Serving the Quantitative Finance Community

 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

February 15th, 2021, 8:26 pm

ML/AI articles would be a joy to read iff

“A good technical writer, trying not to be obvious about it, but says everything twice: formally and informally. Or maybe three times.”

Donald Knuth
 
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

February 20th, 2021, 8:28 pm

What's the jury's take on SGD these days?
Becoming a hornet's nest.
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

March 5th, 2021, 11:49 am

"It has become clear that kernel methods provide a framework for tackling some rather profound issues in machine learning theory. At the same time, successful applications have demonstrated that SVMs not only have a more solid foundation than artificial neural networks, but are able to serve as a replacement for neural networks that perform as well or better, in a wide variety of fields."

Schölkopf and Smola (2002).

Dat's 20 years ago..
 

So, what's up 2002-2022 in AI land?
 
User avatar
snowde
Posts: 9
Joined: December 14th, 2021, 4:15 pm

Re: If you are bored with Deep Networks

December 16th, 2021, 11:19 am

There has recently been some good research done in Yandex showing that compressed neural networks could in fact beet gradient boosting models like XGboost, it's worth a read https://github.com/yandex-research/rtdl
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

September 4th, 2022, 7:59 pm

Maybe Quantum NNs?
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

September 5th, 2022, 8:16 pm

Yes! Good old Fosbury had to show them how to flop.  But the flop could also be a local minimum.

I'd also say the human brain's current understanding of NN is beset by local minima, no?
There was the AI winter but now we have gradient again!  

I bet the arrival of quantum computers will help the high dimensional research a lot and keep the momentum going. Another factor is that we still have a lot to learn about possible ways we can use it. You can change your look on things and suddenly see new application domains. A lot of inspiration from that comes from biology. Things like "attention" have their NN couterpart.

Besides the research we are now also seeing lots of useful application, anyone can make disruptive tools now. A couple of days ago a friend who works at a space agency came over for dinner. They have these hackathons where employees get to do fun projects and his roommate did a cool ml project.

One of the things they do is make custom solar cells depending on the mission. Part of that is to  inspect the produced solar cells for cracks before they send them up to space. They take high resolution images with a robot and then a human goes and stares at them for weeks. However  that is now done by ML with a much better detection rate. They guy demoed it in a single day and now they are taking it to production. Faster inspection, higher quality of inspection, lower cost, shorter deadlines.
Companies that focus on quantum mechanics are earning the majority of their revenue from consulting on upcoming projects rather than from sales of practical applications, 
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

November 17th, 2022, 9:40 pm

Astronomer reinstated on Twitter after meteor video was mistaken for porn

https://nypost.com/2022/11/17/astronome ... n-mistake/
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

November 23rd, 2022, 10:05 pm

Update: Meta’s Galactica AI Criticized as ‘Dangerous’ for Science

[color=rgb(51 51 51/var(--tw-text-opacity))]Black said he asked Galactica about facts he personally knows and “in all cases, it was wrong or biased but sounded right and authoritative. I think it's dangerous.”[/color]

[color=rgb(51 51 51/var(--tw-text-opacity))]It offers authoritative-sounding science that isn't grounded in the scientific method. It produces pseudo-science based on statistical properties of science ‘writing.’ Grammatical science writing is not the same as doing science. But it will be hard to distinguish,[/color]
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

December 12th, 2022, 3:15 pm

The Internet’s New Favorite AI Proposes Torturing Iranians and Surveilling Mosques

https://theintercept.com/2022/12/08/ope ... as-ethics/
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

March 2nd, 2023, 11:56 am

Image

Artificial intelligence is everywhere—powering news feeds, curating search results and invisibly steering our lives. We talk to it and, increasingly, it talks back. And sometimes its answers seem eerily smart.

… Until they don't.


AI was presented the two pictures, one of a cow in a field and one of a cow on the beach and asked to categories items in the images it had surprising results. For the cow in the field “cow” was the first tag the AI gave the image but the AI was unable to tag “cow” with the beach image. This is because AI relies on looking for patterns to assign tags to the images and cows are commonly associated with grass so in the pictures where there was a cow but no grass the AI was unable to form the connection between the animal and the environment.

"Peaches on the beaches" also don't work.

This is a fatal flaw. Ask any cognitive psychologist.
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

March 2nd, 2023, 12:11 pm

Image
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

March 2nd, 2023, 8:22 pm

To me the remarkable thing about ChatGPT is that is can even write reasonably grammatical sentences, given that the academically lazy data scientists did not incorporate any linguistic rules into the system at all and then claim it has knowledge.
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

April 3rd, 2023, 2:11 pm

Image
 
User avatar
Cuchulainn
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: If you are bored with Deep Networks

April 19th, 2023, 1:11 pm

The curse of dimensionality (CoD) for finance applications


I am quite surprised to still read papers dealing with numerical methods attempting to solve [color=var(--color-action)]#Kolmogorov[/color] (AKA Black and Scholes) Partial Differential Equations ([color=var(--color-action)]#PDE[/color]) with deep learning technologies, for pricing or hedging purposes. We are reading such papers since 2015.

I am less surprised to learn that attempts to solve the CoD in a industrial, real life environment failed, because deep learning methods are not efficients methods when it comes to numerical simulations of PDE: I don't think that anybody will ever succeed tackling the CoD with ANNs.

Meanwhile, I would also like to remind that the CoD has been solved more than ten years ago with kernel methods. Today we solve this problem with kernel methods, and very few computational ressources (a laptop is enough), whatever the size of the portfolio or the number of risk sources are.