orting

> -I don’t know whether their policy differs from country to country

Would not be surprised if it did. I remember one rental in the Netherlands where the english text said free parking, but the dutch text said parking available. The problem was AirBnBs translation and they refunded my parking expenses.

A month is not a well defined unit. There are four different month units: 28, 29, 30, 31 days. If you do calculations respecting this there will be no problem.

Complex numbers and vectors are not a good analogy.

This is a good point. From the article "Once the students were selected, the researchers then administered the Major Field Test in Computer Science, an exam that was developed by the U.S. Educational Testing Service and is regularly updated. The exam was translated for the students in China and Russia."

It seems reasonable to expect a test written for a specific education system (either implicitly or explicitly) will be biased against others.

On the other hand, it is not that surprising that the US is a the top. It is a pretty big country with a long tradition for high quality education and a lot of funds going into CS departments. However, I will be surprised if this is not changing towards more dominance by India and China, because both countries are focusing a lot of resources in this area.

There is an interesting version of this with The Netherlands and Belgium The Netherlands: https://www.google.com/maps/place/Baarle-Nassau,+Holland/@51... Belgium: https://www.google.com/maps/place/Baarle-Hertog,+Belgien/@51...

I am not criticizing the study. I am highlighting that the conclusion that the paper arrives at i the correct one: That it warrants large scale studies.

This work is a prime candidate for being misrepresented as showing that this stem cell treatment is effective for age related health issues.

There are 30 participants in the phase 2 trial. There are two treatment groups (100M and 200M) with different dose and one placebo group. Each group has 10 participants.

None of the treatment groups showed adverse effects.

There is a difference between asking "Are there any adverse effects?" and "are there positive effects for parameter 1 to n"? If you ask the second kind of question and do not correct for multiple hypothesis testing, you will make many errors.

The small dose treatment group (100M) showed improvement in many parameters vs placebo, whereas the other (200M) showed improvement in fewer parameters vs placebo. Since no corrections where made for testing, this only tell us that there where no statistically significant adverse effects.

As I noted initially, I think it is interesting. Once we have seen the results of a couple of large studies, we can talk about the effects of this treatment.

The phase 2 study primarily looked at safety. There was no correction for multiple hypothesis testing on the efficacy endpoints. So it seems that the only conclusion that is warranted is that the study shows no adverse health effects and that "larger clinical trials are warranted to establish the efficacy of hMSCs in this multisystem disorder." as they state in the conclusion.

It is interesting if it works, but lets wait for the next phase before assuming it does.

Basically we can represent any signal as an infinite sum of sinusoids. If you know about Taylor expansion of a function, then you know that the first order term is the most important, then the second and so on. Same principle with the sinusoids. So if we remove the sinusoids with very high frequency we remove the terms with least information.

The first thing I noticed was the ringing, which is an artifact of low-pass filtering so it's a nice opportunity to go into problems with that kind of filtering. Other than that I think it was an ok teaser that gives an idea of how compression is done and what the trade-offs are.

It is from Bellman, Held and Karp [0] and has O*(2^n) complexity for both space and time. It is quite simple and based on the idea that

"Every subpath of a path of minimum distance is itself of minimum distance."

[0] https://en.wikipedia.org/wiki/Held%E2%80%93Karp_algorithm

As with a lot of "simple" math, the trick is to actually write it down and calculate it, because our intuition (at least for some of us) is often not the best when it comes to this kind of calculation.

In this case we write down the contingency table. Assuming that the test perfectly detects what we are looking for we find

True positives: 1

False positive: 5% of 1000 = 50

True negative: 949

False negative: 0

Chance of disease given positive results = 1/51 = 1.96%

Forcing a nation is what war is about, and it rarely works out very well.

Negotiation is about finding a solution to a problem that leaves all parties better off if they follow the solution than if they don't. It is not always easy and sometimes coercion, in the form of sactions within EU and UN, is used to make one party realize what is best for them - but this also tends to work out not very well.

Not forcing people to do what you want is often a more succesful way of getting what you need.

There is a lot of stuff the human eye cant see that is very interesting. One of the challenges in medical imaging is getting accurate labelling of images. For many labellings we see large inter- and intra-observer variability. We have both the problem that humans see something that is not interesting and miss something that is interesting.

I currently work on estimating emphysema extent in CT lung scans. Emphysema can be very diffuse and it is not possible to label individual pixels, so instead we try to learn the local emphysema pattern from a global label. Neural networks are interesting for this problem because the learn the features, but it is also a "problem" because the features might not make physically sense, which could make it hard to transfer the model and convince clinicians that they should use it.

> It is not a question of if but when

You can say that about almost anything, and the world is still full of factory workers.

As a PhD student in medical imaging, you must also know that getting fully automating segmentation methods to work to the standard required in the clinic is really hard. And once you solve it for one clinic you will likely not be able to transfer the trained model to another clinic, because scan parameters, patients and workflow are different.

But when we solve the segmentation task, I think most radiologist will clap their hands and move on.

Stating it as "thousands of patient images a day" is misleading. It would be the same as saying you inspected "thousands of parts each day". As the radiologist further down notes, CT scans contains many slices.

While computers don't get tired, they also have a really hard time solving stuff like annotation tasks automatically. One thing is getting a good enough general performance, another is to never make critical errors. I see a huge potential for ML approaches in health care, but primarily as an aid for the health care professionals and not as a full replacement.

I am curious. Could you highlight some of the technical problems with Firefox?

Yes it is obviously an error by the doctor and the pharmacists, no one is saying otherwise. The whole point of the article is to investigate how this error occurred. We live in a world where these errors happens all the time and the best thing we can do is learn from them and if the error analysis boils down to "incompetent doctor", then it is likely that the same error will occur again.

That is a good question, and one we should discuss more actively, because if it can go wrong it will go wrong. What happens when an over eager politician learns that "we can predict with X% accuracy if a person will do something bad next year"? I might be cynical, but I do not expect the result will be an increased interest in how society can help people before they do bad stuff. It would not surprise me if instead the argument would be that extensive surveillance is a great benefit to society because it can identify the bad guys with X% accuracy.

I think it is nicer if you just link to the arxiv page instead of directly to the pdf. Then I can read the abstract before downloading it.

http://arxiv.org/abs/1503.02401

I think the points are good, but I am not very happy about this statement

"When dealing with small amounts of data, it’s reasonable to try as many algorithms as possible and to pick the best one since the cost of experimentation is low. But as we hit “big data”, it pays off to analyze the data upfront and then design the modeling pipeline (pre-processing, modeling, optimization algorithm, evaluation, productionization) accordingly."

If done correctly, then I agree. But we have to be carefull about overfitting when we try out several models or make an initial analysis to determine which model to use. In this sense, choosing a model is no different from fitting the parameters of the model.

A specific critique raised in the press release from IARC is that the study has an

"emphasis on very rare cancers (e.g. osteosarcoma, medulloblastoma) that together make only a small contribution to the total cancer burden."

and that it

"excludes [...] common cancers for which incidence differs substantially between populations and over time."

So it sounds like the generalization hinted at in the abstract shows a bigger misunderstanding of statistics than any in the press release. Would be nice if the paper was not paywalled, so we could actually read it.

Yes I can do basic arithmetic. My point is that the ratio is easy to interpret, the program ran 6.36 times faster. The only reason I can think of for giving a percentage is to make it sound more impressive.

That is nice illustration of premature optimization. Instead of thinking "Lets parallelize", one should measure and find out what causes performance problems. Should one choose to go down the parallel path, it's a good idea to test if hyper threading degrades performance. In my experience it can be expensive use more than the physical cores.

edit: Another issue is that I really dont like when people present speedup in %. How should 540% speedup be interpreted? It makes more sense as a ratio, so we find sequential/parallel = 10067483333/1583584841 ~= 6.36. So the parallel version achieves a speedup factor of 6.36.

In general I think the distinction between physical, neurological and psychological is peculiar. It suggests that the mind is not physical but a metaphysical entity that can be be analysed separately from the body and vice versa.

My experience with both mental illness and physical disease is that a systemic perspective is important if we are to avoid fixating on finding a specific diagnosis and instead focus on quality of life.

I think that programmers reasons for choosing a specific language are as varied as the languages.

Personally I like programming in C++ because the typesystem and abstraction mechanisms allows me to write reasonably correct and concise code and at the same time performance is predictable. I like programming in Python because of the emphasis on readability, the "batteries included" standard library and the scripting capabilities.

Both languages have failings, as do all the other I have tried, but what matters most (for me) is availability (platform support, libraries etc), which is the reason I occasionally write php code.

I dont like this paper. It makes statements based on references to opinion style articles. F.ex

"in practice, run-time type errors in deployed programs are exceedingly rare [TW07]."

If we look at [TW07] they state that

"even very simple approaches to testing capture virtually all, if not all, the errors that a static type system would capture."

But provides no data or reference for that statement.

Another isue is that some references with data are based on small samples and possibly oudated:

"they [dynamc languages] lower development costs [Ous98]"

[Ous98] Compares time-to-implement and code-size for 8 different programs implemented in static and dynamic languages and shows that the dynamic languages are supperior. It is however not clear how much actual implementation is involved, so it may be the case that the difference is caused by diferences in available libraries at the time. In any case, the sample size is small and the article is old (1998) so it is not reasonable to make generalisations for programming in 2009 (or 20014).

[TW07] Laurence Tratt and Roel Wuyts. Dynamically typed languages. IEEE Software, 24(5):28–30, 2007.

[Ous98] John K. Ousterhout. Scripting: Higher-level programming for the 21st century. Computer, 31(3):23–30, 1998.

I remember visiting family in the states in 2008, we had our 9 month old child with us. Arriving in the US we got fast-tracked through security, allowed to bring formula for the baby with us and was generally treated very politely. In London we were asked to open all containers with baby food and taste it.

There are countless anecdotes illustrating how immigration rules and officers are ridiculous and racist, but not all of them are about USA. In many respects immigrants are treated worse in Europe that in the states.

I have no problem with the question "Have they considered purchasing power". I could also come up with a dozen possible issues they might not have taken into consideration. But I would never argue that they havent considered them before I actually checked. And even though it seems reasonable to say

"The relative purchasing power of the money in the test is significant within cultures, let alone across cultures."

I am not convinced that is correct. A quote from the paper describing the study [1] suggest that the amount of money is not crucial

"Indeed, in the UG, raising the stakes to quite high levels (e.g., three months’ income) does not substantially alter the basic results. In fact, at high stakes, proposers tend to offer a little more, and responders remain willing to reject offers that represent small fractions of the pie (e.g., 20%) even when the pie is large (e.g., $400 in the United States). Similarly, the results do not appear to be due to a lack of familiarity with the experimental context. Subjects often do not change their behavior in any systematic way when they participate in several replications of the identical experiment."

The point of the article is that it is the norm to conduct studies where participants are selected from the same non-representative sub-population, and that this methodology is heavily biased. Rejecting this idea, because you find a possible issue with one of the many studies it is based on, seems like a really bad idea.

[1] "Economic man" in cross-cultural perspective: Behavioral experiments in 15 small-scale societies (http://authors.library.caltech.edu/2278/1/HENbbs05.pdf)

So the idea that scientific studies should include people that are not upper middle-class westerns, is wrong because of the purchasing power of $100?

There is a link to the published paper in the article[1]. I have only skimmed it, but there are many different studies and findings. One I think is really interesting is this:

"Research on IQ using analytical tools from behavioral genetics has long shown that IQ is highly heritable, and not particularly influenced by shared family environment (Dickens & Flynn 2001, Flynn 2007). However, recent work using 7‐year old twins drawn from a wide range of socioeconomic statuses, shows that contributions of genetic variation and shared environment varies dramatically from low to high SES children (Turkheimer et al. 2003). For high SES children, where environmental variability is negligible, genetic differences account for 70‐80% of the variation, with shared environment contributing less than 10%. For low SES children, where there is far more variability in environmental contributions to intelligence, genetic differences account for 0‐10% of the variance, with shared environment contributing about 60%. This raises the specter that much of what we think we have learned from behavioral genetics may be misleading, as the data are disproportionately influenced by WEIRD people, and their children (Nisbett 2009)."

What they are arguing is that we have conducted science in a way where we have consistently sampled from a specific sub-population and used the results to generalise about the remaining sub-population. To me it sounds like they are on to something that could change many of the "givens" that are "known to be true". I recently saw a TED talk with Paul Johnson[2] where she discusses the problem that the sex of subjects in medical trials is often ignored leading to results that only holds for men or women.

[1] http://www2.psych.ubc.ca/~henrich/pdfs/Weird_People_BBS_fina...

[2] https://www.ted.com/talks/paula_johnson_his_and_hers_healthc...

You are arguing that we should use anecdotal evidence to guide public health policy. Please stop doing that and use a scientific approach instead.

I remember when I was a carpenter apprentice and asked for personal safety equipment and then being told that "wood dust never harmed anyone" or "I never use hearing protection and my hearing is just fine". I have encountered this attitude in many places, especially in small companies where not following safety rules can be seen directly in the profits. But you also see it in big projects, where the main contractor hires sub-contractors that hires sub-sub-contractors and no one has an overall responsibility.

My argument was badly put.

What is the benefit of using the same language on the client and server? That the developer only needs to know one language?

The server-side code you write has to interact both with the client and the underlying OS. So you should consider how language X interacts with both client and OS, in addition to the intrinsic qualities of the language.

If we adopt a modular approach to software development, then I don't see any universal benefit to using the same language for all modules.

js might be a good server-side language, but choosing it because it works on the client is the wrong reason.