5 ms·
Operations, machine learning and premature babies
- streptomycin 14y agothe telltale signal wasn’t spikes or irregularities, but the opposite. ... Their heart rate was too normal; it didn’t change throughout the day as much as it should. That observation strikes me as revolutionary. It was revolutionary... like 30 years ago when people first noticed that decreased heart rate variability is a marker for bad events, obviously without "a machine learning system that can sift through gigabytes of data". Not that I disagree with the rest of the article, but the anecdotal example is poor, and it just wouldn't be Hacker News if somebody didn't pedantically point that out. Also, FWIW, in the field of predicting bad things in premature babies based on patterns in heart beats, these guys are a bit ahead of IBM, having recently completed a multicenter randomized controlled trial: http://www.heroscore.com/ http://www.heroscore.com/
- ars 14y agoI came here to say the exact same thing. During fetal monitoring the #1 thing they are looking for is heart rate variability. If the heart rare is constant that is very dangerous sign for the fetus. The reason is simple: If the fetus has the oxygen it needs then the heart rate will constantly change, low during rest and high during activity. But if there is a shortage of oxygen the heart will run at the highest it can manage with the oxygen available. This is also true for adults: The military figured out that the best soldiers for special forces training had a constant heart rate - their hart rate doesn't spike during times of stress, making them very good as resisting interrogation. But it has a downside: They don't live as long as people with constantly changing heart rates.
- streptomycin 14y agoThe reason is simple In your example it may be (I don't know much about fetuses), but in general it isn't. The reasons for changes in heart rate variability (HRV) are generally unknown, although there are various hypotheses and theories. If you read the papers by the company I linked to above that does HRV monitoring for neonatal sepsis, they freely admit that they don't know the mechanism behind the signal they detect. The military figured out that the best soldiers for special forces training had a constant heart rate Interesting. Do you have any links about that?
- ars 14y agoThis makes sense for machines - if your server is always running at a constant rate then it's probably maxed out. If the rate changes, and has no "clipping" (flat areas at the top of the peaks) then it has the headroom it needs.
- pjscott 14y agoI was hoping for more than speculation. Sadly, this article didn't contain it, but a bit of googling turned up this paper: http://static.usenix.org/events/sysml08/tech/full_papers/xu/xu.pdf http://static.usenix.org/events/sysml08/tech/full_papers/xu/... These guys wrote some software that looks through the logs from a Hadoop cluster and identifies patterns of anomalies. If a cluster of a few exceptions tend to be spammed in the logs at the same time, it identifies this as a single problem. If a scary-looking "error message" is actually normal, it says that this is normal. As someone who has spent too much time doing that kind of thing manually, the idea of automating this sounds very appealing.