kumoh national institute of technology
Networked Systems Lab.

Review Comment

NSL > Works@NSL> About Review> Review Comment
Cosmas Ifeanyi Nwakanma, Rubina Akter, Jae-Min Lee, and Dong-Seong Kim,"NM500 : An Artificial Intelligence Chip based on Neuromorphic Technology", IEEE Embedded Systems Letters
By : Cosmas
Date : 2020-10-20
Views : 31

19-Oct-2020

Dear Professor Kim:

In view of the reviewer feedback enclosed at the bottom of this letter, I regret to inform you that manuscript # IEEE-ESL-Aug-20-0119 entitled "NM 500: An Artificial Intelligence Chip based on Neuromorphic Technology" submitted to the Embedded Systems Letters has been rejected.

Thank you for considering the Embedded Systems Letters as a venue for the publication of your research. I hope the outcome of this specific submission will not discourage you from the submission of future manuscripts.

Sincerely,
Professor Preeti Ranjan Panda
Editor in Chief, Embedded Systems Letters
panda@cse.iitd.ac.in

AE Comments to Author:

Associate Editor: Jeyapaul, Reiley
Comments to the Author:
The authors present a good understanding of neuromorphic processors, and their use in system design. The authors have presented a report on the application of NM500 processors for laser marking. They have also presented their system design and configuration. There is a need for research on cross-layer design explorations for safety-critical systems.

There are a few key points, from the reviews, which I summarize here for your attention:
1) As the authors note, not much work has been published on the topic of NM500 applications. The authors previous work [14] is the only other work published on the NM500 processor. However, we observe that the submitted work is a report on another case-study on the use of NM500 processor for ASDAS application. The consensus is that the submissions lacks in novelty of technical contribution.

2) The experiments and evaluations do not highlight on why the NM500 is different and unique in comparison to other neurmorphic processors. In addition, the comparitive studies do not include other processors and also are not normalized against established benchmarks (identified in [11]) in literature.

The authors have presented some detailed experimental evaluations on this case-study. For a potential resubmission of the research content, I would recommend that the authors reorganize the manuscript to include details and explain the research process in a way that would help future system designers and researchers to appreciate the use of the NM500 processor (and similar other processors).

Reviewer(s)' Comments to Author:

Reviewer: 1

Comments to the Author
This paper concerns the implementation of an image classification system.

No real details are given about either the hardware or the image recognition task. I also can not see any novel contribution or comparison with previous work. Therefore, I cannot recommend this as a scientific paper.

Reviewer: 2

Comments to the Author
This paper introduces the reader to the NeuroMem NM500 chip but the motivation for the research is not adequately explained. In addition, there are no meaningful comparisons with other state-of-the-art systems based on neuromorphic technology. One of the central claims of the paper is that recognition time is independent of the dataset size but this is not substantiated.