BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20261027T120551Z UID:FF5DE119-CAD8-4FD5-A573-E71936830E31 DTSTART;TZID=America/New_York:20261024T100000 DTEND;TZID=America/New_York:20261024T110000 DESCRIPTION:Abstract:\n\nAs deep learning models scale into multi-billion p arameter spaces\, modern enterprise architectures face an immediate barrie r: the memory bandwidth wall. Standard operator execution patterns bottlen eck high-performance systems by forcing intermediate tensor states to repe atedly cycle through high-bandwidth memory (HBM). This lecture addresses t he mathematical and hardware codesign patterns required to achieve subline ar search mandates and drive maximum runtime execution throughput across r eal-time\, multi-modal environments. Through a deep dive into PyTorch impl ementation mechanics\, we examine the dual paradigms of data vs. compute v ectorization. We analyze how transforming unstructured real-world entities into continuous\, high-dimensional dense numeric representations provides critical topological context for downstream similarity checks. Moving pas t standard eager-mode execution patterns\, we demonstrate how custom OpenA I Triton and CUDA fused memory kernels mathematically combine scale\, mask \, and softmax routines into localized execution blocks inside fast SRAM r egisters\, bypassing HBM (High Bandwidth Memory) latency loops completely. Furthermore\, this session explores advanced index topologies specificall y Hierarchical Navigable Small Worlds (HNSW) coupled with Vector Quantizat ion (VQ) boundaries. By applying Vector Quantization to compress vast floa ting-point vector spaces into low-bit indices\, we illustrate how to bridg e the gap between continuous neural state spaces and discrete tool selecti on loops. Finally\, we synthesize these optimization mechanics within the framework of Agentic AI control loops (ReAct- Reasoning and Acting)\, prov iding a unified approach to cognitive scaling\, multi-agent synchronizatio n\, and real-world hallucination mitigation\n\nhttps://events.vtools.ieee. org/m/575616\n\nSpeaker(s): Dr. Baris Kazar\, \n\nRochester\, New York\, U nited States\, Virtual: https://events.vtools.ieee.org/m/580546 LOCATION:Rochester\, New York\, United States\, Virtual: https://events.vto ols.ieee.org/m/580546 ORGANIZER:mbkazar@gmail.com SEQUENCE:22 SUMMARY:{AI Talks with Tea&Coffee #53 @ RIT} : Advanced Latent Representati ons and Computational Optimization: A Summary of PyTorch Experiences in De ep Architectures\, RAG\, and Agentic AI: Vector Embeddings\, Vectorization and Vector Quantization URL;VALUE=URI:https://events.vtools.ieee.org/m/580546 X-ALT-DESC:Description: <br /><p data-removefontsize="true" data-originalco mputedfontsize="16">Abstract:</p>\n<p data-removefontsize="true" data-orig inalcomputedfontsize="16">As deep learning models scale into multi-billion parameter spaces\, modern enterprise architectures face an immediate barr ier: the memory bandwidth wall. Standard operator execution patterns bottl eneck high-performance systems by forcing intermediate tensor states to re peatedly cycle through high-bandwidth memory (HBM). This lecture addresses the mathematical and hardware codesign patterns required to achieve subli near search mandates and drive maximum runtime execution throughput across real-time\, multi-modal environments.&nbsp\;Through a deep dive into PyTo rch implementation mechanics\, we examine the dual paradigms of data vs. c ompute vectorization. We analyze how transforming unstructured real-world entities into continuous\, high-dimensional dense numeric representations provides critical topological context for downstream similarity checks. Mo ving past standard eager-mode execution patterns\, we demonstrate how cust om OpenAI Triton and CUDA fused memory kernels mathematically combine scal e\, mask\, and softmax routines into localized execution blocks inside fas t SRAM registers\, bypassing HBM (High Bandwidth Memory) latency loops com pletely.&nbsp\;Furthermore\, this session explores advanced index topologi es specifically Hierarchical Navigable Small Worlds (HNSW) coupled with Ve ctor Quantization (VQ) boundaries. By applying Vector Quantization to comp ress vast &nbsp\;floating-point vector spaces into low-bit indices\, we il lustrate how to bridge the gap between continuous neural state spaces and discrete tool selection loops. Finally\, we synthesize these optimization mechanics within the framework of Agentic AI control loops (ReAct- Reasoni ng and Acting)\, providing a unified approach to cognitive scaling\, multi -agent synchronization\, and real-world hallucination mitigation</p>\n<p d ata-removefontsize="true" data-originalcomputedfontsize="16">&nbsp\;</p>\n <p data-removefontsize="true" data-originalcomputedfontsize="16">https://e vents.vtools.ieee.org/m/575616</p> END:VEVENT END:VCALENDAR