ihodes

15 ms·

91.
by ihodes 12y ago
Which project? CTC screening is fascinating. Could you please point me to some good introductory papers/reading?
92.
by ihodes 12y ago
That sounds similar to a recurrent neural net? Backprop definitely works on those with some modifications. I just asked some guys at CMU/Google, they suggested looking more at SGD or some other MCMC methods that may have included the w
93.
by ihodes 12y ago
I hadn't heard of PSO. Looks pretty interesting. What's your use case that you cannot use stochastic gradient descent?
94.
by ihodes 12y ago
You may be looking for stochastic gradient descent (SGD). Training neural nets is very parallelizable using such techniques, and can even be done asynchronously (avoiding having to regularly sync different nodes in the cluster) using system
95.
by ihodes 12y ago
The error rate of 15% (I actually believe it's more like 30% in practice, which is 2x PacBio's) is actually excellent and comparable to PacBio's in that the error is primarily randomly distributed, and not systematic like cur
96.
by ihodes 12y ago
This book[1] is the best resource for a "medium-level" overview I've come across. Once you get beyond that, you're entering textbook and then paper territory. [1]: http://www.amazon.com/Immune-System-Incl
97.
by ihodes 12y ago
Not at all a stupid question. I don't even have a great answer for you. My understanding is that, at least for some cases, you can target e.g. radiation enough to shrink the tumor enough for these therapeutic vaccines to be effective.
98.
by ihodes 12y ago
Maybe not faster than I'd think :) I work in the field as well. Maybe even with your father!
99.
by ihodes 12y ago
It's one way we're curing cancer. Other promising avenues include personalized cancer vaccines which essentially recruit your immune system to attack the tumor[1], or other viral vectors such as the recent Mayo success[2] using th
100.
by ihodes 12y ago
Good to know, and that makes sense to me. Thank you for adding a warning, as well. ack turns out to be much faster than grep on these large files, FWIW. Thanks for making this superb tool :)
101.
by ihodes 12y ago
Big problem with ag is that it appears to be broken on even moderately larger files where ack works just fine: https://github.com/ggreer/the_silver_searcher/issues/384
102.
by ihodes 12y ago
That turns out to be the focus of a lot of research right now[1][2]. The computational aspect of that has a lot to do with 1. faster, more accurate analysis of genetic code 2. faster, more accurate simulation of interactions
103.
by ihodes 12y ago
The actual paper (parts are accessible & interesting): http://arxiv.org/pdf/1404.3840v1.pdf
104.
by ihodes 13y ago
Probably the four most important things you can do to change your odds of making it past 80 are: 1. Not smoking. 2. Eating healthily (fiber, vitamins, low sugar; this is a nascent field). 3. Exercising regularly. 4. Wearin
105.
A rich REPL for Clojure in the notebook style.(github.com)
6 points by ihodes 13y ago|2 comments
106.
Yesql - Clojure SQL queries rethought(github.com)
2 points by ihodes 13y ago|0 comments
107.
by ihodes 13y ago
This article always struck me as being composed of two distinctly different opinions. One, Aaron's assertion that employees at Google are being infantilized. Two, that Google's mission and their overall morale and sense of purpose
108.
by ihodes 13y ago
Thanks! Mind sharing your email (or pinging me)? I'd love to hear about what you found.
109.
by ihodes 13y ago
Ping me at isaachodes at gmail.com if you have specific questions (and I've love to hear about your use-case as well). Encrypted data at rest and transit are upcoming (SSL is very soon, encrypted by the secret key (as an optional measu
110.
by ihodes 13y ago
Sorry about that! Yeah, that's something a few people use to track & graph things (either y/n questions, 0-n question, or numeric questions) via scheduled emails.
111.
by ihodes 13y ago
Yes, that is in progress.
112.
by ihodes 13y ago
Probably wouldn't for that, but there's quite a bit more going on there, not the least of which is XPath and JSONPath querying. I'll be posting more on this later, but this is just a part of another project as well. But since
113.
Show HN: I Made a Web API for Textual Data(skr.io)
27 points by ihodes 13y ago|13 comments
114.
by ihodes 13y ago
I agree completely. DEFINITELY something I'm working on; the syntax is not ideal. That's not even quite how it looks anymore (and even that isn't where I want it to be yet), though I didn't have time to switch it out for
115.
by ihodes 13y ago
I debated using the word "build" in lieu of "implement". Designing building logic is an ongoing process; implementing it is what we're seeking to make easier.
116.
by ihodes 13y ago
Absolutely agree; the business logic is the tough part. That's why it should be easy it iterate and verify it, which is the goal of what we're doing here. The sooner you can get a product out, the sooner you can spot your edge cas
117.
by ihodes 13y ago
They are going after a similar market, but the key difference (and it's very important in terms of expressiveness) is that they have a workflow engine, but this is a state machine at its core.
118.
by ihodes 13y ago
Great! I will be in touch very soon, looking forward to chatting.
119.
by ihodes 13y ago
Right now I have Twilio, Urban Airship, Mailgun, and also arbitrary webhooks. More to come soon. It's actually a technology of Marion Tech, copperthoughts is just the silly name of my blog :)
120.
by ihodes 13y ago
First I'll say that I am nowhere near an expert on working with workflow engines. If you wouldn't mind, I'd love to chat with you about your experiences with them. The main difference is it's an extended state machine, n

More ›