3550742
doi
10.5281/zenodo.3550742
oai:zenodo.org:3550742
user-cernopenlab
Real-Time Server Monitoring and CNN Inference on FPGA
Debdeep Paul
info:eu-repo/semantics/openAccess
Creative Commons Attribution 4.0 International
https://creativecommons.org/licenses/by/4.0/legalcode
CERN openlab
summer-student programme
<p>Neutrinos are subatomic particles, very similar to an electron, but without any electrical charge and <br>
a very negligible rest mass. They are the most abundant and perhaps the most mysterious matter <br>
particles in the universe! </p>
<p>Detecting and tracking neutrinos poses a different challenge both in terms of hardware and <br>
algorithms. Since they have very little interaction with matter, they are incredibly difficult to detect. <br>
So, every bit of information produced by the neutrino interactions is required to be stored and <br>
analyzed at an incredible fast rate. </p>
<p>Present day State-of-the-Art Convolutional Neural Networks (CNNs) have shown an un-precedented <br>
accuracy on many modern AI applications. They have become the de-facto standard for a wide range <br>
of tasks ranging from computer vision to machine translation due to their high accuracy and <br>
robustness. </p>
<p>Research has proved that CNNs also hold the potential to excel in High Energy Physics (HEP) <br>
applications as compared to traditional methods in identifying particle interactions in sampling <br>
calorimeters or Time Projection Chambers. However, these algorithms are both computationally <br>
and memory extensive which limits them from running on normal CPU for real-time and power <br>
constrained applications. Thus, customized hardware implementation of machine learning <br>
algorithms can be a promising solution for higher performance and improved throughput. </p>
<p> </p>
<p>Another important aspect is to constantly monitor the server which hosts the FPGAs for the CNN <br>
inference and sent alert messages in crucial situations. </p>
<p>This project involved the development of a lightweight, easy to use solution for the server <br>
monitoring so that all the important metrices can be visualized at a glance in real-time. Secondly, to <br>
compare the CNN inference between the server and the FPGA in terms of accuracy, power <br>
consumption and latency. The project also focuses on dealing with the numerical overflow problem <br>
due to fixed point computations in FPGA during CNN inference. </p>
Zenodo
2019-11-22
info:eu-repo/semantics/report
3550741
user-cernopenlab
1689325998.167568
1351743
md5:401b55ca9d73b87b7afbbaccec12f5e2
https://zenodo.org/records/3550742/files/Report_Debdeep_Paul.pdf
public
10.5281/zenodo.3550741
isVersionOf
doi