04676376v1
https://doi.org/10.1145/3673038.3673106

Scheduling with Fully Compressible Tasks: Application to Deep Learning Inference with Neural Network Compression

Autoři: da Silva Barros, Tiago Giroire, Frédéric Aparicio-Pardo, Ramon a další

Přispěvatelé: da Silva Barros, Tiago Giroire, Frédéric Aparicio-Pardo, Ramon a další

Zdroj: 2024 IEEE 24th International Symposium on Cluster, Cloud and Internet Computing (CCGrid)

Témata: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO.INFO-NI] Computer Science [cs]/Networking and Internet Architecture [cs.NI], [INFO.INFO-DS] Computer Science [cs]/Data Structures and Algorithms [cs.DS], scheduling, [INFO] Computer Science [cs], compressible tasks, approximation algorithms, convex programming, neural network compression, [INFO.INFO-RO] Computer Science [cs]/Operations Research [math.OC]

Popis souboru: application/pdf

AdaQAT: Adaptive Bit-Width Quantization-Aware Training

Autoři: Gernigon, Cédric Filip, Silviu-Ioan Sentieys, Olivier a další

Přispěvatelé: Gernigon, Cédric Filip, Silviu-Ioan Sentieys, Olivier a další

Zdroj: 2024 IEEE 6th International Conference on AI Circuits and Systems (AICAS). :442-446

Témata: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], FOS: Computer and information sciences, Computer Science - Machine Learning, Artificial Intelligence (cs.AI), Computer Science - Artificial Intelligence, Adaptive Bit-Width Optimization, Neural Network Compression, [INFO.INFO-AO] Computer Science [cs]/Computer Arithmetic, Quantization Aware Training, Machine Learning (cs.LG)

Popis souboru: application/pdf

Přístupová URL adresa: http://arxiv.org/abs/2404.16876
https://hal.science/hal-04549245v1

A comprehensive review of model compression techniques in machine learning

Autoři: Pierre Vilar Dantas Waldir Sabino da Silva Lucas Carvalho Cordeiro a další

Zdroj: Dantas, P V, da Silva Junior, W S, Cordeiro, L C & Carvalho, C B 2024, 'A Comprehensive Review of Model Compression Techniques in Machine Learning', Applied Intelligence, vol. 54, no. 22, pp. 11804-11844. https://doi.org/10.1007/s10489-024-05747-w

Témata: computational efficiency, lightweight design approaches, architectural innovations, technological evolution in machine learning, model generalization, neural network compression

Přístupová URL adresa: https://research.manchester.ac.uk/en/publications/1c15cef0-c6f5-4d13-a6d5-c3c23c9ce389
https://doi.org/10.1007/s10489-024-05747-w

Smoothed per-tensor weight quantization: a robust solution for neural network deployment

Autoři: Xin Chang

Zdroj: International Journal of Electronics and Telecommunications, Vol vol. 71, Iss No 3 (2025)

Témata: per-tensor quantization, edge device, neural network compression, Electrical engineering. Electronics. Nuclear engineering, TK1-9971, Telecommunication, TK5101-6720

Popis souboru: electronic resource

Relation: https://journals.pan.pl/Content/135755/23_4966_Chang_L_sk.pdf; https://doaj.org/toc/2081-8491; https://doaj.org/toc/2300-1933

Přístupová URL adresa: https://doaj.org/article/bccde7525d194aab9d62899f43c1fa50

SGRACE: Scalable Architecture for On-Device Inference and Training of Graph Attention and Convolutional Networks

Autoři: Nunez-Yanez, Jose Luis Mousanejad Jeddi, Hadi

Zdroj: IEEE Transactions on Very Large Scale Integration (vlsi) Systems.

Témata: Hardware, Training, Field programmable gate arrays, Computer architecture, Quantization (signal), Computational modeling, Complexity theory, Sparse matrices, Instruction sets, Hardware acceleration, Accelerator architectures, graph neural networks, high-level synthesis (HLS), neural network compression, reconfigurable architectures

Popis souboru: print

Přístupová URL adresa: https://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-217251

Rapture of the deep: highs and lows of sparsity in a world of depths

Autoři: Gribonval, Rémi Riccietti, Elisa Le, Quoc-Tung a další

Přispěvatelé: Gribonval, Rémi Riccietti, Elisa Le, Quoc-Tung a další

Témata: Linear inverse problem, [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO.INFO-TS] Computer Science [cs]/Signal and Image Processing, Neural Network Compression, [MATH.MATH-OC] Mathematics [math]/Optimization and Control [math.OC], Sparsity, Butterfly factorization

Popis souboru: application/pdf

Přístupová URL adresa: https://inria.hal.science/hal-04954574v4

Small is Beautiful: Compressing Deep Neural Networks for Partial Domain Adaptation

Autoři: Yuzhe Ma Xufeng Yao Ran Chen a další

Zdroj: IEEE Transactions on Neural Networks and Learning Systems. 35:3575-3585

Témata: Task analysis, Taylor series, Adaptation models, Training, Computational modeling, Deep learning, Neural network compression, Supervised learning, Transfer learning

Přístupová URL adresa: https://pubmed.ncbi.nlm.nih.gov/35939474

Efficient N:M Sparse DNN Training Using Algorithm, Architecture, and Dataflow Co-Design

Autoři: Chao Fang Wei Sun Aojun Zhou a další

Zdroj: IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems. 43:506-519

Témata: FOS: Computer and information sciences, Computer Science - Machine Learning, Computer Science - Artificial Intelligence, deep neural networks (DNNs), pruning, Algorithm-hardware codesign, 7. Clean energy, neural network compression, sparse training, Machine Learning (cs.LG), DNN training, Artificial Intelligence (cs.AI), Hardware Architecture (cs.AR), SDG 7 - Affordable and Clean Energy, Computer Science - Hardware Architecture, SDG 7 – Betaalbare en schone energie

Přístupová URL adresa: http://arxiv.org/abs/2309.13015
https://research.tue.nl/en/publications/a0a0e678-c270-4c6f-b608-68d3f66ff83f
https://doi.org/10.1109/TCAD.2023.3317789

Channel Pruning of Transfer Learning Models Using Novel Techniques

Autoři: Pragnesh Thaker Biju R. Mohan

Zdroj: IEEE Access, Vol 12, Pp 94914-94925 (2024)

Témata: Deep compression of CNN, structured pruning, 0202 electrical engineering, electronic engineering, information engineering, channel pruning, Electrical engineering. Electronics. Nuclear engineering, 02 engineering and technology, transfer learning, neural network compression, TK1-9971

Přístupová URL adresa: https://doaj.org/article/e45c645f1caf4c91bc29eae38e469b3a

MDPruner: Meta-Learning Driven Dynamic Filter Pruning for Efficient Object Detection

Autoři: Lingyun Zhou Xiaoyong Liu

Zdroj: IEEE Access, Vol 12, Pp 136925-136935 (2024)

Témata: meta-learning, Object detection, convolutional neural network compression and acceleration, filter pruning, Electrical engineering. Electronics. Nuclear engineering, TK1-9971

Přístupová URL adresa: https://doaj.org/article/b6fbded8ea804a0693e9dad921788442

Enhancing Deep Compression of CNNs: A Novel Regularization Loss and the Impact of Distance Metrics

Autoři: Pragnesh Thaker Biju R. Mohan

Zdroj: IEEE Access, Vol 12, Pp 172537-172547 (2024)

Témata: Deep compression of CNN, structured pruning, channel pruning, Electrical engineering. Electronics. Nuclear engineering, transfer learning, neural network compression, TK1-9971

Přístupová URL adresa: https://doaj.org/article/068852194fc44fd884860f48a6a59a85