Dear Members,
I live in Romania, am married, have three children and am in love with AI. Over the past 18 months a friend of mine and I started working on a project that regards the use of AI on the capital market.
In this project (that we now work on with 4 people) my role is the role of the product owner (product owner as in a SCRUM team).
I read a lot of research papers during those 18 months and still haven't found in any paper the approach that we are going to pursue.
We don't claim to find the “holy grail”, even though we do dream of that . But we do believe that we can get very close to it.
We know we are not the only ones that are attempting to forecast the market; a lot of bright minds are or have been working on this already.
Imagine having two soccer teams: one with a high budget and top players and one with a modest budget, unknown players but with a brilliant coach. On which team would you place your bet? (Think of Bertha Berlin who now plays for the title in Germany and defeat all big budget teams).
What I'm trying to say is that the training stage of AI is the most important things: We should be careful with what data we show and what data we don't show to the AI for learning.
Most of the studies I've read about would simply take time series and technical indicator data for the market and train neural networks or Genetic Programs on all that data and nothing more than that.
On the internet thanks to people like Klaus Meffert with JGAP, Eibe Frank and Mark Hall from WEKA, Paolo Marrone from JOONE and Steffen Nissen from FANN, and many others who have developed libraries and made their work public everyone is able to experiment with AI.
You may say that what we are doing has been done before; but we plan to filter our training data well according to certain specifications and rules. Think about what Ford, Google and McDonalds have in common.. they all took something that excited already and turned it into something better. Google didn't invent the search engine and Ford didn't invent the car.
This leads us to a well known expression: Our limits are merely the limits of our imagination.
What we would like to do is to get all those AI components together under one roof and use all of them to forecast the market.
What we would like to do is to get all those applications together under one roof and start to use in forecast. why to bring all those applications under one roof? because everyone in part have very heavy very nice good stuff and in our opinion it would do much much better forecast.
We've already made quite some progress: We've setup a server with SVN on which we work and we've almost completed data collection, (multiplexer) server and database modules and are well on our way with the AI modules.
Currently though, we are little bit short on people. We are looking for good programmers that have experience in the AI field.
We will keep you informed of our progress in this thread.
With respect,
Kerosen
I live in Romania, am married, have three children and am in love with AI. Over the past 18 months a friend of mine and I started working on a project that regards the use of AI on the capital market.
In this project (that we now work on with 4 people) my role is the role of the product owner (product owner as in a SCRUM team).
I read a lot of research papers during those 18 months and still haven't found in any paper the approach that we are going to pursue.
We don't claim to find the “holy grail”, even though we do dream of that . But we do believe that we can get very close to it.
We know we are not the only ones that are attempting to forecast the market; a lot of bright minds are or have been working on this already.
Imagine having two soccer teams: one with a high budget and top players and one with a modest budget, unknown players but with a brilliant coach. On which team would you place your bet? (Think of Bertha Berlin who now plays for the title in Germany and defeat all big budget teams).
What I'm trying to say is that the training stage of AI is the most important things: We should be careful with what data we show and what data we don't show to the AI for learning.
Most of the studies I've read about would simply take time series and technical indicator data for the market and train neural networks or Genetic Programs on all that data and nothing more than that.
On the internet thanks to people like Klaus Meffert with JGAP, Eibe Frank and Mark Hall from WEKA, Paolo Marrone from JOONE and Steffen Nissen from FANN, and many others who have developed libraries and made their work public everyone is able to experiment with AI.
You may say that what we are doing has been done before; but we plan to filter our training data well according to certain specifications and rules. Think about what Ford, Google and McDonalds have in common.. they all took something that excited already and turned it into something better. Google didn't invent the search engine and Ford didn't invent the car.
This leads us to a well known expression: Our limits are merely the limits of our imagination.
What we would like to do is to get all those AI components together under one roof and use all of them to forecast the market.
What we would like to do is to get all those applications together under one roof and start to use in forecast. why to bring all those applications under one roof? because everyone in part have very heavy very nice good stuff and in our opinion it would do much much better forecast.
We've already made quite some progress: We've setup a server with SVN on which we work and we've almost completed data collection, (multiplexer) server and database modules and are well on our way with the AI modules.
Currently though, we are little bit short on people. We are looking for good programmers that have experience in the AI field.
We will keep you informed of our progress in this thread.
With respect,
Kerosen
Attached File(s)
intelligentics-design.pdf
13 KB
|
792 downloads
Understanding the PAST, we can watch the PRESENT and look into the FUTURE