Machine Learning and Neural Networks

Who claimed such? Nobody, however you made a ridiculous claim without the slightest factual backup. I just felt it was worth mentioning how wrong you are on that.

 

Ok

 

I use them. Ill say...if you think youre simply going to train a neural network to make accurate trading decisions, you are mistaken. For me, "the sauce" was figuring out how to properly stage the data for the network and how to combine the network with other tools. Sequence learning can be tricky...

 

when you said"they trained deep learning networks on asset pricing data", do you mean they feed high,low,open,close data to the DNN, and DNN produces profitable model? or you mean they first meta-label the HLOC data, and then let dnn work on those meta-labelled data?

 

Both, look up the definitions of supervised VS unsupervised learning.

 

A friend of mine is seeing some promising results building trading models using AWS SageMaker (https://aws.amazon.com/sagemaker/). It's not magic and requires upfront work but based on what he has shown me thus far it looks to be worth the effort.

 

This is what I failed to say apparently

 

There is no substitute for market knowledge which is why most programmers don't become traders. But if you have trading knowledge/smarts - I'm sure it can be a useful tool?

I'm exploring ML myself at the moment, but not just throwing raw data at it and expecting magic. Rather, I apply it to my existing data sets/variables and try to see if I can discover relationships/correlations that I have not found myself yet. I have moderate expectations, but try to keep an open mind.

 

Funny how things work. Spoke with this friend today and he has contracted me to integrate my product (FasterQuant) and what I use for live trading (AlgoTerminal) with SageMaker. Should be a fun project!

 

The balling is spreading to too many threads man.