Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
the cool thing about this AI is that there is no coding needed, I think that is super cool since i do not code, I sent the last two days getting the python on my macbook so that it had all to the required modules to be able to run ludwig, I got ludwig running about an hour ago, the next few days i am going to learn how to get it running and how to train it. here is a youtube video explaining how to do just that. I cant wait to get to the point that i have it training on market data.
"Learning to Trade: The Cost Of Tuition"
- a roadmap of my lessons learned as taught by the market
"Applied machine learning is basically feature engineering."
— Andrew Ng
The problem with this is virtually all the work is in pre-processing data and feature engineering. The ML algorithms in scikit-learn work practically the same way. If you had the perfect data file you just feed them in that is it. All the work though is in creating the data file.
That took me quite awhile to understand.
A support vector classifier can't be just??
clf = svm.SVC(gamma='scale')
clf.fit(X, y)
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,057 since Dec 2013
Thanks Given: 4,409
Thanks Received: 10,225
I think having a dedicated "Machine Learning and Artificial Intelligence" sub forum would be a good idea. I would also personally be willing to go through and identify old threads that should be moved into the new sub forum. Question is should it be Elite only or not? I would lean towards no but a lot of the ML threads here already are actually in Elite only sections. Could have both that would probably be overkill. Right now the ML threads are spread between "Traders Hideout", "Matlab, R project and Python", "Elite Circle" and "Elite Automated Trading".
@NJAMC @Mabi @rleplae @srgtroy what do you think?
@xplorer is this something you could do (if it has support) or would this need the big guy?
I would have never started the AI and Trading thread had I seen this thread, which I am now seeing for the first time. We do need an AI and Trading thread, perhaps with a ML subforum, to centralize what is an increasingly important subject. But I'm happy to delete my thread and/or hand it over to a more informed authority like @rleplae or @NJAMC or someone else equally knowledgeable. I was just looking for a place to talk about AI and Trading and hadn't found one.
Bottom line:
1. I think we need a centralized thread, perhaps with one or more subforums.
2. Any new threads or subthreads should not be elite, the more the merrier imho.
That is part of it, but there are other parts needed if you are not running standard IID (Independent and Identical Data) upon a well understood problem. For example, image feature detection is one that is well understood now, but you will not find much in the way of Time Series yet or Reinforcement Learning (other than a few samples).
The issue is thought needs to go into the model to make sure it doesn't over or under fit. Also, in the case of RL, stability is a problem which needs to be fine tuned. In the case of RL, the model can oscillate as it is being trained which forces an infinitely long training cycle.
I suspect an easier tool like the one Uber released is good for learning basics and experimenting with ML but might not be a good choice for a problem that hasn't been solved yet. Trading data within the stock market is a particularly difficult dataset to pre-process since it is non-stationary and non-linear. This means that some of the basic ML techniques will not work due to the IID assumption in the input data (which isn't explicitly stated within the sample problems). I don't think the Uber tool will have the flexibility to stabilize some problems.
But again, it might develop over time and I haven't gone through the technical documentation to check that statement. But proceed with caution, use this to learn about ML. Most of the top Deep Learning libraries are heading to this "No coding" type of structure (although you are coding through the config file so don't kid yourself). Tensorflow has Keras now as a preferred front end starting with TensorFlow 2.0. It is done with a simple Python script much like demoed in the youtube above.
I don't know that we have critical mass yet for another sub-forum as I don't see a lot of activity under that topic. Might be me, but seems like there is only a hand-full of us with this interest. Would a sub-forum bring more interest and help organize the information, maybe!
If one is formed, my preference would be to keep it Elite as there is potentially very powerful threads that will be formed. It is best not to have that become a product by non-Elite users by taking this communities thoughts. It seems like that should be an Elite topic. Just my 2 cents..