128 points by neural_markets 1 year ago flag hide 10 comments
fintech_fan 1 year ago next
This is really interesting! It could be a game changer for quantitative finance if the predictions are accurate and reliable.
machine_learner 1 year ago next
Absolutely! I'm excited to see new developments in deep learning applications. Would be good to see some comparisons with other prediction approaches.
data_sci_enthusiast 1 year ago prev next
This paper mentions some tests, but it would be nice to have a comprehensive study of the model on real-life, long- term market trend data.
code_master 1 year ago next
Agreed, and this would be an even greater contribution to the community if the framework were open source. Thoughts on this?
original_author 1 year ago prev next
It's been an ongoing debate about open-sourcing this framework. Due to ethical and legal considerations, we've decided to make it invite-only for now. However, we're making a more lightweight version open source for educational purposes as part of our broader initiative to democratize AI.
security_analyst 1 year ago next
That's an understandable approach to ensure ethical use and accountability for its predictions. Thanks for the open-source contribution!
llm_grad 1 year ago prev next
While this framework has potential, I'd also be interested in seeing how the framework handles the problem of overfitting. Can anyone share their thoughts on this? Has anyone tried it with other data sets or complex markets?
tensor_enthusiast 1 year ago next
I think this method using regularization techniques like dropout and L1/L2 weights, helps prevent overfitting quite a bit. I'd love for someone to stress-test it though!
ai_democra 1 year ago prev next
This framework might face the challenge of low interpretability, often associated with DNNs? Is there a potential for Explainable AI / Interpretable ML methods integration?
leaderboard 1 year ago next
There's definitely an discuss on integrating explainability methods with the existing model to make it more transparent.