There was quite an energetic discussion on discord about the Neurolab Extension.
https://discord.com/channels/775345259642748928/819002533111529543
Carova in particular had some good suggestions and insights, some of which Dion implemented.
This thread had 9 feature request votes and discussed features in Neurolab a few years back: https://www.wealth-lab.com/Discussion/Additional-Machine-Learning-Algorithms-and-GPU-support-6014.
My knowledge of ANNs ended when I stopped using them in about 2010. So interested in what the community here thinks may be worthy of a Neurolab feature request in 2024.
For me it would be:
1. Optimal signal. Predict buy and sell directly ala TradingSolutions. https://www.youtube.com/watch?v=HeVn0riFoy4
2. Recurrent Neural Network support as they are well suited to time series analysis. Though I note that the Sharp Learning Library does not support this net. https://github.com/mdabros/SharpLearning
3. Pre-processing as mentioned in the discord channel. ROC is probably the most important one and already available as an input.
Interested in the communities opinions of what new features in Neurolab could be.
https://discord.com/channels/775345259642748928/819002533111529543
Carova in particular had some good suggestions and insights, some of which Dion implemented.
This thread had 9 feature request votes and discussed features in Neurolab a few years back: https://www.wealth-lab.com/Discussion/Additional-Machine-Learning-Algorithms-and-GPU-support-6014.
My knowledge of ANNs ended when I stopped using them in about 2010. So interested in what the community here thinks may be worthy of a Neurolab feature request in 2024.
For me it would be:
1. Optimal signal. Predict buy and sell directly ala TradingSolutions. https://www.youtube.com/watch?v=HeVn0riFoy4
2. Recurrent Neural Network support as they are well suited to time series analysis. Though I note that the Sharp Learning Library does not support this net. https://github.com/mdabros/SharpLearning
3. Pre-processing as mentioned in the discord channel. ROC is probably the most important one and already available as an input.
Interested in the communities opinions of what new features in Neurolab could be.
Rename
Just a side note, we switched from SharpLearning to Encog as the NN engine, but aren't opposed to switching again if something better comes up in the discussion.
https://www.heatonresearch.com/encog/
https://www.heatonresearch.com/encog/
Capabilities to generate dendrograms in order to identify families of indicators who are too correlated, in order to prevent redundancy, would be a good add on.
I asked ChatGPT what it "thinks". RNNs dominate the response (1, 2, & 7) and some new ones I have never heard of. How meta is this!
Your Response
Post
Edit Post
Login is required