123 points by mlwhiz 6 months ago flag hide 23 comments
hacker1 6 months ago next
This is really impressive! I can't wait to see how it's applied in real-world scenarios.
datascientist 6 months ago next
I've been wondering about its limitations as well. Have the developers shared any concerns or areas for improvement?
codewizard 6 months ago next
One limitation I've seen mentioned is the training set size requirement. It seems to work optimally with at least 100k records. What do you think about that, users?
hacker4 6 months ago next
That's a good point. I was wondering about working with smaller data sets to save on training time.
mlexpert 6 months ago prev next
I've been following the development of this project, and the results are truly astonishing. Great work to the team!
hacker2 6 months ago next
Do you think it's possible to implement it in my project? I'm building a predictive analytics tool that could benefit from the added performance.
hacker2 6 months ago next
That would be monumentally helpful. I'd appreciate any guidance from you or the developers.
hacker3 6 months ago next
The tutorial helped a lot! Thanks for providing clear instructions.
hacker5 6 months ago next
Have you tried the new HN search bar? Type in "MLExpert tutorial" to find the tutorial directly.
mlexpert 6 months ago prev next
Absolutely! This algorithm can be integrated with your tool by adding a few additional lines of code. I'll write up a tutorial on how to do that later today.
mlexpert 6 months ago next
The tutorial is now up on GitHub. Interested users, please see the link below.
newuser 6 months ago prev next
@MLExpert Can the algorithm be applied to non-numerical data?
mlexpert 6 months ago next
Not out-of-the-box, but it can be adapted to process non-numerical data with basic preprocessing techniques.
aienthusiast 6 months ago next
@MLExpert Thanks for sharing your thoughts on the non-numerical data. Any chance of a quick example using pandas?
mlexpert 6 months ago next
I'll create an example and post it in a separate thread. Keep an eye out for it!
statsguru 6 months ago prev next
Have you done any benchmarks comparing it to Google's AI/ML products or types of architectures?
mlexpert 6 months ago next
Yes, we're planning on sharing the benchmarks next week, although it seems like a tight race against some of Google's solutions.
nerdherd 6 months ago prev next
Will this algorithm be open-sourced, or would you consider licensing it for production use?
mlappsteam 6 months ago next
We're seriously considering open-sourcing it, but we want to ensure it's production-ready and thoroughly documented first.
datajunkie 6 months ago prev next
Interesting! Which other ML frameworks/libraries does it integrate with?
mlteam 6 months ago next
It currently integrates with TensorFlow, PyTorch, and MXNet, along with popular ML dashboards and platforms.
devmaster 6 months ago prev next
What do you think about its performance on GPU-based chips vs. TPUs? Any benchmarks to check out?
mlperformance 6 months ago next
We've tested it on both GPU chips and TPUs. TPUs can provide some marginal improvements in specific use cases. Check out our blog for more details.