34 points by mytechuser 5 months ago flag hide 14 comments
hnsurfer 5 months ago next
This looks great for those starting with machine learning! I really appreciate the simplicity and clear documentation. Thanks for sharing!
johncoder 5 months ago next
I agree @hnsurfer, the onboarding process is seamless and the examples cover a wide range of common tasks. Keep up the good work!
hnsurfer 5 months ago next
@johncoder the API is neat and everyone should give it a try. Will it cover deep learning models in the future?
mlbeast 5 months ago next
@hnsurfer great point, they do seem to have a good roadmap lined up. I think their decision to exclude it now is to ensure there is a strong core library first
annadata 5 months ago prev next
Just read the blog post and tried out the library. With this level of simplicity, I hope more people get into machine learning now!
hnsurfer 5 months ago next
@annadata that's the idea, to bring machine learning to a wider audience. I think the author can work on a faster or more optimized version if there's a demand for it.
mlbeast 5 months ago prev next
I've tried a few ML libraries and I must say, this one looks the most user-friendly. It's a nice project!
superdatafan 5 months ago next
It looks interesting, @mlbeast. Any suggestions on how to get started with predictions on a huge dataset without using all my memory?
mlbeast 5 months ago next
@superdatafan There are many techniques to handle large datasets. One approach is using iterative algorithms such as stochastic gradient descent/online learning that don't require memoryhungry operations like matrix inversion.
mypreviouswork 5 months ago prev next
Congratulations on sharing your creation on Hacker News! I still can't understand how your library is simple and user-friendly without a command-line interface to run everything, how would that work?
helpfulcoder 5 months ago next
@mypreviouswork It's a great question, I was thinking the same thing. I think they are using a web-based notebook environment to both explain and test your code with examples.
justanotheruser 5 months ago next
@helpfulcoder I think they also included a decorator for running via standard Python scripts, isn't that right?
helpfulcoder 5 months ago next
@justanotheruser I agree with you, that's a useful feature. With the variety of options available, users can choose what's right for their workflow.
mypreviouswork 5 months ago prev next
Ahh, I missed that. Well, it seems that it does get the job done. I'm still not sure I personally prefer a web-based solution for ML.