Forums  > Basics  > SVM vs. Neural Network  
     
Page 2 of 2Goto to page: 1, [2] Prev
Display using:  

Baltazar


Total Posts: 1768
Joined: Jul 2004
 
Posted: 2007-05-07 10:17
well, i did not try it yet.
Busy doing a lot of other stuff.

Regarding the composition of classifiers, i like the stuff from Vovk quite a lot and i wonder if something usefull can come of this.

If i had time to continue this, that would be what i would check.

Qui fait le malin tombe dans le ravin

tptspecial


Total Posts: 254
Joined: Dec 2006
 
Posted: 2007-05-07 21:42

Thank you baltazaar,

Yes, VOVK papers were indeed very good. I just completed my knn classifier. it is giving me 76% correct classification. How do I make sure that this kind of performance is repeated when I go Live?? I will be including SVm, TSVM and PNN in my ensemble and see what kind of performance comes out.

I think I just have to put it into SIM mode and see some live results. I still need to work on generating Posterior probabilities so that I can only take those that are really highly probable according to the model. Seems like lot of work ahead

Thanks,

 

 


Baltazar


Total Posts: 1768
Joined: Jul 2004
 
Posted: 2007-05-08 08:59
"How do I make sure that this kind of performance is repeated when I go Live?"

That's the whole point: you cannot make sure of this.
All the machine learning is based on the idea that "it will behaves in the future as it behaved in the past".
Basic ones need the distribution to be the same and to hae independant draws
More complex ones try to relax some of this assumption.

Some thing i'd like to see is the confidence from vovk applied to financial markets.
The assumption is that distribution do not change so it is not verified here, but i'd be curious to see how robust it is.
It is related to posterios, they just call it confidence. (iirc)

Qui fait le malin tombe dans le ravin

nuno


Total Posts: 277
Joined: Jul 2006
 
Posted: 2007-05-25 03:37
you might also look at papers using boosting, esp those coming from the Penn-Lehman competition.

here's a helpful google search to get started

hedgeQuant


Total Posts: 233
Joined: Dec 2006
 
Posted: 2007-06-19 16:32

Actually, I'm more of a Support Quaternion Machine (SQM)-kind-of-guy....

Well can anybody really think of a practical application of Quaternions...

- HQ


FDAXHunter
Founding Member

Total Posts: 8371
Joined: Mar 2004
 
Posted: 2007-06-19 17:20
Yes, I can. Quaternions are used extensively throughout the 3D graphics community. Quaternion math is the de-facto standard when it comes to doing rotations in 3D space.

The Figs Protocol.

pj


Total Posts: 3421
Joined: Jun 2004
 
Posted: 2007-06-19 18:24
>Quaternions are used extensively throughout the 3D graphics community.
Right oh,
Happy birthday, FDAXHunter!

BTW you should definitely try and take some sort of holiday... good for the soul. (Donal Gallagher)

nikol


Total Posts: 594
Joined: Jun 2005
 
Posted: 2018-10-19 00:02
gekko
Disocvered this thread on searching for neural networks,
I have read (on other forums) that NNs are no longer used by the financial institutions, after some banks made losses in mid 1990s, will be good to hear abt it if some one knows..


FDAX
That would be correct. The mid/late-1990s saw a lot of interest in neural networks (including yours truly).

10 years later, nobody is using them (except for very specific and niche applications, which are usually more gimmick than anything else).

Why? Turns out that they are difficult to handle. Parsimonity is usually the key message in any financial application and there are other, better numerical tools available for handling small parameter spaces.
It can be hard to understand just what is going on inside a neural networks inner layers, so, if the system even reaches moderate complexity it can become a nightmare to just understand what is causing the system to behave the way it does. Definitely not something people in finance care for.


Today banks do it again.

Curious to see those old news or at least cases, so I can dig in
Previous Thread :: Next Thread 
Page 2 of 2Goto to page: 1, [2] Prev