Wilmott Forums - Are the Neural Networks an effective weapon?

Forum Navigation:

new topic

FORUMS > Technical Forum

< refresh >

Topic Title: Are the Neural Networks an effective weapon?
Created On Sun Dec 21, 08 09:37 PM

Topic View:


	Pages: [ << 1 2 3 Previous ]

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Sat Dec 27, 08 12:13 AM

Neural Networks, Financial Trading and Efficient Markets Hypothesis - Skabar and Cloete

Have not read it carefully, but at first glance it seems to have a short, but decent review of the literature and some test data. Published in 2002.

More recently from Skabar:

A Kernel-Based Technique for Direction-of-Change Financial Time Series Forecasting

Lecture Notes In Computer Science; Vol. 5102 archive
Proceedings of the 8th international conference on Computational Science, Part II table of contents

Section: Workshop on Computational Finance and Business Intelligence table of contents

Publisher is Springer-Verlag, 2008.

"This paper presents a generative approach to direction-of-change time series forecasting. Kernel methods are used to estimate densities for the distribution of positive and negative returns, and these distributions are then combined to produce probability estimates for return forecasts. An advantage of the technique is that it involves very few parameters compared to regression-based approaches, the only free parameters being those that control the shape of the windowing kernel. A special form is proposed for the kernel covariance matrix. This allows recent data more influence than less recent data in determining the densities, and is important in preventing overfitting. The technique is applied to predicting the direction of change on the Australian All Ordinaries Index over a 15 year out-of-sample period."

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Sat Dec 27, 08 at 12:42 AM by trackstar

Quote

Top

Bottom

Cuchulainn
Senior Member

Posts: 16557
Joined: Jul 2004

Sat Dec 27, 08 10:13 AM

Quote
It's probably not loaded.

Blanks?

-------------------------
www.datasimfinancial.com

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Sat Dec 27, 08 12:43 PM

Even blanks can be dangerous at close range, esp. in the hands of an ice princess.

Anyway, it's a good picture and certainly:

Happiness is a Warm Gun

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Sat Dec 27, 08 at 01:46 PM by trackstar

Posts: 87
Joined: Nov 2008

Sun Dec 28, 08 07:01 PM

Hey macbar,

Thanks for the information.

Here are some links from me though...in exchange

. The links are from Harvard, MIT, Oxford, Cambridge, Tokyo University, Princeton(I am student at princeton btw

.
The problem with the links though is that they view the subject on very theorethical view - I mean they don't give source code of their solutions or do not even mention the software being used for the work - just formulas and long sentences with theories...
Of course, I might be wrong at some places, since I have read them some time ago...but they are still worth reading, at least for a brain teasing and exercising and better understanfing of the logic behind the networks.
But my recommendation - code your own network! This is what I am doing now on C sharp. I have one working script on Mathematica - but it generates 0.2 RMSE in the very best cases, for stocks with low volatility. I am targeting - 0.1

:

Harvard:
http://www.eecs.harvard.edu/~parkes/cs286r/spring08/cours.pdf

Princeton:
http://www.csbmb.princeton.edu/ncc/PDFs/Neural%20Economics/Cohen%20(JEP%2005).pdf

Oxford:
http://users.ox.ac.uk/~mast0315/QuRegNeuralNet.pdf

Cambridge:
http://www.alab.t.u-tokyo.ac.jp/~shino/scanned_paper/Neural_Coding_Mech_Tac_Pattern_Recog_Relative_Cont_Slowly_Rapidly_Adapt_Roughness.pdf

MIT:
http://dspace.mit.edu/bitstream/handle/1721.1/17488/46818399.pdf?sequence=1

Only the first one from Harvard is based exclusively on what everybody here is focused: time series and prediction using neural networks.
But the next from MIT, ox. cam, etc...give decent information on the NN theory as well.

Enjoy.

Quote

Top

Bottom

daveangel
Senior Member

Posts: 5126
Joined: Oct 2003

Sun Dec 28, 08 07:16 PM

Quote
I mean they don't give source code of their solutions or do not even mention the software being used for the work - just formulas and long sentences with theories...

well that shouldn't pose much of a problem for an award winning programmer like yourself ...

btw which award was that ?

Quote
The links are from Harvard, MIT, Oxford, Cambridge, Tokyo University, Princeton(I am student at princeton btw

You missed out the truly great Universities such as Hull and Bangalore. In my experience it takes someone from Oxbridge approximately 15 minutes after you first meet them to let you know they went to Oxbridge. I think in your case you did it in 25 posts...

-------------------------
Knowledge comes, but wisdom lingers.

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Sun Dec 28, 08 07:56 PM

And yet the topic does hold some interest for me...so to conduct a brief search for younger profs doing practical research in AI/Finance.

I tend to look inside *and* outside the Global Ivies for several reasons.

Among them: campus politics and bureaucracy, course load, tenure track stress, the reality of peer-review journal topics cool and uncool, and overall quality of life.

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Tue Dec 30, 08 at 01:08 AM by trackstar

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Sun Dec 28, 08 10:02 PM

somedevil,

Thank you for the links; I will have a look. Here is a book that might be of interest to you, if you have time for the archaeology of NNs.

Talking Nets: An Oral History of Neural Networks edited by James A. Anderson and Edward Rosenfeld.
The MIT Press, hardcover 1998, paperback, 2000

I believe that it is still in print:

Talking Nets - MIT Press

Good bedtime reading, imo.

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Sun Dec 28, 08 at 10:05 PM by trackstar

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Sun Dec 28, 08 10:15 PM

University settings in winter:

UCalifornia - Santa Barbara aerial

MIT Dome In Snow

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Tue Jan 13, 09 at 05:15 PM by trackstar

Posts: 87
Joined: Nov 2008

Mon Dec 29, 08 06:35 AM

Mackbar,

Thanks again...i will see it later - if I have some spare time...not to mention that I will be out of the country in a few hours(well...about 48 hours actually

.

Daveangel,

Was that a joke about Bangalore and Hull? To be quite honest, I didn't know about Bangalore...and initiated quick research online - just to see that it's not included in the Top 100 rankings of universities on 2 websites with high page rank....after that - I lost interest. Not to mention that, typing "bangalore" on google - returns results just of some Indian city and not the university itself. Call me stupid, but no - most people worldwide are clueless of the indian cities - despite Mumbai and Delhi - Bangalore sounds like a "gang bang" in all honesty

. Now...speaking of Hull - OK, I have heard of it - but then...it's not included in almost any ranking of top university.
A simple comparison between Princeton and Bangalore though.
Princeton - ranked in top 10 usually.
Bangalore - not ranked anywhere.

Documents google from princeton: over 1 million(1,220.000).
Documents google from bangalore university: 4,300(even less than that!).

Make your own conclusions...

Other than that, I will take your note of Oxbridge as a compliment

...though I don't get your overall suspicion in my programming abilities. However...I assume that you are a smart guy and you understand that like I love to say: The economics is a zero-sum game. In basic English - No, I am not that stupid to give here a source code for free, so that other traders can take advantage.If I mention my awards, this will include links to NN software that I have developed 3 years ago...sooo nope.

Have a happy new year though and wish you over 1000 times your net profit for 2008

Quote

Top

Bottom

Cuchulainn
Senior Member

Posts: 16557
Joined: Jul 2004

Mon Dec 29, 08 07:56 AM

The Long Room

-------------------------
www.datasimfinancial.com

Edited: Mon Dec 29, 08 at 07:57 AM by Cuchulainn

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Mon Dec 29, 08 12:32 PM

Try IIT - Indian Institute of Technology and IIS - Indian Institute of Science - Bangalore campus.

Cuch: Trinity College, Dublin, a bibliophile's fantasy!

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Mon Dec 29, 08 at 01:26 PM by trackstar

Quote

Top

Bottom

trackstar
Senior Member

Posts: 9600
Joined: Aug 2008

Mon Dec 29, 08 02:14 PM

A few more comments on universities, technical schools, perceived rank and innovation.

As impressive as the PR machines are at high-ranked schools, if you are looking for innovators, you will miss alot of action if you limit your search to them.

Google came out of Stanford, but University of Illinois, Urbana-Champaign produced Netscape, building ties to VC Jim Clark, and two former students of UIUC were among the founders of YouTube. Those firms did *not* come out of U Chicago, which is a common misperception.

Clark is now the father-in-law of Chad Hurley, one of the cofounders of YouTube. How is that for social networking?

.

Internationally there are places with great regional acclaim, but not well-known in the US or Europe, outside of certain academic/tech circles.

In Singapore, NTU and NUS stand out, drawing many students from the entire region. The Indian system mentioned above (IIT/IIS) has similar cachet within individual schools. We could go on and on, but others here could comment on traditionally strong or rising schools in Brazil, Russia, China,...

You also have to look for pockets of specialties. The University of Edinburgh had an AI program quite a few years ago and now has a School of Informatics. I have not visited, but I would check it out, if I were looking for people with an apparent license to do interdisciplinary work.

University of Edinburgh - Informatics

History of AI at Edinburgh

Sometimes you find dead ends, but sometimes there are diamonds.

-------------------------
"You can discover more about a person in an hour of play than in a year of conversation."
- Plato

Edited: Mon Dec 29, 08 at 02:45 PM by trackstar

Posts: 87
Joined: Nov 2008

Tue Dec 30, 08 02:39 PM

Mackbar,

Thanks...and yep, I am aware of that.
Google and most search engines have severe problems with dynamic data indexing and hence believe it or not - google can search only in 30% of the documents that are hosted online. Aside from that, a simple robots.txt file can prevent the Googlebot from indexing. Needless to say, yahoo, google and ask.com produce different results due to different spiders. I guess I should regret that

, but driven by habit, I tend to use only Google on daily basis. I almost neer switch to other engines to search for documents. Of course mp3 search, movies and software are a different story. One can use ftp, emule, other search engines, sf.net, download.com...and some google queries.

Aside from the search engines technology(this was a favourite topic of mine some time ago as I have "devoted" months of my life to them...), it is obvious that good knowledge doesn't come always from the universities.

For example, Cornelius Vanderbildt who has been the richest person of his time is known to have said: "If I had education, I will lack time to learn something useful"

- I am quoting this by memory...perhaps the original sentence is different. Also...it's not just him...even I consider about more than 50% of the topics studied here at Princeton for a total waste of time - and...I can prove it

Posts: 174
Joined: Aug 2008

Tue Dec 30, 08 03:55 PM

UIUC is actually not a bad University for a lot of subjects even if it is in the middle of nowhere.

Quote

Top

Bottom

ThinkDifferent
Member

Posts: 159
Joined: Mar 2007

Mon Jan 05, 09 03:09 AM

Quote
If I mention my awards, this will include links to NN software that I have developed 3 years ago...sooo nope.

...I think you already mentioned your awards in your 1st or 2nd post. Everyone here is highly impressed with your achievements.

Quote

Top

Bottom

daveangel
Senior Member

Posts: 5126
Joined: Oct 2003

Mon Jan 05, 09 08:43 AM

Quote
Originally posted by: ThinkDifferent
Quote
If I mention my awards, this will include links to NN software that I have developed 3 years ago...sooo nope.

...I think you already mentioned your awards in your 1st or 2nd post. Everyone here is highly impressed with your achievements.