madisonmay

9 ms·

31.
Finetune: Scikit-learn style model finetuning for NLP(github.com)
2 points by madisonmay 8y ago|0 comments
32.
by madisonmay 8y ago
The critical thing to note is that although the initial language model takes a month to train on 8 GPUs, fine-tuning the language model for a specific task is much much cheaper (a few hours on a single GPU).
33.
by madisonmay 8y ago
Initially embeddings of OOV terms are initialized to mean embeddings, and then the language model fine-tuning step allows the model to adjust these values. See [18] here: https://github.com/fastai/fastai/blob/
34.
by madisonmay 8y ago
Yes absolutely want to do comparisons to ELMo. Perhaps even more interesting than comparison would be modifications to ULMFit to incorporate good ideas from the AllenNLP ELMo paper. The learned weighting of representation layers seems lik
35.
by madisonmay 9y ago
In Knuth's words: "Premature optimization is the root of all evil". Although I guess in this context, you're emphasizing that it's more valuable to solve the right problem poorly than it is to solve the wrong proble
36.
by madisonmay 9y ago
Like most language, meaning is dependent on context. But in general it seems to be good practice to steer clear from pronouns associated with a particular gender when there are other perfectly good terms to be used (In this context, "
37.
by madisonmay 9y ago
Nope, not just fluff. New ML model architectures get too much hype, while it's relatively simple tools like this that actually make the difference in whether or not ML can applied to industry problems. The low hanging fruit in the ML
38.
by madisonmay 10y ago
How close are you to a linux release?
39.
Data Science Deployments with Docker(indico.io)
1 points by madisonmay 10y ago|0 comments
40.
Machine Learning: Models with Learned Parameters(indico.io)
314 points by madisonmay 10y ago|31 comments
41.
by madisonmay 10y ago
Lead Frontend Developer @ indico data solutions =============================================== Experience: 3+ years professional experience Location: Boston, MA ----------- Overview ----------- indico is a machine learning company turning
42.
by madisonmay 10y ago
Haha, having mucked around with similar mechanisms for dealing with concurrency via decorators I have to agree with you that this is likely to cause hard to debug behavior (especially since direct modification of the AST is going on behind
43.
by madisonmay 10y ago
My best guess is that this is to circumvent issues with naming encountered by using decorated functions in conjunction with multiprocessing pools. Typically you would run into serialization errors.
44.
by madisonmay 11y ago
So in other words, O(n) where n is the number of elements in the trie? That's rough.
45.
by madisonmay 11y ago
I've used this thing to reduce the size of scikit-learn's TfidfVectorizer, which stores a giant dictionary of words --> occurrence counts, and have seen close to 10X memory reductions. Great project!
46.
by madisonmay 11y ago
Honestly this addition was so seamless I thought it had always been that way. Definitely liked a few things on twitter before I realized what I was doing.
47.
by madisonmay 11y ago
Fellow BKB Somerville member here! Whaddup!
48.
by madisonmay 11y ago
Sort of related (but the other way around): https://www.blockspring.com/
49.
by madisonmay 11y ago
Always great to see content from Kyunghyun Cho. Keep up the great work, and have fun at NYU this fall!
50.
Data Science and Disney Movies(indico.io)
2 points by madisonmay 11y ago|0 comments
51.
Why You Shouldn't Build Your Own Machine Learning Models(indico.io)
3 points by madisonmay 11y ago|0 comments
52.
Show HN: CommonRegex(github.com)
1 points by madisonmay 11y ago|0 comments
53.
by madisonmay 11y ago
Yup, got it working on 14.04 here. My guess is you're having issues setting up the scientific python stack. You might want to try the anaconda distribution ( https://store.continuum.io/cshop/anaconda/ ) if you
54.
by madisonmay 11y ago
You also might want to check out Whetlab: https://www.whetlab.com/ .
55.
by madisonmay 12y ago
Always excellent content from Christopher Olah. So few people do such a clear job of explaining these concepts -- this is technical writing done right.
56.
by madisonmay 12y ago
I think part of it is also the big fish, little pond --> little fish, big pond phenomenon. When high school teaches you to define yourself as intelligent, academic, and driven and then you're placed in an environment where everyone
57.
by madisonmay 12y ago
If you're interested on a radical new take on engineering education, take a look at Olin College of Engineering. It's not perfect, but it's a move in the right direction -- Olin has traded strict competition for teamwork, te
58.
by madisonmay 12y ago
I disagree with this sentiment. Suicide rates this high are not acceptable side effects of simply studying engineering. Yes, it's typically a tough program, but action needs to be taken to address a deeper issue. It's the intense
59.
by madisonmay 12y ago
thetalkingmachines.com
60.
by madisonmay 12y ago
Any plans to come to Boston?

More ›