A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Filters
Prioritized Subnet Sampling for Resource-Adaptive Supernet Training
[article]
2023
arXiv
pre-print
Preserved Fulltext
Other Versions
In this paper, we propose prioritized subnet sampling to train a resource-adaptive supernet, termed PSS-Net. ...
A resource-adaptive supernet adjusts its subnets for inference to fit the dynamically available resources. ...
CONCLUSION We have proposed Prioritized Subnet Sampling (PSS-Net) to train a resource-adaptive supernet. ...
arXiv:2109.05432v2
fatcat:wdfinvlhsnh3zht3zkbhhsdg2a
DNA Family: Boosting Weight-Sharing NAS with Block-Wise Supervisions
[article]
2024
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2024; you can also visit the original URL.
The file type is application/pdf.
In the supernet training stage, we decouple each block, significantly increasing subnet sampling efficiency compared to prior weight-sharing NAS methods. ...
Then, they extract weights from the supernet for each subnet for validation and use this validation accuracy to rate the subnet. ...
arXiv:2403.01326v1
fatcat:uiyteif5lvb47pcr54np55ubu4
UFO: Unified Feature Optimization
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
This paper proposes a novel Unified Feature Optimization (UFO) paradigm for training and deploying deep models under real-world and large-scale scenarios, which requires a collection of multiple AI functions ...
Instead, it aims to make the trimmed model dedicated for one or more already-seen task. ...
Dynamic Task Prioritization [14] automatically prioritizes more difficult tasks by adaptively adjusting the mixing weight of each task's loss objective. ...
arXiv:2207.10341v1
fatcat:hl2bzzbxabb5hapb7b6gvtmskq
Once for Both: Single Stage of Importance and Sparsity Search for Vision Transformer Compression
[article]
2024
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2024; you can also visit the original URL.
The file type is application/pdf.
Such a bi-mask search strategy is further used together with a proposed adaptive one-hot loss to realize the progressive-and-efficient search for the most important subnet. ...
In this work, for the first time, we investigate how to integrate the evaluations of importance and sparsity scores into a single stage, searching the optimal subnets in an efficient manner. ...
Existing Transformer Architecture Search (TAS) works [5, 32, 48] mainly follow the SPOS NAS [18] scheme, which first trains the supernet from scratch by iteratively training the sampled subnets, then ...
arXiv:2403.15835v1
fatcat:jiqiu4lauzhbhcr3ulkjufwhve
DREAM: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2022 pre-print arXiv:2212.03414v1 fatcat:j2ujf4bmebhzdorkwjn67jnagm
DREAM utilizes tunable parameters that provide fast and effective adaptivity to dynamic workload changes. ...
In addition, RTMM workloads require real-time processing, involve highly heterogeneous models, and target resource-constrained devices. ...
Supernet facilitates the training of multiple models with a single training process, which provides scalability for the model development process. ...
arXiv:2212.03414v2
fatcat:yeoaxazlpnf3jmp26hpstisgre
DREAM: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
2023
Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 4
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2024; you can also visit the original URL.
The file type is application/pdf.
DREAM utilizes tunable parameters that provide fast and effective adaptivity to dynamic workload changes. ...
In addition, RTMM workloads require real-time processing, involve highly heterogeneous models, and target resource-constrained devices. ...
We thank Harshit Khaitan and Don Stark for their support. ...
doi:10.1145/3623278.3624753
fatcat:rya74g4iszg67ogk5scgl36n6i
Generalized Global Ranking-Aware Neural Architecture Ranker for Efficient Image Classifier Search
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Meanwhile, the global quality distribution facilitates the search phase by directly sampling candidates according to the statistics of quality tiers, which is free of training a search algorithm, e.g., ...
Neural Architecture Search (NAS) is a powerful tool for automating effective image processing DNN designing. The ranking has been advocated to design an efficient performance predictor for NAS. ...
We acknowledge the anonymous reviewers for their valuable comments. We thank Miao Guo for the helpful discussions and beautiful illustrations. ...
arXiv:2201.12725v2
fatcat:wv3vtx3hkbdhbfudkbyxrgmqt4
DiffusionNAG: Predictor-guided Neural Architecture Generation with Diffusion Models
[article]
2024
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2024; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2305.16943v2 fatcat:ksuxcdxksvcrpapwnf6jniagz4Other Versions
Existing NAS methods suffer from either an excessive amount of time for repetitive sampling and training of many task-irrelevant architectures. ...
Moreover, with the guidance of parameterized predictors, DiffusionNAG can flexibly generate task-optimal architectures with the desired properties for diverse tasks, by sampling from a region that is more ...
MobileNetV3 In our training pipeline, we fine-tuned a subnet of a pretrained supernet from the MobileNetV3 search space on the ImageNet 1K dataset for CIFAR-10, CIFAR-100, Aircraft, and Oxford-IIIT Pets ...
arXiv:2305.16943v4
fatcat:7pfvcozb7fgariwh7cu7revzne
Intrusion-Detection Systems
[chapter]
2012
Handbook of Computer Networks
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2014; you can also visit the original URL.
The file type is application/pdf.
and software assurance. of maturity to warrant a comprehensive textbook treatment. ideas for books under this series. ...
future research in information security and, two, to serve as a central reference source for advanced and timely topics in information security research and development. ...
On the other hand, if the training samples are not general enough, they might lead the system to a bias, making it too specialized for the characteristics of the training samples alone, thus almost canceling ...
doi:10.1002/9781118256107.ch26
fatcat:aeidzkegvfc27dqqmztiayv3dm