_pastel

PicnicHealth (YC S14) | Design, Front-End, ML, Full-Stack | SF | ONSITE, FULL-TIME, VISA | https://team.picnichealth.com/jobs?src=hn Picnic’s mission is to structure the world’s medical data to make it useful. We work directly with patients to collect, digitize, and manage their complete medical records, giving them with control over their care. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building the data sets that power some of today’s most cutting edge medical research. Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Our product has literally saved some of our users' lives by unearthing issues in their medical records. We're growing rapidly — on pace for 20x this year. Our team is smart, hard-working, and passionate about fixing healthcare.

Fun game: read through the summary of each experiment (psyarxiv.com/9654g) and try to guess whether it replicated or not before they tell you. I got 20/28.

Some of them were obvious (#3 - no, using a bad font does not make people answer questions more carefully; #22 - yes, people attribute more responsibility to adults than to babies). But there were some big surprises for me:

#6 - Participants read a hypothetical water conservation dilemma, then rate how much they trust the people in the scenario to conserve water. Sometimes the people are called "consumers" and sometimes "individuals". Does it make a difference? Apparently yes. When they're called "consumers", you expect them to conserve significantly less water than when they're "individuals".

#24 - if you ask "How satisfied are you with your relationship?" and then "How satisfied are you with your life-as-a-whole?", are the results more highly correlated than if you ask those questions in the opposite order? I thought for sure they would be - relationship happiness is one of several contributing factors to your overall happiness, so if it's fresh on your mind, you'll weigh it more heavily when you answer the life-as-a-whole question. Apparently not.

This article has by far my favorite explanation of GANs for image generation:

  To generate the image, the algorithm compared its own work to those in the data set until it could not tell them apart.

Amazingly succinct, lay-person friendly, and still pretty accurate.

I have a pet implementation of the frequency map that I'm overly fond of, for ascii strings:

  (1) keep an array of length 127 that you re-use and set to 0 between calls
  (2) for each character in the first string, increment the array at the character's index
  (3) for each character in the second string, decrement the array at the character's index

If you end up with all 0s, it's an anagram.

This is a perfect explanation of why OCR of medical records is so difficult.

As an engineer working on this, it's frustrating that the computer vision research community treats document OCR as a solved problem. New papers and competitions are exclusively in fancy settings, like slanted text outdoors on a busy street, or handwritten Urdu with only a small set of training labels.

Meanwhile our medical data is locked up in a scan of a fax of a scan of a fax, with a mixture of printed text and handwritten notes, with table lines crossing the text. Paid services like Google's Cloud Vision API struggle at best, and the best available open source software (tesseract) isn't even close.

PicnicHealth (YC S14) | Design, Front-End, ML, Full-Stack | SF | ONSITE, FULL-TIME, VISA | https://team.picnichealth.com/jobs?src=hn

Picnic’s mission is to structure the world’s medical data to make it useful. We work directly with patients to collect, digitize, and manage their complete medical records, giving them with control over their care. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building the data sets that power some of today’s most cutting edge medical research. Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Our product has literally saved some of our users' lives by unearthing issues in their medical records. We're growing rapidly — on pace for 20x this year. Our team is smart, hard-working, and passionate about fixing healthcare.

PicnicHealth (YC S14) | Design, Front-End, ML, Full-Stack | SF | ONSITE, FULL-TIME, VISA | https://team.picnichealth.com/jobs?src=hn

Picnic’s mission is to structure the world’s medical data to make it useful. We work directly with patients to collect, digitize, and manage their complete medical records, giving them with control over their care. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building the data sets that power some of today’s most cutting edge medical research. Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Our product has literally saved some of our users' lives by unearthing issues in their medical records. We're growing rapidly — on pace for 20x this year. Our team is smart, hard-working, and passionate about fixing healthcare.

Picnic | Front-End, Full-Stack, ML Engineering | SF | Full Time, Onsite

Picnic's mission is to structure the world's medical data to make it useful. We work directly with patients to collect, digitize, and manage their complete medical records, giving them with control over their care through a personal health timeline. We do the dirty work of fetching and parsing any record from any doctor, anywhere in the US. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building data sets that power cutting edge medical research. We’re going through a period of extreme growth — on track for 20x over the next year, coming off our recent $1M finish in Google Cloud’s Machine Learning Startup Competition.

Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Learn more at https://team.picnichealth.com/jobs

It's an acronym: Scale Normalization for Image Pyramids with Efficient Resampling.

PicnicHealth (YC S14) | Design, Front-End, ML, Full-Stack | SF | ONSITE, FULL-TIME, VISA | https://team.picnichealth.com/jobs?src=hn

Picnic’s mission is to structure the world’s medical data to make it useful. We work directly with patients to collect, digitize, and manage their complete medical records, giving them with control over their care. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building the data sets that power some of today’s most cutting edge medical research. Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Our product has literally saved some of our users' lives by unearthing issues in their medical records. We're growing rapidly — on pace for 20x this year. Our team is smart, hard-working, and passionate about fixing healthcare.

> Indeed, statistical significance testing revealed no difference between students of any tier when it came to interview performance.

Ah, the classic confusion between not finding enough evidence to prove there is a difference, and finding enough evidence to prove there is no difference.

The juniors from elite schools in particular have fewer 1s and 2s and more 3s and 4s than the other juniors. Really, you found statistical evidence that they aren't from different distributions? I'd love to see it.

Picnic | Front-End, Full-Stack, ML Engineering | SF | Full Time, Onsite

Picnic’s mission is to structure the world’s medical data to make it useful. We work directly with patients to collect, digitize,and manage their complete medical records, giving them with control over their care through a personal health timeline. We do the dirty work of fetching and parsing any record from any doctor, anywhere in the US. Beyond serving patients directly, we partner with biotech, genomics, and pharma companies who sponsor PicnicHealth accounts for research volunteers. Through this work we’re building data sets that power cutting edge medical research. We’re going through a period of extreme growth — on track for 20x over the next year, coming off our recent $1M finish in Google Cloud’s Machine Learning Startup Competition.

Our stack is React/Redux, Node, Python, Keras, PostgreSQL, Kubernetes.

Learn more at https://team.picnichealth.com/jobs

It's more like the opposite problem.

Psychology researchers try to justify deep, compelling theories of human behavior with small experiments that are hard to reproduce.

Deep Learning experiments are comparatively easy to run and reproduce because the field has both a common set of performance benchmarks and a large community of people implementing cutting-edge ideas in common frameworks. The harder part is building useful theory on top of those experiments.

I'm curious: have you had those experiences everywhere in Thailand, or only at the most famous tourist sites?

I've read this about Chinese tourists for years. But all the Mainlander tourists I've spotted in Taiwan and Hong Kong have been normally behaved, which has always been mildly disappointing. Until I finally visited some of the more touristy Thailand temples and saw the behavior you describe.

My theory is that the sort of Chinese tourist who fits this stereotype will almost exclusively follow a guided itinerary with a big tour group.

Can anyone give me some intuition into why 77k people are employed by the Social Security Administration? [1] That's bigger than Google, and I didn't think their scope was that wide.

[1] http://usafacts.org/government-finances/employment?compariso... under "Secure the Blessings of Liberty to Ourselves and Our Posterity". The combined count is "n/a" - maybe that's because the State & Local count has some overlap with the Federal count? So I took the maximum of the two.

Note also their x axis is on a log scale, which makes the edges linear instead of curved. E.g.:

plot(a + b, log(a/b), xlim=c(1, 2), log="x")

It really doesn't matter - at least, not for the statistical error the parent is talking about. The effect isn't related to whether we are sampling from a larger population of programmers.

Suppose there were no difference between the usage of each language, and people just program on the weekends vs weekdays with some probability independent of language. Then, if a language has lots of users, it will likely have close to the average weekend/weekday proportion. The fewer users the language has, the more likely that it has an uneven weekend/weekday proportion just by chance. And if you plot the weekend/weekday proportions vs. the number of users, you expect a funnel shape just like the one in the article.

Therefore, the plot in the article - by itself - provides no evidence that there is any difference between the usage of different programming languages.

I wonder if some of your difficulties are related to the particular structure of Reversi.

The obvious heuristic in the endgame - maximize the number of your pieces - is actually reversed in the opening and midgame, where you want to minimize your number of pieces (other considerations like edge play and parity being equal). So it's possible your NN managed to do worse than random because it learned an endgame heuristic and generalized it improperly.

If so, you may want to consider explicitly wiring in the turn counter as an input.