HNHacker News
TopNewBestAskShowJobs

wjn0

348 karma · joined October 5, 2015

submissionscomments
wjn0··on SymPy makes math fun again
WolframAlpha + student licence is comparable for some tasks, I've found (hard integrals from applied probability theory).
wjn0··on AI Models Predict Breast Cancer with Radiologist-Level Accuracy
This is a crucial point. There's a good chance these are problems that could be solved with more data. However, I think philosophically the challenge of integrating these models in a modern healthcare system is more fundamentally related to explainability. A crucial part of a radiologist's job is the "why?" - why did you make this diagnosis, why did you flag this patient, etc. While there are models (especially in image processing) that tackle this problem, I'm not personally aware of them being used in a clinical setting. It's difficult from not only a machine learning perspective, but also in terms of HCI/UX.
wjn0··on Another Book on Data Science – Learn R and Python in Parallel
This is really interesting! I had always just claimed that Python was faster (see the benchmark I linked above [1]) based on personal experience. I wonder if this internal implementation has something to do with it...

[1] https://julialang.org/benchmarks/

wjn0··on Another Book on Data Science – Learn R and Python in Parallel
Speed. [1]

See the yellow benchmark (matrix multiply). I suspect it's memory-related.

[1] https://julialang.org/benchmarks/

wjn0··on Another Book on Data Science – Learn R and Python in Parallel
> > R is far superior in its the quality of the packages, their documentation, their ability to behave predictably on a given data type.

> I not only disagree but I think that the exact opposite is true for each one of these points. But if things are working well in our shop, I'm not going to try to convince you otherwise.

I partially agree with you here. I'm extremely careful about what non-standard packages I use in R. Code quality varies wildly outside of these, likewise for documentation. But outside of neural networks, I've never found a package in Python that I felt better about in terms of code quality or documentation than its equivalent in R.

wjn0··on Another Book on Data Science – Learn R and Python in Parallel
> Not an advantage if you ask me - exactly because data.frame is built in, people have been building their own versions (tibble, data.table) instead of improving it. That's how R ended up with 3 different structures that are similar but have inconsistent apis and behaviour.

I've been fortunate to only work on projects that use built-in data frames, never encountered tibble or data.table in the wild.

> there is no one go-to library for ML in R (like sklearn in Python) and each package has it's own strange interface and implementation.

I still disagree here - one example being the unified interface for generalized linear models. Also, the vast majority of classifiers (RF, SVM, etc.) have similar or identical interfaces. Also, the unified `predict` interface as well. Granted, `sklearn` does have a consistent API as well.

That said, some of this is just a personal preference for the vaguely functional interface in R. The object-orientedness in Python feels a little forced for some tasks in `sklearn`.

wjn0··on Another Book on Data Science – Learn R and Python in Parallel
I agree for the most part, but R does have a few things beyond the tidyverse: built-in dataframe support, lots of domain-specific packages, more consistent interfaces for basic statistics and machine learning models, etc. Python is definitely better for matrices (because of NumPy) and anything involving custom gradient descent methods (because of TensorFlow).

I think 90% of data science content is for beginners because anything more advanced isn't best described as data science. As soon as you get beyond the initial stages of data analysis (cleaning and processing data), you're doing something best described as some other word (statistics, machine learning, etc.) - although, granted, there isn't much content in these areas if you don't know _exactly_ what you're looking for.

wjn0··on Carnival Cruises emits ten times more sulphur oxide than all of Europe’s cars
Surely an entity such as the EU could go so far as to ban the sale of bunker oil to passenger cruise lines with offices in the EU, or something to that effect?
wjn0··on Why are CVS receipts so long? (2018)
It almost certainly is. Likewise for the dye.
wjn0··on Fun with Snapchat's Gender Swapping Filter
From a machine learning standpoint, I would guess that this is achieved with a partially supervised approach. The idea would be to embed an entire face dataset from a number of different poses to a latent space by some method (could be as simple as PCA, or as complex as an arbitrary neural network). An arbitrary classification method (using a partial labelling of faces by gender) yields a hyperplane which optimally separates the labelled points. Then, something like a simple reflection across this hyperplane yields another point in the latent space, and the generator yields a corresponding face of the opposite gender.

Anyone know if this has been implemented for a task like this?

wjn0··on Two L.A. universities quarantine hundreds students and staff in measles outbreak
True, it's not a perfect system. But that would likely not be the case for a vaccination database, anyway.
wjn0··on Two L.A. universities quarantine hundreds students and staff in measles outbreak
It seems like vaccination info is not kept at the state or federal level. This information is no doubt extremely dangerous - imagine if a malicious actor could get ahold of the "non-vaccinated list." But right now, it's stored at high schools, universities, and workplaces around the country with standard HIPAA-grade security. As an example of what kind of government oversight I'm talking about: in Canada, there are reporting requirements for certain STD infections (including treatable ones: think chlamydia, or gonorrhea). I think some US states have reporting requirements as well.

In cases like this, where there's already an active quarantine, if the CDC or the state kept track of non-vaccinated people, they could be quarantined as well. The idea being, even if they're not showing symptoms, they're at high risk for carrying the disease. This seems like a reasonable trade-off for choosing not to vaccinate yourself, and my guess is people who can't be vaccinated for medical reasons would be glad to consent to quarantine.

wjn0··on Graph Convolutional Networks (2016)
Yes. This problem is of particular importance in drug response prediction, where small molecules are represented as graphs where the atoms are nodes and the bonds are edges. These are difficult to work with because of the variability in the number of nodes.

[1] https://arxiv.org/abs/1509.09292

[2] https://arxiv.org/abs/1603.00856

[3] http://proceedings.mlr.press/v48/niepert16.html

[4] https://persagen.com/files/misc/xu2017seq2seq.pdf (PDF)

wjn0··on Graph Convolutional Networks (2016)
I think it's highly task-specific. In network biology for example, taking advantage of the network structure, either through node embeddings or directly, is almost always helpful (even necessary). I'm not as familiar with applications to social networks, or graph DBs.
wjn0··on Graph Convolutional Networks (2016)
Variational graph auto-encoders, also by Kipf and Welling, is another great short, understandable paper. [1]

[1] https://arxiv.org/abs/1611.07308

wjn0··on Applied Category Theory
Check out UMAP [1]. It's an alternative to t-SNE for dimension reduction that's grounded in topological category theory.

[1] https://arxiv.org/abs/1802.03426

wjn0··on A guide to difficult conversations
OK, but that's not a problem with NVC. "You are sloppy" is an evaluation. An observation would be "the reports you produce consistently have errors" or something like that.
wjn0··on The Adult Brain Does Grow New Neurons After All, Study Says
Well, fundamentally, neurogenesis is just another form of cell division. The most well known disease of cell overproliferation is cancer. Most brain cancers are of the glial cells (i.e. not neurons, but their support cells). There are some exceptions (e.g. some types of medulloblastomas).
wjn0··on Major survey finds worms are rare or absent in 20% of fields in England
> 42% of fields studied had poor earthworm diversity

> 21% of fields lacked surface-dwelling worms

Seems pretty clear to me. Some vs. none...

wjn0··on Logical difficulties in modern mathematics (2012)
Interesting that you chose statistics as your example. I've always considered statistics to be the one branch of math I've studied where intuition keeps pace with depth and rigour. The caveat being that I am by definition an amateur when it comes to any other branch of math.
wjn0··on Scientists are transplanting faeces into obese people with type 2 diabetes
So, let me make sure I have this right:

1. Fecal transplants are pseudoscience, because we don't know if they work.

2. We should assume they work, and identify specific microbes within them that are relevant or dangerous.

That's contradictory. Further, without data from examining #1, #2 is almost impossible - there are at least 1,000 species in our gut at all times. The only identification process I know of that could be reasonably affordable is 16S RNA sequencing, which is still expensive and would be nearly impossible to do properly on the 0.5 - 1.0 kg sample of gut bacteria (yes, that's the weight of just bacteria) from the average human. Not to mention that evidence suggests that gut bacteria act in concert, that is, we'd have to examine combinatorially many combinations of these 1,000+ species to get any real data.

This is just nonsense.

wjn0··on Scientists are transplanting faeces into obese people with type 2 diabetes
The trial is ongoing [1], so it's not surprising you can't find any information on it.

[1] http://www.isrctn.com/ISRCTN30359027

wjn0··on GNU Octave: A high-level interactive language for numerical computations [pdf]
> It's very hard to recommend that students and colleagues grow invested in toolchains involving proprietary software that they may not reliably be able to afford/activate/use in the future, particularly when some students may end up working in developing nations.

The future is here. As a research student in a developed nation, it's very annoying to deal with MATLAB projects. Our lab has a fairly strict no-new-code-written-in-MATLAB policy. Getting a personal licence is fairly straightforward as a student (for now), but running it on a cluster, or anything beyond my personal computer is very annoying.

wjn0··on The Chinese scientist who allegedly created CRISPR babies is being detained
I like that analogy. An even better one: porting the code of `sed' to a new operating system without fully understanding it, using `sed' to automatically correct bugs in fundamental operating system code without checks, recompiling said OS, and hoping it works perfectly.
wjn0··on Police decrypt messages after breaking pricey IronChat crypto app
Well, I took it to mean PGP vs IronChat. I'll take an open source implementation of a well-established protocol over a closed source implementation of another well-established protocol any day.
wjn0··on Police decrypt messages after breaking pricey IronChat crypto app
> Blackbox-security.com, the site selling IronChat and IronPhone, quoted Snowden as saying: "I use PGP to say hi and hello, i use IronChat (OTR) to have a serious conversation," according to Web archives. It wasn’t immediately known if the endorsement was authentic.

This strikes me as... very inauthentic. I think anyone with even a basic understanding of crypto would do things the other way around.

wjn0··on Alarm as China eases 25-year ban on rhino and tiger parts
Just because fascism and "communism" (as in, communism as implemented in China or the USSR) were at odds in the mid-20th century, doesn't mean they're incompatible.
wjn0··on How Vancouver Became a Money Laundering Paradise [audio]
"Intelligence" is an interesting (fictional) TV show that looks at the links between drugs, financial crime, Canadian-US relations, and the intelligence community in Vancouver. Supposedly was cancelled at the request of the Harper government for hitting too close to home re: water rights.
wjn0··on A step toward systems that can assess competence using neurological data
I think it's important to be careful with the vocabulary here (if only because of the site we're on).

There can be metrics that are targets which still remain good metrics. For example, in many machine learning competitions, the submissions optimize a known, given metric; but the test data is not known. Therefore, it is still a good metric.

I agree with the sentiment in this case, though.

wjn0··on A step toward systems that can assess competence using neurological data
Very curious about this as well!
← PreviousPage 2 of 5Next →