{"id":546513,"date":"2018-10-30T02:58:05","date_gmt":"2018-10-30T09:58:05","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-research-item&p=546513"},"modified":"2019-10-21T13:30:17","modified_gmt":"2019-10-21T20:30:17","slug":"compiling-kb-sized-machine-learning-models-to-constrained-hardware","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/compiling-kb-sized-machine-learning-models-to-constrained-hardware\/","title":{"rendered":"Compiling KB-Sized Machine Learning Models to Tiny IoT Devices"},"content":{"rendered":"

Recent advances in machine learning (ML) have produced KiloByte-size models that can directly run on constrained IoT devices. This approach avoids expensive communication between IoT devices and the cloud, thereby enabling energy-efficient real-time analytics. However, ML models are expressed typically in floating-point, and IoT hardware typically does not support floating-point. Therefore, running these models on IoT devices requires simulating IEEE-754 floating-point using software, which is very inefficient. We present SeeDot, a domain-specific language to express ML inference algorithms and a compiler that compiles SeeDot programs to fixed-point code that can efficiently run on constrained IoT devices. We propose 1) a novel compilation strategy that reduces the search space for some key parameters used in the fixed-point code, and 2) new efficient implementations of expensive operations. SeeDot compiles state-of-the-art KB-sized models to various microcontrollers and low-end FPGAs. We show that SeeDot outperforms 1) software emulation of floating-point (Arduino), 2) high bitwidth fixed-point (MATLAB), 3) post-training quantization (TensorFlow-Lite), and 4) floating- and fixed-point FPGA implementations generated using high-level synthesis tools.<\/p>\n","protected":false},"excerpt":{"rendered":"

Recent advances in machine learning (ML) have produced KiloByte-size models that can directly run on constrained IoT devices. This approach avoids expensive communication between IoT devices and the cloud, thereby enabling energy-efficient real-time analytics. However, ML models are expressed typically in floating-point, and IoT hardware typically does not support floating-point. Therefore, running these models on […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"msr-content-type":[3],"msr-research-highlight":[],"research-area":[13556,13552,13560],"msr-publication-type":[193716],"msr-product-type":[],"msr-focus-area":[],"msr-platform":[],"msr-download-source":[],"msr-locale":[268875],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-546513","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-hardware-devices","msr-research-area-programming-languages-software-engineering","msr-locale-en_us"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2019-6-24","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"ACM","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"file","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/10\/pldi19-SeeDot.pdf","id":"596629","title":"pldi19-seedot","label_id":"243109","label":0}],"msr_related_uploader":"","msr_attachments":[{"id":596629,"url":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2019\/07\/pldi19-SeeDot.pdf"},{"id":546516,"url":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/10\/paper.pdf"}],"msr-author-ordering":[{"type":"text","value":"Sridhar Gopinath","user_id":0,"rest_url":false},{"type":"text","value":"Nikhil Ghanathe","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Vivek Seshadri","user_id":36323,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Vivek Seshadri"},{"type":"user_nicename","value":"Rahul Sharma","user_id":36308,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Rahul Sharma"}],"msr_impact_theme":[],"msr_research_lab":[199562],"msr_event":[],"msr_group":[],"msr_project":[416408,546507],"publication":[],"video":[],"download":[621708],"msr_publication_type":"inproceedings","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/546513"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":6,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/546513\/revisions"}],"predecessor-version":[{"id":616599,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/546513\/revisions\/616599"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=546513"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=546513"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=546513"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=546513"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=546513"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=546513"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=546513"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=546513"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=546513"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=546513"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=546513"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=546513"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=546513"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=546513"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=546513"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}