shoyer

22 ms·

181.
by shoyer 11y ago
As a lower level employee, you often lack the context to understand what the most important priority is. Even if you do have that context, you need to be able to convince management. Simply doing "what you think is right" is not a
182.
by shoyer 11y ago
This sort of thing works fine. We use it for pandas all the time.
183.
by shoyer 11y ago
I recommend PyToolz, "set of utility functions for iterators, functions, and dictionaries": https://github.com/pytoolz/toolz The functions in PyToolz are short, well tested and idiomatic Python (thought the f
184.
by shoyer 11y ago
Indeed, came here to say exactly this. K-d trees are even built in to SciPy, which the author probably already has installed (Scikit-Learn also has an implementation).
185.
by shoyer 11y ago
Yeah, Numba makes it awesomely easy to write fast functions in Python that release the GIL. You can already do this directly with dask.array by passing a Numba compiled function to the map_blocks method: http://dask.readthedocs.o
186.
by shoyer 11y ago
Indeed, those are both great options when possible. But easy access to parallel computing is still quite useful. For interactive analysis or building statistical models, you probably do indeed still probably want your data fit in memory. Bu
187.
Xray and Dask: Out-Of-Core, Labeled Arrays in Python(continuum.io)
53 points by shoyer 11y ago|6 comments
188.
by shoyer 11y ago
As an occasional numpy contributor, I would call this a bug.
189.
by shoyer 11y ago
No, squared error (the Brier score) is absolutely a strictly proper scoring rule for classification, which is the claim made in the article. See the linked Wikipedia article: https://en.wikipedia.org/wiki/Scoring_rule#B
190.
by shoyer 11y ago
I agree -- you shouldn't read sources you don't want to treat as derivative while you're writing an independent implementation. So yes, the code you've already written is GPL. But suppose you decide to reimplement the
191.
by shoyer 11y ago
Are you really suggesting that it's not possible for you write an independent implementation of an algorithm from the version you once read? Is this sort of "clean room" approach typical? It strikes me as absurdly cautious.
192.
by shoyer 11y ago
Most of this is luck. It takes a certain number of upvotes to make it to the front page and only links that happens to cross that magic threshold accumulate significant numbers of votes.
193.
Numba vs. Cython: How to Choose(eng.climate.com)
2 points by shoyer 12y ago|0 comments
194.
by shoyer 12y ago
Exponential improvements aren't very helpful if you're working on exponentially hard problems. Exponential growth of technology hasn't solved machine intelligence (by direct simulation) for essentially the same reason why it
195.
by shoyer 12y ago
Here's the original journal article (which is open access): http://www.ehjournal.net/content/pdf/s12940-015-0003-1.pdf
196.
by shoyer 12y ago
Cython is usually the most best option for maintainable, high performance numerics with Python these days -- it's the option of choice for scikit-learn and pandas. Numba can be easier in some cases (no manual typing necessary) but is s
197.
by shoyer 12y ago
It's not what they spent developing Sovaldi -- it's the fortune the industry as a whole spends developing drugs that never make it to market. That said, it's quite likely true that marketing is a bigger share of costs than R&
198.
by shoyer 12y ago
Really? The article talks about what external parties think about why these laws are being written, but doesn't any provide any direct evidence. And none of those even talk about "approved" sources.
199.
by shoyer 12y ago
I agree, but in some fields this is already done. For example, This is exactly why geoscientists have standardized on the fully self-described netcdf file format. With netcdf, you can specify units, axis labels and other metadata very strai
200.
by shoyer 12y ago
I agree. There is no debate or "problem" as far as quantum physicists are concerned. I'm not sure most of us would call our belief "many worlds", but you'll find very few who think that collapse is a real pheno
201.
by shoyer 12y ago
> A PhD is one of the few opportunities in academia to do true, unconstrained, open ended research. Sure, but is this really in conflict with publishing articles? The sort of scientific advances that require a full book or dissertation f
202.
by shoyer 12y ago
"Debunk" was my word, not Scott's, and you're right, it's overly harsh. I've edited my post. But Scott did convince me, at least, that this paper is nowhere near as relevant or important as the authors make it
203.
by shoyer 12y ago
I wrote Scott Aaronson about this and he showed me where he already addressed this on his blog: http://www.scottaaronson.com/blog/?p=2053#comment-269770 Note the response from one of the authors of this paper: http:&#
204.
by shoyer 12y ago
Could not agree more. I'd love to use an alternative to Uber that is less scummy, but I'm turned off by Lyft's style.
205.
by shoyer 12y ago
Here's the notebook on NBViewer: http://nbviewer.ipython.org/github/tmsgost/Bayes_Impact/blob...
206.
Show HN: Cython implementation of Python's collections.OrderedDict(github.com)
11 points by shoyer 12y ago|0 comments
207.
by shoyer 12y ago
Indeed, the evidence is pretty damning. My guess is that this ends with him fired from Ubert. Now that Lyft has presented the evidence, that may be the only way for Uber to avoid liability itself.
208.
by shoyer 12y ago
Totally agreed, I am much more excited about Numba right now. Numba has some really nice tricks up its sleeve -- I especially like the ability to easily write (g)ufuncs. There are a lot of edge cases to nail down for these JIT compilers. Ev
209.
by shoyer 12y ago
Based on the description in the docs ( http://pythonhosted.org/hope/lang.html#loops ), my guess is that "for i in range(N)" is directly translated into something like "for (i = 0; i < N; i++)" in t
210.
by shoyer 12y ago
Here's the doc page that lists the supported language features: http://pythonhosted.org/hope/lang.html Compared to numba, it looks like hope adds: (1) support for recursion, (2) automatic simplification of express

More ›