{"633860":{"#nid":"633860","#data":{"type":"event","title":"CSE Faculty Candidate Talk - Amir Gholaminejad","body":[{"value":"\u003Cp\u003ETo join the meeting on a computer or mobile phone:\u0026nbsp;\u003Ca href=\u0022https:\/\/bluejeans.com\/505935226\u0022 title=\u0022https:\/\/bluejeans.com\/505935226\u0022\u003Ehttps:\/\/bluejeans.com\/505935226\u003C\/a\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003EPhone Dial-in\u003C\/p\u003E\r\n\r\n\u003Cp\u003E+1.888.748.9073 (United States(Primary))\u003C\/p\u003E\r\n\r\n\u003Cp\u003E+1.844.540.8065 (United States(Primary))\u003C\/p\u003E\r\n\r\n\u003Cp\u003E+1.408.419.1715 (United States(San Jose))\u003C\/p\u003E\r\n\r\n\u003Cp\u003E+1.408.915.6290 (United States(San Jose))\u003C\/p\u003E\r\n\r\n\u003Cp\u003E(Global Numbers)\u003C\/p\u003E\r\n\r\n\u003Cp\u003EMeeting ID: 505 935 226\u003C\/p\u003E\r\n\r\n\u003Cp\u003ERoom System\u003C\/p\u003E\r\n\r\n\u003Cp\u003E199.48.152.152 or bjn.vc\u003C\/p\u003E\r\n\r\n\u003Cp\u003EMeeting ID: 505 935 226\u003C\/p\u003E\r\n\r\n\u003Cp\u003EWant to test your video connection?\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Ca href=\u0022https:\/\/bluejeans.com\/111\u0022\u003Ehttps:\/\/bluejeans.com\/111\u003C\/a\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003ETalk Title: \u0026nbsp;\u003C\/strong\u003E\u003Cem\u003EAn Integrated Approach for Efficient Neural Network Design, Training, and Inference\u003C\/em\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003ETalk Abstract:\u0026nbsp;\u003C\/strong\u003EOne of the main challenges in designing, training, and implementing Neural\u0026nbsp;Networks is their high demand for computational and memory resources. Designing a model for a new task requires searching through an exponentially large space to find the right architecture, which requires multiple training runs on a large dataset.\u0026nbsp; This has a prohibitive computational cost, as training each candidate architecture often requires millions of iterations.\u003Cbr \/\u003E\r\nEven after the right architecture with good accuracy is found, implementing\u0026nbsp;it on a target hardware platform to meet latency and\u0026nbsp;power constraints is not straightforward.\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nI will present a framework that efficiently utilizes reduced-precision computing to address the above challenges by considering the full stack of designing, training, and implementing the model on a target platform.\u0026nbsp; This is achieved through careful analysis of the numerical instabilities associated with reduced-precision matrix operations, incorporation of a novel second-order, mixed-precision quantization approach, and a framework for hardware aware neural network design.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003EBio:\u0026nbsp;\u003C\/strong\u003EAmir Gholami is a postdoctoral research fellow in BAIR Lab at UC\u0026nbsp;Berkeley. \u0026nbsp;He received his PhD in Computational Science and Engineering Mathematics from UT Austin, working with Prof. George Biros on bio-physics based image analysis, a research topic which received UT Austin\u0026rsquo;s best doctoral dissertation award in 2018. Amir has extensive experience in High Performance Computing, second-order optimization methods, image registration, and large scale inverse problems, developing codes that have been scaled up to 200K cores. He is a Melosh Medal finalist, recipient of best student paper award in SC\u0026#39;17, Gold Medal in the ACM Student Research Competition in 2015, as well as best student paper finalist in SC\u0026rsquo;14.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"CSE is hosting a faculty candidate seminar by Amir Gholaminejad,  a postdoctoral research fellow in BAIR Lab at UC Berkeley."}],"uid":"34540","created_gmt":"2020-03-27 17:36:41","changed_gmt":"2020-03-27 17:36:41","author":"Kristen Perez","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2020-04-09T12:00:00-04:00","event_time_end":"2020-04-09T13:00:00-04:00","event_time_end_last":"2020-04-09T13:00:00-04:00","gmt_time_start":"2020-04-09 16:00:00","gmt_time_end":"2020-04-09 17:00:00","gmt_time_end_last":"2020-04-09 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"624060","name":"Center for High Performance Computing (CHiPC)"},{"id":"47223","name":"College of Computing"},{"id":"50877","name":"School of Computational Science and Engineering"}],"categories":[],"keywords":[{"id":"166896","name":"seminar"},{"id":"4305","name":"cse"},{"id":"184339","name":"faculty candidat4e"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[],"invited_audience":[{"id":"78761","name":"Faculty\/Staff"},{"id":"78771","name":"Public"},{"id":"78751","name":"Undergraduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}