{"id":{"repo_id":"nus","oai_identifier":"oai:scholarbank.nus.edu.sg:10635/309529"},"canonical_url":"https://search.dev.ndltd.org/etd/nus/oai:scholarbank.nus.edu.sg:10635/309529","repository":{"repo_id":"nus","name":"National University of Singapore","base_url":"https://scholarbank.nus.edu.sg/oai/request"},"display":{"title":"NEURAL ARCHITECTURE DESIGN AND APPLICATIONS","abstract":"Neural architecture design is crucial in AI development. This thesis first examines the macroscopic architecture of Transformers, challenging the belief that their attention-based token mixer is key. By replacing the attention module with a simple spatial pooling operator, we create PoolFormer, which achieves competitive performance across vision tasks. This supports our hypothesis that the general Transformer architecture, dubbed MetaFormer, is essential for model performance. MetaFormer ensures consistent performance and works well with various token mixers, often yielding state-of-the-art results. Additionally, we address efficiency in neural architecture by proposing InceptionNeXt, which decomposes large-kernel depthwise convolutions into smaller components, inspired by Inceptions. InceptionNeXt improves throughput and maintains performance, offering a more efficient baseline for future architecture design.","abstract_html":"Neural architecture design is crucial in AI development. This thesis first examines the macroscopic architecture of Transformers, challenging the belief that their attention-based token mixer is key. By replacing the attention module with a simple spatial pooling operator, we create PoolFormer, which achieves competitive performance across vision tasks. This supports our hypothesis that the general Transformer architecture, dubbed MetaFormer, is essential for model performance. MetaFormer ensures consistent performance and works well with various token mixers, often yielding state-of-the-art results. Additionally, we address efficiency in neural architecture by proposing InceptionNeXt, which decomposes large-kernel depthwise convolutions into smaller components, inspired by Inceptions. InceptionNeXt improves throughput and maintains performance, offering a more efficient baseline for future architecture design.","abstract_has_math":false,"creators":["YU WEIHAO"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-03-15","date_published":"2024-03-15","updated_at":"2026-07-24T03:32:30Z","subjects":["CNN","Transformer","Neural Networks","Neural Architecture"],"languages":[],"rights":[],"rights_urls":["https://scholarbank.nus.edu.sg/bitstreams/c1143ab1-76aa-4e04-822e-bed9a2ba05ee/download"],"identifier_entries":[]},"links":{"outbound_url":null,"outbound_label":null,"outbound_source":null},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["YU WEIHAO"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2024-03-15"]},{"key":"dc:relation.isreferencedby","label":"Dc Relation Isreferencedby","values":["https://scholarbank.nus.edu.sg/handle/10635/309529"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["CNN","Transformer","Neural Networks","Neural Architecture"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["https://scholarbank.nus.edu.sg/bitstreams/c1143ab1-76aa-4e04-822e-bed9a2ba05ee/download"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://scholarbank.nus.edu.sg/bitstreams/b2b1b3c0-1c60-497b-a9c2-1076c7d1a58b/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Neural architecture design is crucial in AI development. This thesis first examines the macroscopic architecture of Transformers, challenging the belief that their attention-based token mixer is key. By replacing the attention module with a simple spatial pooling operator, we create PoolFormer, which achieves competitive performance across vision tasks. This supports our hypothesis that the general Transformer architecture, dubbed MetaFormer, is essential for model performance. MetaFormer ensures consistent performance and works well with various token mixers, often yielding state-of-the-art results. Additionally, we address efficiency in neural architecture by proposing InceptionNeXt, which decomposes large-kernel depthwise convolutions into smaller components, inspired by Inceptions. InceptionNeXt improves throughput and maintains performance, offering a more efficient baseline for future architecture design."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["9f3c6f2aa8f96291ef77b0a18484521d","94f70dc3fe417073adf1d74ae9343ffe","49acae2c8013bd3a530a947c4c2e51dd"]},{"key":"dc:title","label":"Title","values":["NEURAL ARCHITECTURE DESIGN AND APPLICATIONS"]}]}],"canonical_facts":{"dc:creator":["YU WEIHAO"],"dc:date.issued":["2024-03-15"],"dc:description.abstract":["Neural architecture design is crucial in AI development. This thesis first examines the macroscopic architecture of Transformers, challenging the belief that their attention-based token mixer is key. By replacing the attention module with a simple spatial pooling operator, we create PoolFormer, which achieves competitive performance across vision tasks. This supports our hypothesis that the general Transformer architecture, dubbed MetaFormer, is essential for model performance. MetaFormer ensures consistent performance and works well with various token mixers, often yielding state-of-the-art results. Additionally, we address efficiency in neural architecture by proposing InceptionNeXt, which decomposes large-kernel depthwise convolutions into smaller components, inspired by Inceptions. InceptionNeXt improves throughput and maintains performance, offering a more efficient baseline for future architecture design."],"dc:format.checksum.md5":["9f3c6f2aa8f96291ef77b0a18484521d","94f70dc3fe417073adf1d74ae9343ffe","49acae2c8013bd3a530a947c4c2e51dd"],"dc:identifier.uri":["https://scholarbank.nus.edu.sg/bitstreams/b2b1b3c0-1c60-497b-a9c2-1076c7d1a58b/download"],"dc:relation.isreferencedby":["https://scholarbank.nus.edu.sg/handle/10635/309529"],"dc:rights":["https://scholarbank.nus.edu.sg/bitstreams/c1143ab1-76aa-4e04-822e-bed9a2ba05ee/download"],"dc:subject":["CNN","Transformer","Neural Networks","Neural Architecture"],"dc:title":["NEURAL ARCHITECTURE DESIGN AND APPLICATIONS"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T03:32:30Z"}