jph00

49 ms·

601.
by jph00 8y ago
The results on this paper are 4x worse than the state of the art of a couple of years ago. Personally I think the paper oversells their results. Creating a much worse model by making it much simpler doesn't seem that interesting, to me
602.
Fast.ai Embracing Swift for Deep Learning(fast.ai)
7 points by jph00 8y ago|0 comments
603.
by jph00 8y ago
The authors made various claims or implications not backed up by their experiment. However, it absolutely does not answer the question that is asked in the title of the paper, and the process they use is incapable of answering that question
604.
by jph00 8y ago
Also AWS has p3dn.24xlarge with 100Gbps :)
605.
by jph00 8y ago
BTW, if you're interested in a code-first walkthru of NMF at a higher level, we cover it in this Jupyter notebook and video: https://nbviewer.jupyter.org/github/fastai/numerical-linear-...
606.
by jph00 8y ago
In other fields such as infosec, responsible disclosure is a standard approach. You don't just throw a zero-day out there because you can. Whilst the norms for AI research needn't be identical, they should at least be informed by
607.
by jph00 8y ago
Nope I didn't say that - and I haven't been involved with fm for years.
608.
by jph00 8y ago
FYI, Google just completed adding all the pre-requisites for this course to their Colab service, which provides a GPU-powered Jupyter Notebook for free. There are no usage restrictions, other than that each session must be no longer than 12
609.
by jph00 8y ago
It's free and there are no ads, and you can run it for free on colab or GCP. I have no idea about where your sagemaker comment comes from.
610.
by jph00 8y ago
About 14 hours of lessons per part for fast.ai. A lot of Udacity courses I've seen are shorter than that. I certainly wouldn't call it "brief"!
611.
by jph00 8y ago
The changes were just during the development phase. Should be quite settled now.
612.
by jph00 8y ago
Whilst it's absolutely not required knowledge for most deep learning, if you are interested in more fully understanding the chain rule, my friend Terence Parr (the ANTLR guy) and I wrote a fairly thorough tutorial for matrix calculus
613.
by jph00 8y ago
Many thanks for letting me know - I've fixed that link now. FYI the ML course is now at: http://course18.fast.ai/ml
614.
by jph00 8y ago
There's a lot of Python learning resources here: https://forums.fast.ai/t/recommended-python-learning-resourc... My main recommendation is to try to do as many things as you can with code, instead of manually or w
615.
by jph00 8y ago
Object detection is in part 2. You can use the 2018 part 2 for now. We had hoped to move it to part 1 this year, but couldn't figure out a way to make it easy enough yet.
616.
by jph00 8y ago
I discussed this a bit here: https://www.fast.ai/2017/09/08/introducing-pytorch-for-fasta... Also some info in the fastai release post: https://www.fast.ai/2018/10/02/fastai-ai&
617.
by jph00 8y ago
Yes it'll be a bit more work for you, but it should be OK. I provide links to more math background where required in the course.
618.
by jph00 8y ago
For sure. It's a giant step over last year's course.
619.
by jph00 8y ago
Many thanks. I'm really happy with how it turned out. I hope you all like it! :)
620.
Practical Deep Learning for Coders 2019(fast.ai)
556 points by jph00 8y ago|75 comments
621.
by jph00 8y ago
fastmail lets you write complete scripts for filtering using the Sieve programming language. https://www.fastmail.com/help/technical/sieve.html
622.
Pytorch tutorial: “What is torch.nn, really?”(pytorch.org)
3 points by jph00 8y ago|0 comments
623.
by jph00 8y ago
I'm pretty happy with how this new project turned out - I was able to wrap C++11's random distribution library in a way to feels like idiomatic Swift code. Although Swift doesn't support C++ directly, creating a C API as show
624.
C++11, random distributions, and Swift(fast.ai)
2 points by jph00 8y ago|1 comments
625.
by jph00 8y ago
Oh BTW there's already something for Swift inspired by python's Requests lib: https://github.com/JustHTTP/Just/blob/master/README.md
626.
by jph00 8y ago
That should be doable with a library, right? Is there any need for language syntax for this?
627.
by jph00 8y ago
Swift's APIs tend on the verbose side, but they can be wrapped up into much more concise versions. Since few people have been using Swift in this way, those wrappers haven't really appeared yet (and because Swift packaging is stil
628.
111 Donald Knuth Lectures Uploaded by Stanford(youtube.com)
27 points by jph00 8y ago|1 comments
629.
by jph00 8y ago
"If the array’s Element type is a struct or enumeration, Array and ContiguousArray should have similar efficiency."
630.
by jph00 8y ago
Very interesting perspective - many thanks for sharing.

More ›