Arash

Interested in Computer Science & Mathematics, @ELLISforEurope PhD student, Sharif alum, INTJ

[Arash @ArashJ_m]

@Matiuee اگه چیزی میزان سختیش خیلی زیاد به نظر میاد، معمولا معنیش اینه هنوز کامل بکگراند مورد نیاز (مثلا کورس های پیش‌نیاز) رو نداری. با مطالعه بیشتر بهتر میشه.

[Arash @ArashJ_m]

کتاب Linear Algebra Done Right به فارسی ترجمه شده! به نظرم یکی از بهترین کتاب ها برای یادگیری جبر خطی هست که هم rigorous هست و هم شهود خیلی خوبی میده.

[Sheldon Axler @AxlerLinear]

The translation of my book Linear Algebra Done Right into Farsi is now freely available at https://t.co/ii1ovMFKvH (click on the fourth link on that webpage to get the Farsi version).

[Dmitriy Drusvyatskiy @ddrusvyat]

This spring I taught a grad course on “All of Convex Quadratic Optimization.”

That may sound narrow, but quadratic problems are simple enough to analyze precisely, yet rich enough to reveal many of the core phenomena that drive modern optimization and learning.

I wrote up the course notes here:
https://t.co/kuLZi4rFRG

[Arash @ArashJ_m]

@ehsanasgharzde اگر تو زمینه مشخصی ریسرچ نمیکنی، به نظرم مطالعه کورس های advanced دانشگاه های خوب (ایرانی یا خارجی) فایده بیشتری از خوندن مقاله رندوم داره.

[Arash @ArashJ_m]

@Alirezahasstam تعداد بالاتر پیپر لزوما معنی خوبی نداره. مخصوصا الان که با LLM ها راحت میشه بدون داشتن ایده خوب پیپر تولید کرد.

[Mahdi Soltanolkotabi @mahdisoltanol]

A linear autoencoder learns PCA. This has been understood since Baldi–Hornik ’89.

But what happens if you make the model nonlinear by adding just a ReLU in the middle?

Surprisingly, the training dynamics of even this minimal nonlinear variant remain poorly understood.

In our paper, we take a stab at this by characterizing how a one-hidden-layer ReLU network, trained from random initialization while updating both layers, recovers a linear model.

https://t.co/b5D6M87xR9

No fixed features. No preprocessing tricks. No “train only the last layer” gimmicks.

Just standard gradient descent on the empirical loss, with both layers trained end-to-end.

At first glance this may sound almost comically simple: “Congrats, you proved neural nets can learn linear functions?!”

But the dynamics turn out to be surprisingly subtle, and the proof techniques ended up being some of my favorite ones we’ve developed in the last few years.

Quick thread below on why this problem is challenging 🧵

I also genuinely think the techniques here may be useful for understanding more contemporary models and training phenomena, including reasoning and post-training. Stay tuned!

[Stat.ML Papers @StatMLPapers]

High-Dimensional Statistics: Reflections on Progress and Open Problems https://t.co/EvglC0dVqH

[Frank Nielsen @FrnkNlsn]

Book #OTD

"The Cauchy-Schwarz Master Class"

https://t.co/JdDkOjzAqK

[Caglar @caglar_ee]

Video lectures, CMU 10 799 Diffusion & Flow Matching spring 2026, by Yutong (Kelly) He
https://t.co/PXxEffP5Ql

[Lenka Zdeborova @zdeborova]

Challenge for ML theory folks:
What’s the cleanest solvable data model where nonlinear autoencoders provably beat PCA?
We suggest one here: https://t.co/HVpiXLCCIG
Happy to collect pointers & ideas!

[Elad Hazan @HazanPrinceton]

Psyched to be teaching COS511 (Theoretical ML) again! 🚀 Course website below. New this year: video lectures—coming soon.
https://t.co/ninbmNdbYB

[Arash @ArashJ_m]

@bemoniri @docmilanfar Could you share some interesting resources? Especially video lectures!

[Math Cafe @Riazi_Cafe_en]

UC Berkeley's "Machine Learning" lecture notes

PDF: https://t.co/ktJDR3gUEf

[Gappy (Giuseppe Paleologo) @__paleologo]

1. Ridge regression is heavily used in systematic investing both in the p << n and in the p >> n cases (last one, less so). I don't think that the use is very deeply motivated, other than the old standard argument in favor (see, e.g., El. Stat. Learning).

[Arash @ArashJ_m]

@pseudophakic_sh اختیار دارید. قبلا ندیده بودم این کتاب رو ولی الان یه نگاه انداختم به نظر جالب میاد :دی

[Arash @ArashJ_m]

@SharabKhanum ویدئوهای یادگیری ��اشین از دکتر مهدیه سلیمانی از دانشگاه شریف به نظرم خوب هستن. توی سایت مکتب خونه میتونی پیداش کنی.

[Math Cafe @Riazi_Cafe_en]

Probability Theory and Computational Mathematics
by Joel Tropp

PDF: https://t.co/j4a9vSvcst

[Stat.ML Papers @StatMLPapers]

Lecture Notes on High Dimensional Linear Regression https://t.co/aBRU4v9rY7