jph00

84 ms·

841.
by jph00 9y ago
That's my view too. Freezing more layers at the start can also help a lot with fragile training.
842.
by jph00 9y ago
Changing to a different fixed color space is unlikely to help. However adding a learnable color space transformation is a great (and rarely known) trick. Both are discussed and compared in this terrific paper: https://arxiv.org
843.
by jph00 9y ago
It's not so much that there's just one or two specific results, but rather that there's just far more researchers working on this now, and quite a few are working on more theoretical stuff. Sometimes, that theory results in r
844.
by jph00 9y ago
Part 2 recording is finished! Just need to get it online :)
845.
by jph00 9y ago
I've really enjoyed talking to Jacques about his lego project over the last few days, and I hope that it will lead to some additional learning materials on course.fast.ai (which he kindly links to in his article) as a result. It's
846.
by jph00 9y ago
Most research is done on tiny images (224*224) and uses therefore large batch sizes (128). So there's a lot of compute to do. Even more so if you're using RNNs with large backprop windows. There are quite a few examples of folks k
847.
by jph00 10y ago
> I have a feeling people are going to reply without reading this through and assume the author poses a Deep Learning vs Classic CV sort of argument, in a deep-learning-is-overrated sort of way. Whereas it seems to me he is merely sayin
848.
by jph00 10y ago
Great to see this picked up on HN - it's really interesting work. Terence is my colleague at USF (in fact he's the guy that got me to go there!) and it was great to see the journey from "could this possibly work" to &quo
849.
Why securing investment can be a challenge for female entrepreneurs(theage.com.au)
2 points by jph00 10y ago|1 comments
850.
Why We Need to Democratize Artificial Intelligence Education(forbes.com)
7 points by jph00 10y ago|0 comments
851.
by jph00 10y ago
I've seen dozens of examples of people with less than a year of ML experience get great outcomes from deep learning. It's no longer true that this is such an exclusive field.
852.
by jph00 10y ago
To answer your question about the http://course.fast.ai MOOC: the course provides an AMI that has everything you need pre-installed. Just follow the directions in the setup video, which includes a script that sets up the instanc
853.
by jph00 10y ago
No, the person I know who is in the last round is an economics major with no Masters/PhD. She has a very impressive background in industry. Although a lot of folks in the course do indeed have graduate degrees in other fields (includin
854.
by jph00 10y ago
To give a more exact estimate - if you don't use spot instances ($0.20/hour) it'll cost $0.90/hour, and if you do the suggested 70 hours of work for the course, that's $63. And then there's around $6/month
855.
by jph00 10y ago
One forum user put together a <$700 complete PC with an adequate GPU. You definitely don't need >1 GPU. Or you can use spot instances, as a sibling comment mentions - about $0.20/hour generally.
856.
by jph00 10y ago
If you're the kind of person that's interested in taking up this challenge, but you currently have the coding skills without the deep learning skills, we built something that can equip you with most of the current best practices i
857.
by jph00 10y ago
Regarding AWS, one participant has created a system that lets you use spot instances for the course. It's published on the forum. Great way to save $$$ (400% or more...)
858.
by jph00 10y ago
Sure - but on pytorch they suffer the kernel launch overhead each time through the loop, whereas on tensorflow and theano they do not. Which really impacts the kinds of algorithms that work well on each platform. Does that seem like a reaso
859.
by jph00 10y ago
Generally for structured data (i.e. each column represents a distinct type of information, such as 'revenue' or 'color') you'll want random forest or GBM. For unstructured data, where you'll need lots of comple
860.
by jph00 10y ago
I'm intrigued!... The kernel calling overhead and lack of any GPU while/scan/map/etc for Pytorch seems like a limitation, but I guess on 2nd thoughts you can still do all the keras fit/predict stuff and auto-connect
861.
by jph00 10y ago
If you try the course, be sure to make use of the forums for it too: http://forums.fast.ai . As you'll see, they're extremely active and helpful for all deep learning students (and all practitioners in general). Disc
862.
by jph00 10y ago
It's trivial with 'max' merge param: https://github.com/fchollet/keras/pull/3128
863.
by jph00 10y ago
Yay, making deep learning uncool again! ;) ( http://www.fast.ai/about/ ) But seriously - does it even make use to have a "define by run" dynamic framework as a backend? It seems to me that keras is particularly
864.
by jph00 10y ago
Why does the article say that Ben Hamner was involved in the founding in 2010? He joined years later. Some basic fact checking would be nice, even in tech articles... (Ben has been a great contributor, mind you.)
865.
by jph00 10y ago
> From the beginning of this event until 11:37AM PST, we were unable to update the individual services’ status on the AWS Service Health Dashboard (SHD) because of a dependency the SHD administration console has on Amazon S3. Ensuring
866.
Deep Learning for Coders – Full notes and transcripts now available(fast.ai)
9 points by jph00 10y ago|0 comments
867.
by jph00 10y ago
This is good advice - in fact at http://course.fast.ai our main teaching focus is transfer learning
868.
by jph00 10y ago
Try http://course.fast.ai . We built it because we observed just the issue you have in most teaching materials for deep learning. We do cover all of the details eventually, but only when you're ready, and in a code first wa
869.
by jph00 10y ago
The Keras functional API are just simple operations on tensors too.
870.
by jph00 10y ago
> This doesn't feel fair to say, just like it wouldn't if you replaced "Tensorflow" with "C" and "Keras" with "Python". They operate on fundamentally different levels and provide a simila

More ›