user avatar
Richard Peng
@rpeng233
Associate Professor @SCSatCMU, works on efficient graph/sparse matrix algorithms. Adjunct Professor @UWCheritonCS.
Beamsville, Ontario
cs.cmu.edu/~yangp
Joined September 2018
336
Following
2,321
Followers
Posts
  • user avatar
    Richard Peng
    @rpeng233
    May 1, 2021
    "Tenure is a privilege that carries significant responsibility and obligation to students, the profession, and society at large. You are entering a new phase of your career that allows you to more creatively and fully express your work..." ... I haz tenure😅, thx 4 all teh halp
  • user avatar
    Richard Peng
    @rpeng233
    Dec 2, 2023
    The ultimate row sampling / non-linear dimensionality reduction theorem? arxiv.org/abs/2311.18145
    27K
  • user avatar
    Richard Peng
    @rpeng233
    Nov 11, 2021
    Holy bovine, a proper matrix anti-concentration bound arxiv.org/abs/2111.05553 (it works for square matrices with correlated entries, and reduces the exponent in the runtime upper bound for sparse poly-conditioned linear systems by 0.06)
  • user avatar
    Richard Peng
    @rpeng233
    Jul 21, 2025
    Replying to @jasondeanlee
    In my opinion (as contest coach), IMO`25 is a problem setting accident (rk9-91 all solved 5 problems, suppose to be 50 golds) that's almost as bad as IOI`24 (rk 10 -150 all solved 2 problems, 30 golds). Basically there were no doable medium/hard problems on either.
    6.9K
  • user avatar
    Richard Peng
    @rpeng233
    Mar 3, 2022
    Replying to @BooleanAnalysis and @thegautamkamath
    ఠ_ఠ you follow CF... Disclaimer: that post was written on a contest programming (CP) forum, where trees / tree data structures / recursions are dank memes. Thankfully these days there is (usually) @yangpliu with the more balanced / sane take:
  • user avatar
    Richard Peng
    @rpeng233
    Dec 28, 2024
    Replying to @PradyuPrasad
    I've taught design & analysis of algo, theory of computing, and discrete math. Never heard of FizzBuzz before... is it another LeetCode Meme?
    2.6K
  • user avatar
    Richard Peng
    @rpeng233
    Apr 1, 2019
    Will be talking on TCS+ this Wednesday 1700UTC on vertex sparsification based dynamic graph data structures: tcsplus.wordpress.com/2019/03/28/tcs…. Tune in to see the hybridization of eigenkitty, offline processing, and graph algorithms being presented online, with memes and sparsifiers.
    tcsplus.wordpress.com
    TCS+ talk: Wednesday, April 3rd — Richard Peng, Georgia Tech
    The next TCS+ talk will take place this coming Wednesday, April 3rd at 1:00 PM Eastern Time (10:00 AM Pacific Time, 18:00 Central European Time, 17:00 UTC). Richard Peng from Georgia Tech will spea…
  • user avatar
    Richard Peng
    @rpeng233
    Jun 24, 2025
    Oh.... deer .... (from a student's ESA`25 outcome)
    3.2K
  • user avatar
    Richard Peng
    @rpeng233
    Feb 23, 2022
    Replying to @WeinNicole
    Controversial opinion w.r.t. 2 & 3: I feel they are partly by-products of how much we under-emphasize the role of memory in problem solving processes. I feel a lot of 'quick solves/observations' are due to matching to setups that one is already familiar with.
  • user avatar
    Richard Peng
    @rpeng233
    Dec 2, 2023
    Replying to @rpeng233
    It says that any (well-behaved) function over an inherently d-dimensional space can be approximated by a subset of about d such functions, and gives an algorithm to do so. Previously this was only known for norms. The paper also significantly demystifies the concentration proof.
    2.3K
  • user avatar
    Richard Peng
    @rpeng233
    May 20, 2020
    Are there general matrix anti-concentration bounds? E.g. showing the sum of isotropic matrices reweighed by Gaussians has >1/poly(n) min-singular value with 1 - 1/poly(n) probability? The bounds that I can find seem to rely on independence of matrix entries.
  • user avatar
    Richard Peng
    @rpeng233
    Jan 22, 2020
    Congratulations Julia Chuzhoy of @TTIC_Connect, recipient of the 2020 @theNASciences Michael and Sheila Held Prize, for advances in graph algorithms, hardness of approximation, and structural graph theory! #NASaward #ComputerScience bit.ly/Chuzhoy
  • user avatar
    Richard Peng
    @rpeng233
    Sep 12, 2021
    998244353
  • user avatar
    Richard Peng
    @rpeng233
    Feb 16, 2020
    Replying to @chrmanning
    As someone who has looked at some of the 10^7 lines of code in Sandia's Trilinos package (for solving large linear systems), I feel there is also a factor of at least 10 or so from the software engineering/legacy aspects. This type of inefficiency is far more complicated though.