use machine learning to discover perfect trading strategies

 

Last year the best human go player in the world was beaten by a machine, now a machine can beat any human in poker, or basically any game, so really it seems 

like a good idea to create a software that test trades multiple times and get better, the difference is that a machine can trade and test billions of trades in a week, that a 

trader would take a life, so how can i code this EA that learns? do you know any book that can help on this, what should i look into to be able to code this self learning EA?

 
mrluck012:

Last year the best human go player in the world was beaten by a machine, now a machine can beat any human in poker, or basically any game, so really it seems 

like a good idea to create a software that test trades multiple times and get better, the difference is that a machine can trade and test billions of trades in a week, that a 

trader would take a life, so how can i code this EA that learns? do you know any book that can help on this, what should i look into to be able to code this self learning EA?

“Machine Learning” is highly complex and requires an extremely advanced mathematical, statistical and coding knowledge and skill!

You can barely code the simplest of scripts. How do you expect to be able to learn and code “machine learning” at your level of skill?

Before you can run, you must first learn to walk, and before you can walk, you first need to learn to stand up right and before that learn to crawl first!

“Machine Learning” is way, way, way out of your league at the moment! Dedicate some time and effort to becoming a better and more skilled coder first!

 

my code improved a lot since our last talk, now i can code any ea that i want, so what books do you recommend to learn that complicated topic? i can see the term "black boxes" but i think machine learning is 

more advanced than that. Where do i start?

 
mrluck012:

my code improved a lot since our last talk, now i can code any ea that i want, so what books do you recommend to learn that complicated topic? i can see the term "black boxes" but i think machine learning is 

more advanced than that. Where do i start?

If your skill was advanced enough, you would have already known the answer to your question. That fact that you are searching for such material, shows that you are no where near the level of expertise required.

You still don't get it - it takes many years to reach that level!

But if you really insist on it, just do simple search here on this very site or better still, on Google. You will find plenty of material to get lost in while you go down the rabbit hole to Wonderland!

PS! Here on this site, search for Neural Networks which is one aspect of A.I. but not the only one!

EDIT: Here is a comparison of different Deep Learning frameworks:

Knock yourself out, but don't complain if you are not able to understand it!

Comparing Top Deep Learning Frameworks: Deeplearning4j, Torch, Theano, TensorFlow, Caffe, Paddle, MxNet, Keras & CNTK
  • Chris Nicholson, Adam Gibson
  • deeplearning4j.org
Deeplearning4j is distinguished from other frameworks in its API languages, intent and integrations. DL4J is a JVM-based, industry-focused, commercially supported, distributed deep-learning framework that solves problems involving massive amounts of data in a reasonable amount of time. It integrates with Kafka, Hadoop and Spark using an...
 
mrluck012:

my code improved a lot since our last talk, now i can code any ea that i want, so what books do you recommend to learn that complicated topic? i can see the term "black boxes" but i think machine learning is 

more advanced than that. Where do i start?


In increasing difficulty, try the following books:

1. https://www.amazon.com/Introduction-Statistical-Learning-Applications-Statistics/dp/1461471370/ref=sr_1_1?ie=UTF8&qid=1504131012&sr=8-1&keywords=statistical+learning

2. https://www.amazon.com/Elements-Statistical-Learning-Prediction-Statistics/dp/0387848576/ref=sr_1_2?ie=UTF8&qid=1504131012&sr=8-2&keywords=statistical+learning

3. https://www.amazon.com/Pattern-Recognition-Learning-Information-Statistics/dp/0387310738/ref=sr_1_2?ie=UTF8&qid=1504131038&sr=8-2&keywords=pattern+recognition


There is no need to say that you need calculus (and some advanced calculus), linear algebra and introductory probability as a background.

Also some coding skills in R or python would help, as most of the books about the subject use one of them.

 
yes that seem very complicated what i see on deeplearning4j.org , please send me free stuff. Imagine that, you just click a button then you have perfect systems, those who understand machine learning can make huge profit
 
mrluck012: yes that seem very complicated what i see on deeplearning4j.org , please send me free stuff. Imagine that, you just click a button then you have perfect systems, those who understand machine learning can make huge profito they don't

No they don't! Profit does not come that easily! Even with "machine learning", the Market does not make it easy for any retail trader!

 

If there is such a thing as what the thread imply then it could have done already, but market behavior cannot be learned easily even by computers because it is governed by emotions and emotions is always random and unpredictable, trading is not a science that is why it cannot be coded into something a perfect strategy.

 
I think many miss a big issue, the hardware.  
 

I think I hit the mother lode when it comes to resources on machine learning. There is easily months and months of learning resources in this thread.

I have started a course on Quantra on machine learning which has a lot of things on offer but I will still bookmark this thread for future use.

But coming back to the first thought which started the thread, I guess machine learning is eventually going to influence (if not replace manual) trading in a big way in the future. So fingers crossed, I would like to be on the winning side when this happens.