Michael Wu

Michael Wu

Auto-generated profile, based on publicly available information

Michael Wu

Computer Science, Yale University, New Haven, Connecticut, United States

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 December 2012
Verified
Conference Paper
Closed access
AVAida Vosoughi
Aida Vosoughi
MWMichael Wu
Michael Wu
JCJoseph R. Cavallaro
Joseph R. Cavallaro
To comply with the evolving wireless standards, base stations must provide greater data rates over the serial data link between base station processor and RF unit. This link is especially important in distributed antenna systems and cooperating base stations settings. This paper explores the compression of baseband signal samples prior to transfer over the above-mentioned link. We study lossy and lossless compression of baseband signals and analyze the cost and gain of each approach. Sample quantizing is proposed as a lossy compression scheme and it is shown to be effective by experiments. With QPSK modulation, sample quantizing achieves a compression ratio of 4:1 and 3.5:1 in downlink and uplink, respectively. The corresponding compression ratios are 2.3:1 and 2:1 for 16-QAM. In addition, lossless compression algorithms including arithmetic coding, Elias-gamma coding, and unused significant bit removal, and also a recently proposed baseband signal compression scheme are evaluated. The best compression ratio achieved for lossless compression is 1.5:1 in downlink. Our simulations and over-the-air experiments suggest that compression of baseband signal samples is a feasible and promising solution for increasing the effective bit rates of the link to/from remote RF units without requiring much complexity and cost to the base station.

8 May 2023 in IEEE Micro
Verified
Journal Article
Closed access
KSKarthik Sriram
Karthik Sriram
IKIoannis Karageorgos
Ioannis Karageorgos
XWXiayuan Wen
Xiayuan Wen
JVJán Veselý
Ján Veselý
NLNick Lindsay
Nick Lindsay
ABAbhishek Bhattacharjee
Abhishek Bhattacharjee
Brain-computer interfaces (BCIs) enable direct communication with the brain, providing valuable information about brain function and enabling novel treatment of brain disorders. Our group has been building HALO , a flexible and ultra-lowpower processing architecture for BCIs. HALO can process up to 46 Mbps of neural data, a significant increase over the interfacing bandwidth achievable by prior BCIs. HALO can also be programmed to support several applications, unlike most prior BCIs. Key to HALO ’s effectiveness is a hardware accelerator cluster, where each accelerator operates within its own clock domain. A configurable interconnect connects the accelerators to create data flow pipelines that realize neural signal processing algorithms. We have taped out our design in a 12nm CMOS process. The resulting chip runs at 0.88V, peraccelerator frequencies of 3–180 MHz, and consumes at most 5.0mW for each signal processing pipeline. Evaluations using electrophysiological data collected from a non-human primate confirm HALO ’s flexibility and superior performance per watt.

22 June 2023
Verified
Conference Paper
Open access
MWMichael Wu
Michael Wu
KJKetaki Joshi
Ketaki Joshi
ASAndrew Sheinberg
Andrew Sheinberg
GCGuilherme Cox
Guilherme Cox
AKAnurag Khandelwal
Anurag Khandelwal
ABAbhishek Bhattacharjee
Abhishek Bhattacharjee
Memory prefetching improves performance across many systems layers. However, achieving high prefetch accuracy with low overhead is challenging, as memory hierarchies and application memory access patterns become more complicated. Furthermore, a prefetcher's ability to adapt to new access patterns as they emerge is becoming more crucial than ever. Recent work has demonstrated the use of deep learning techniques to improve prefetching accuracy, albeit with impractical compute and storage overheads. This paper suggests taking inspiration from the learning mechanisms and memory architecture of the human brain---specifically, the hippocampus and neocortex---to build resource-efficient, accurate, and adaptable prefetchers.