A Compact Neural Architecture Search for Accelerating Image Classification Models

Cited 1 time in webofscience Cited 0 time in scopus
  • Hit : 154
  • Download : 0
Nowadays, Automated Machine Learning (AutoML)has gradually become an inevitable trend providing automaticand suitable solutions to address AI tasks without needingmore efforts from experts. Neural Architecture Search (NAS), asubfield of AutoML, has generated automated models solving fun-damental problems in computer vision such as image recognition,objects detection. NAS with differentiable search strategies hasreduced significantly the GPU time that occupancy on calculation.In this paper, we present an effective algorithm that allowsexpanding search spaces by selecting operation candidates fromthe initial set with different ways in concurrent execution. Theextended search space makes NAS having more opportunities tofind good architectures simultaneously by running the group ofsearch spaces in overlapping time periods instead of sequentially.Our approach, is called Accelerated NAS, shortens 1.8× search-ing time when comparing to previous works. In addition, theAccelerated NAS generates potential neural architectures havingcomparable performances with the low inference time.
Publisher
The korean institute of communications and information sciences (KICS)
Issue Date
2021-10-22
Language
English
Citation

12th International Conference on ICT Convergence (ICTC) - Beyond the Pandemic Era with ICT Convergence Innovation, pp.1713 - 1718

ISSN
2162-1233
DOI
10.1109/ICTC52510.2021.9620797
URI
http://hdl.handle.net/10203/289094
Appears in Collection
EE-Conference Papers(학술회의논문)
Files in This Item
There are no files associated with this item.
This item is cited by other documents in WoS
⊙ Detail Information in WoSⓡ Click to see webofscience_button
⊙ Cited 1 items in WoS Click to see citing articles in records_button

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0