Average person.
Joined January 2021
- Tweets 215
- Following 51
- Followers 69
- Likes 46
128 Photos and videos
Jun 14
I am invited to Banff International Research Station for Math Innovation & Discovery birs.ca/ program: Complex Optimization & Variational Inequalities at Institute for Adv. Study in Math
I thank the organizers & hope my Oct27 teaching schedule will not burn me š
1
25
Jun 10
My collaborators will present our joint work at NUMTA26 Calabria & EUROPT26 Linz
We attack a bilevel nonsmooth nonconvex problem that I don't know how to solve
I hide as the 3rd (last) author. Please point your cannon criticism at me. (...be gentle. I will diverge to infinity)
17
with biology people, we use matrix-exp, FrƩchet derivative & Peterson-Stochmeyer method to exactly solve dx/dt=A(x-x*) for some gene learning, our approach beat other methods (I don't know biology, I just want to learn & do applied math for fun)
10.64898/2026.05.18.726023v1
2
487
during SIAMOPT26 Edinburgh
my Waterloo MMath student Sammy passed thesis defence, joining KAUST
my Southampton (to-be) PhD student James got full scholarship
I learned metric entropy-free complexity & game theory for computational brain damage
first time in Scotland (freedommm!)
45
I will be in SIAM Conf. Optimization 26 presenting how to make everybody happy by mixing subdifferential graph theory network economics stochasticity Bernstiein inequality epsilon-net covering sparsity trust-region sequential quadratic program constraint qulification blablabla..
3
123
May 18
2 papers accepted in EUSIPCO26
1.Agnostic Blind Source Separation on Subsea Cables via Sum-of-Norms Bispectrum Tensor Factorization
What: cumulant kill Gaussian
2. NMF on Oblique Manifold For Earth Observation Data Analysis
What: land change over time by oblique manifold NMF
4
186
May 12
Tomorrow: my first tele MMath thesis defense
Sammy Shi will defend "On initializations of NMF"
No spoilers (see abstract for detail)
Just to post here to support Sammy, he is reaaaaally gooooood student
52
May 11
I let my (good) undergrad student to experience his first
bloodbath: giving presentation on formalization of optimization algorithm convergence theory in front of logic dudes: model theorist/constructivist/intuitionist
(I guess no ultrafinitist there?)
1/n
1
4
113
May 11
after hearing several seminars givne by logic dudes, I realized that math dudes "have common ground" while logic dudes fight with each other: they run 3000 steps to prove (a b) c=a (b c), which is crazy for āworking mathematicianā
2/n
1
48
May 11
hearing questions from these logic dudes on Lean, now I donāt know where is math moving to, maybe to vacuum, may be to truth, but not my issue, I let my grant funder decide
(a small paper on automating optimization proof in lean incoming later, maybe)
n/n
1
52
pre-LLM era, I always "see" convex analysis in epigraph, which is more "intuitive" to me
now, outsourcing (some) analysis to AI, I teach convex analysis (indicator, prox, proj, conj) all in epigraph, focus more on the meaning, at least one student told me he like it haha
46
On Monday, one of my UG Part-III Project student Xinyuan Guo will be presenting his work on Logic Seminar Ā
Monday 11-May-2026 13:00-14:00 Building 32 Room 3079
University of Southampton Ā
A Lean 4 Framework for Formal Verification of Selected Optimisation Algorithms
1
75
finally find a use of category
1. we hv a function v: AāY on sets A,Y
2. actually we don't care about a specific v but ALL v: we let Yį“¬ be set of all v
3. if function Ī¦: AāYį“¬ is surj, then all f: YāY has fix-pt
(nice to expand my math-1 notes on Cantor's diagonal argument)
11 Sep 2025
wikipedia is useful for looking up obscure category theory results -- though even for that I imagine LLMs are already better
otherwise it's a partisan fever swamp
47
Apr 24
I joined the University of Waterloo Faculty of Maths as Adjunct Professor
<long paragraph of AI-generated gibberish filler words>
[End of announcement]
2
61
Apr 21
Prof.Martinez took 3000yr old tablet image
the data stored behind my office, in the world I am the only CS professor 2meter from 3000yr old Mesopotamiaš¤£
if I run matrix algo, I am the FASTEST Natural Language Processing human with 10Ā¹ā¹bit/s
(time to convert it to homeworkš¤£)
72
Apr 18
Eigendecomposition is a special case of SVD
Inner product is a special case of bilinear form
Matrix is a special case of Representation Theory
Euclidean distance is a special case of Bregman divergence
Gaussian distribution is a special case of log-momment generating function
Apr 16
Ordinary Least Squares is a Special Case of Transformer ift.tt/h3UFDBs
2
124
Apr 17
Reading 11pm while having conference 9am
I just learned that cumulant is the coordinate of graded tensor algebra and it gives a much better explaination why 4th order cumulant kills Gaussian noise
(should never read or study midnight before sleep again!!!!!!!!)
2
52
Apr 15
reading that topolgical Galois theory is quite joyfull during bus trip (I learned something new!)
Apr 14
My thoughts on this hit "universal gate of continuous functions" (as some have suggested). TL;DR to save you a click: For a more conventional definition of "elementary function", if my math is right, it doesn't do what the tin says. š
61
Mar 25
Three humans in a room.
Algebraic-geometry PhD student talking varieties, rings, Nullstellensatz to my visiting topology PhD student on quotients, kernels, persistent homology, bifiltration.
Me (analysis supervisor): WTF you two talking ā does it converge?
4
156