Computer Science
CS 7670: Seminar in Computer Systems
Seminar - 2 credits
ND
EI
IC
FQ
SI
AD
DD
ER
WF
WD
WI
EX
CE
- Deep learning (or neural networks) is popular---in fact, in my opinion---fiery.
- It improves and sometimes revolutionizes many fields, for example, CV, NLP, and biology, to name a few.
- If you work in systems (like OSes, databases, and networked systems) and are now wondering, how about systems?
- This is the seminar for you.
- This seminar studies neural networks for systems (NN4Sys).
- NN4Sys is emerging and has potential.
- People have made many proposals for NN4Sys in the past five years.
- Some are revolutionary (my opinion), some are useful, and some are failed attempts (of course).
- In this seminar, we will study two papers a week.
- One is a systems paper (could be a paper in the 1980s), and the other is a recent paper using neural networks to solve the same or a similar problem.
- One example pair is Masstree[1] and Learned Index[2].
- We will touch at least operating systems (SOSP, OSDI), databases (SIGMOD, VLDB), and networked systems (SIGCOMM, NSDI).
- Note that this is a seminar in systems.
- So, it is expected that you to have solid systems background and know how to program.
- It is not expected for you to understand machine learning or neural networks (in fact, you will learn the basics of neural networks from this seminar).
- If you're interested, join us. [1] Mao, Yandong, Eddie Kohler, and Robert Tappan Morris. "Cache craftiness for fast multicore key-value storage.".
- EuroSys'12. [2] Kraska, Tim, Alex Beutel, Ed H.
- Chi, Jeffrey Dean, and Neoklis Polyzotis. "The case for learned index structures." SIGMOD'18.
Deep learning (or neural networks) is popular---in fact, in my opinion---fiery. Show more.