iiinova — where your digital self keeps evolving

Peyman Milanfar @docmilanfar.bsky.social · 14/04/2025

I was invited to give this plenary talk at the Electronic Imaging Symposium this past February. The recording is now available on YouTube www.youtube.com/watch?v=s59U...

0141

Peyman Milanfar @docmilanfar.bsky.social · 13/04/2025

Silicon Valley crosswalk buttons have been hacked to hilarious effect

1276

Peyman Milanfar @docmilanfar.bsky.social · 12/04/2025

What I’ve learned here is that some will not get it, even if explained in the simplest terms.

100

Peyman Milanfar @docmilanfar.bsky.social · 12/04/2025

it's called "volatility drag" - look it up. and it doesn't need fixed percent. I used a fixed number so it would be easier for people to understand. apparently it wasn't easy enough

100

Peyman Milanfar @docmilanfar.bsky.social · 12/04/2025

Most people don't understand that volatility alone can kill your investments start with capital P lose 5% one week: (1-0.05)*P gain 5% the next week: (1+0.05)*(1-0.05)*P. repeat the pattern for 1yr (52 weeks): (1+0.05)^26 *(1-0.05)^26 *P = 0.88P You lost 12% of your money.

3121

Peyman Milanfar @docmilanfar.bsky.social · 07/04/2025

It's a cruel irony that those who orchestrated this stock market crash are already insulated from its effects. They will be fine. The real tragedy is that it's the ordinary person – the retiree, the small investor, the hardworking individual – who will suffer the most.

1171

Peyman Milanfar @docmilanfar.bsky.social · 05/04/2025

"That Mr. Musk has come to hold so many of the same beliefs about social engineering and economic planning as his grandfather is a testament to his profound lack of political imagination, to the tenacity of technocracy and to the hubris of Silicon Valley." www.nytimes.com/2025/04/04/o...

nytimes.com

Opinion | The Rise and Fall of Elon Musk (Gift Article)

His belief that liberal democracy has failed and that technologists should lead can be traced to the unusual life of his grandfather.

0140

Peyman Milanfar @docmilanfar.bsky.social · 04/04/2025

Warning signs you're dealing with a cynic: * says “that’ll never work...” before any evidence * often assumes the worst of people * every interaction is a 0-sum game * uses criticism to signal competence 2/2

1233

Peyman Milanfar @docmilanfar.bsky.social · 04/04/2025

Every profession has its personality types. In my experience the most difficult people to work with are cynics - no matter how talented they may be. Healthy skepticism is positive and curious. Cynicism is not - it's judgmental. The difference is under-appreciated. 1/2

1292

Peyman Milanfar @docmilanfar.bsky.social · 04/04/2025

A wonderful scholar and person, he is. Taught me so much.

010

Peyman Milanfar @docmilanfar.bsky.social · 03/04/2025

Meta has one VP for every 300 employees. They’re a dime a dozen.

010

Peyman Milanfar @docmilanfar.bsky.social · 03/04/2025

This tour de force overview by the incomparable Alan Willsky (my former advisor) is well worth reading, and a great reference for anyone studying AR models for generation. citeseerx.ist.psu.edu/document?rep...

170

Peyman Milanfar @docmilanfar.bsky.social · 03/04/2025

Autoregressive models have become more popular recently. Speaking to people, I get the sense most folks are unaware of the history of the topic in signal processing. In particular, AR across scale got a lot of attention in the 90s. There are a lot of great papers on the subject but one stand out…..

2485

Peyman Milanfar @docmilanfar.bsky.social · 31/03/2025

I think the "zero" connotation came before "digit" though. So perhaps arabic inherit that more directly from sanskrit

000

Peyman Milanfar @docmilanfar.bsky.social · 31/03/2025

A cipher (or cypher) is an algorithm for performing encryption or decryption - the foundation of all of cryptography research and practice. Not widely appreciated is that the word cipher has its direct origin in the arabic word صفر (sifr) meaning zero. Andصفر itself roots to sanskrit शून्य (śūnya)

3232

Peyman Milanfar @docmilanfar.bsky.social · 27/03/2025

A man walks into a library and asks whether they have a book on Pavlov’s dog and Schrodinger’s cat. The librarian thinks for a moment and says “it rings a bell, but I’m not sure if it’s in or not”

0245

Peyman Milanfar @docmilanfar.bsky.social · 26/03/2025

I posted this thread four months ago. To my knowledge not a single one of these 10 basic features has been implemented on this website. It can’t be that hard.

250

Peyman Milanfar @docmilanfar.bsky.social · 21/03/2025

Bonus: this form is flexible enough to adapt both the length and to different audiences. n/n

020

Peyman Milanfar @docmilanfar.bsky.social · 21/03/2025

And finally, use the last 5-10 minutes to summarize and talk about what you think are important new directions to pursue in research and how these could have impact for the team and the products of the company, or the department and their areas of interest and growth. 3/n

140

Peyman Milanfar @docmilanfar.bsky.social · 21/03/2025

Second, you can spend the next 20-25 minutes doing a deep dive on just one of them - say, the one you're most proud of, or the one that matches the hiring team/departments interests and needs best. 2/n

130

Peyman Milanfar @docmilanfar.bsky.social · 21/03/2025

First, a short snippet of your overall set of technical accomplishments. For instance, you might spend the first 15 minutes giving short overviews of a few of your works/papers - it's also very good to draw a theme across them if there is one. 1/n

220

Peyman Milanfar @docmilanfar.bsky.social · 21/03/2025

In some ways it's harder than ever to land a research job these days - both in academia and in industry. As you prepare for a job talk, consider giving it the following structure - it's hard to give a bad talk that's built like this.: 0/n

3303

Reposted by Peyman Milanfar

Dmytro Mishkin @ducha-aiki.bsky.social · 19/03/2025

The Power of Context: How Multimodality Improves Image Super-Resolution Kangfu Mei, Hossein Talebi, Mojtaba Ardakani, Vishal M. Patel @docmilanfar.bsky.social Mauricio Delbracio tl;dr: condition SR model on RGB+depth+segmentation (which you can predict from RGB)-> PROFIT arxiv.org/abs/2503.14503

0153

Peyman Milanfar @docmilanfar.bsky.social · 19/03/2025

I had these slides in a deck for a couple of years. For tired of seeing people use SURE to derive Tweedie. Though I’m sure it will keep happening. 😀

120

Reposted by Peyman Milanfar

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

Tweedie's formula is super important in diffusion models & is also one of the cornerstones of empirical Bayes methods. Given how easy it is to derive, it's surprising how recently it was discovered ('50s). It was published a while later when Tweedie wrote Stein about it 1/n

16414

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

One more thing: Most people I've seen recently derive Tweedie from Stein's Unbiased Risk Estimator (SURE). I don't get that - it's messy and requires much stronger regularity conditions. I mean - Tweedie invented the formula before Stein invented SURE after all! @gabrielpeyre.bsky.social FYI

360

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

So to build an MMSE denoiser, all you need is the marginal density P(y) of the noisy measurements; which you can empirically build given just noisy samples. This explains why you don't need clean image to learn to denoise, or to estimate score. 7/n

150

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

And.... you have Tweedie's formula for the MMSE denoiser in explicit form. Specifically, it's the sum of a "trivial" max likelihood estimate; and a Bayes "correction" 6/n

161

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

Now divide both sides by P(y). This is not problematic since P(y) - being a Gaussian "blurred" version of the prior P(x) - does not vanish. 5/n

141

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

The key to arriving at Tweedie's formula is to differentiate the marginal density P(y). 4/n

140

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

Note that the normalizing term in the denominator is the marginal density of y. 3/n

130

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

The MMSE denoiser is known to be the conditional mean f̂(y) = 𝔼(x|y). In this case, we can write the expression for this conditional mean explicitly: 2/n

130

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

16414

Peyman Milanfar @docmilanfar.bsky.social · 18/03/2025

Why not do us both a favor and stop following me.

100

Peyman Milanfar @docmilanfar.bsky.social · 17/03/2025

"This is a pretty big accusation. Have you asked them if they’ve done such a thing? Easy for them to say no and experience the consequent decline in reputation were it to come out. I know folks involved with this and they’re ethical people, I can’t imagine them doing this"

100

Peyman Milanfar @docmilanfar.bsky.social · 16/03/2025

There’s no merit to this work except as a deliberate and troubling strategy: to both create a heavily-cited, machine-generated 'first'; and to effectively normalize such publications. This agenda should be deeply concerning to the AI community. 2/n

170

Peyman Milanfar @docmilanfar.bsky.social · 16/03/2025

Don’t be fooled - SakanaAILabs has very likely already submitted dozens of manuscripts to various conferences to test out its model before productionizing in this sham “experiment”. It’s not an experiment, it’s arbitrage - a core expertise of David Ha from his finance days 1/n

1160

Peyman Milanfar @docmilanfar.bsky.social · 16/03/2025

061

Peyman Milanfar @docmilanfar.bsky.social · 16/03/2025

Insights we have been teaching in image processing courses for decades. Nice to see computer vision catching up. 😀

120

Peyman Milanfar @docmilanfar.bsky.social · 16/03/2025

The point of this post is not whether the idea is novel or not, but rather that this company has very likely already submitted dozens of manuscripts to various conferences to test out its model before making public this sham “experiment”.

320

Peyman Milanfar @docmilanfar.bsky.social · 12/03/2025

The tech industry in Silicon Valley runs on the considerable talents and work ethic of folks who were not born here. Many of these folks were also trained in this country in our Universities. This is our strength. If we mess with this virtuous cycle, it is at our own peril.

0252