BEGIN:VCALENDAR
VERSION:2.0
X-WR-CALNAME:EventsCalendar
PRODID:-//hacksw/handcal//NONSGML v1.0//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:America/New_York
LAST-MODIFIED:20240422T053451Z
TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:College of Engineering,Thesis/Dissertations
DESCRIPTION:Topic: DSM-RC: A Hardware-Efficient Digital Reservoir Computing
  Architecture With Delta–Sigma-Inspired Dynamics and Count-Domain Readou
 t Abstract: Reservoir computing (RC) is a recurrent neural computing appro
 ach that offers opportunities for efficient hardware implementation by res
 tricting training primarily to the output layer. However, conventional res
 ervoir architectures can require substantial computational resources, memo
 ry, and arithmetic operations, limiting their suitability for resource-con
 strained digital systems. This research investigates the design, implement
 ation, and evaluation of a hardware-efficient digital reservoir computing 
 architecture for the classification of finite input streams. The architect
 ure emphasizes reduced arithmetic complexity and memory requirements throu
 gh binary-state processing, count-based feature extraction, and multiplier
 -free classification. The proposed architecture employs a sparse ternary t
 apped projection to drive a parallel bank of heterogeneous reservoir nodes
  inspired by first-order delta–sigma modulation. Each node incorporates 
 state accumulation, one-bit quantization, and output-dependent bipolar fee
 dback to produce nonlinear temporal responses. To reduce memory requiremen
 ts, binary reservoir outputs are summarized through online output-density 
 and transition-count features rather than storing complete temporal trajec
 tories. A multiplier-free classification stage based on sign-separated bit
 -plane arithmetic is developed to perform inference using shifts, addition
 s, subtractions, and comparisons. The MNIST dataset is used as a represent
 ative classification benchmark to investigate the classification accuracy,
  computational complexity, hardware resource utilization, throughput, and 
 power consumption. Advisor(s): Dr. Md Sakib Hasan, Associate Professor, De
 pt. of Electrical & Computer Engineering, UMASS Dartmouth Committee Member
 s: Dr. Sezer Goren, Associate Teaching Professor, Dept. of Electrical & Co
 mputer Engineering, UMASS Dartmouth; Dr. Liudong Xing, Commonwealth Profes
 sor, Dept. of Electrical & Computer Engineering, UMASS Dartmouth NOTE: All
  ECE Graduate Students are ENCOURAGED to attend. All interested parties ar
 e invited to attend. Open to the public. *For further information, please 
 contact Dr. Md Sakib Hasan via email at mhasan19@umassd.edu.\nEvent page: 
 https://www.umassd.edu/events/cms/12-11-26-elec-research-component-of-phd-
 qualifier-exam-by-mehedi-hasan-.php
X-ALT-DESC;FMTTYPE=text/html:<html><body><p>Topic: DSM-RC: A Hardware-Effic
 ient Digital Reservoir Computing Architecture With Delta–Sigma-Inspired 
 Dynamics and Count-Domain Readout</p>\n<p>Abstract: Reservoir computing (R
 C) is a recurrent neural computing approach that offers opportunities for 
 efficient hardware implementation by restricting training primarily to the
  output layer. However\, conventional reservoir architectures can require 
 substantial computational resources\, memory\, and arithmetic operations\,
  limiting their suitability for resource-constrained digital systems. This
  research investigates the design\, implementation\, and evaluation of a h
 ardware-efficient digital reservoir computing architecture for the classif
 ication of finite input streams. The architecture emphasizes reduced arith
 metic complexity and memory requirements through binary-state processing\,
  count-based feature extraction\, and multiplier-free classification. The 
 proposed architecture employs a sparse ternary tapped projection to drive 
 a parallel bank of heterogeneous reservoir nodes inspired by first-order d
 elta–sigma modulation. Each node incorporates state accumulation\, one-b
 it quantization\, and output-dependent bipolar feedback to produce nonline
 ar temporal responses. To reduce memory requirements\, binary reservoir ou
 tputs are summarized through online output-density and transition-count fe
 atures rather than storing complete temporal trajectories. A multiplier-fr
 ee classification stage based on sign-separated bit-plane arithmetic is de
 veloped to perform inference using shifts\, additions\, subtractions\, and
  comparisons. The MNIST dataset is used as a representative classification
  benchmark to investigate the classification accuracy\, computational comp
 lexity\, hardware resource utilization\, throughput\, and power consumptio
 n.</p>\n<p>Advisor(s): Dr. Md Sakib Hasan\, Associate Professor\, Dept. of
  Electrical & Computer Engineering\, UMASS Dartmouth</p>\n<p>Committee Mem
 bers: Dr. Sezer Goren\, Associate Teaching Professor\, Dept. of Electrical
  & Computer Engineering\, UMASS Dartmouth\; Dr. Liudong Xing\, Commonwealt
 h Professor\, Dept. of Electrical & Computer Engineering\, UMASS Dartmouth
 </p>\n<p>NOTE: All ECE Graduate Students are ENCOURAGED to attend. All int
 erested parties are invited to attend. Open to the public.</p>\n<p>*For fu
 rther information\, please contact Dr. Md Sakib Hasan via email at mhasan1
 9@umassd.edu.</p><p>Event page: <a href="https://www.umassd.edu/events/cms
 /12-11-26-elec-research-component-of-phd-qualifier-exam-by-mehedi-hasan-.p
 hp">https://www.umassd.edu/events/cms/12-11-26-elec-research-component-of-
 phd-qualifier-exam-by-mehedi-hasan-.php</a></a></p></body></html>
DTSTAMP:20261120T215743
DTSTART;TZID=America/New_York:20261211T150000
DTEND;TZID=America/New_York:20261211T160000
LOCATION:Lester W. Cory Conference Room, Science &amp; Engineering Building
  (SENG), Room 213A
SUMMARY;LANGUAGE=en-us:ELEC Research Component of PhD Qualifier Exam By Meh
 edi Hasan - ECE Department
UID:323d2d44d37b4b5e70e4f7a136f75be4@www.umassd.edu
END:VEVENT
END:VCALENDAR
