cgs1019

`cat -n` prints with line numbers, which is occasionally handy.

Oh yeah that's gonna be a dense read then :)

I really like Kevin Murphy's books. I learned a lot from them when I was getting started on this stuff. He's just released a two part follow up to his 2012 book. Both parts are available online.

I also really like David Mackay's book (also free online). It's...idiosyncratic. But fantastic!

Best of luck with your learning! Start simple, go slow. Write code every step of the way!! ^_^

To grok the ELBO, don't start with VAEs; start with basic variational inference. https://arxiv.org/abs/1601.00670 is good, though maybe increasingly dated...Once you're happy with VI in general, https://arxiv.org/abs/1606.05908 is a nice read on the VAE.

Hi, tech lead for TFP here. The wording here was unclear -- sorry! We're fixing it presently.

We are not migrating away from TF; far from it!

The change here was to interoperate with TF and JAX (and numpy!), by way of some rewrite trickery under the hood. Essentially, we wrote a translation layer that implements the TF API surface (or, the parts we actually use) in terms of numpy & JAX primitives [1]. This lets us leave most TFP code intact, written in terms of the TF API, but interoperate with JAX by way of the API translation layer. (Actually we implemented numpy support first, and mostly got JAX for "free" since JAX is largely API-compatible with numpy).

Sorry for any confusion!

We're pretty stoked about this work, so happy to answer any other questions you may have (also feel free to chime in on the github tracker or email [email protected])

[1] - https://github.com/tensorflow/probability/tree/master/tensor...

The link cites a vice article that cites a guardian article: https://www.theguardian.com/world/2016/jul/01/philippines-pr...

Plain pdf: http://arxiv.org/pdf/physics/0312087.pdf

Neutrons are electrically neutral but they still interact electromagnetically (they're composed of charged quarks). They have a magnetic moment, for instance [1]. So they wouldn't be dark the way "dark matter" appears (heh) to be.

[1] - https://en.wikipedia.org/wiki/Neutron_magnetic_moment

Google's auto library, auto/value in particular, is worth a look: https://github.com/google/auto/tree/master/value

I've always thought it would be fun to somehow incorporate technology into climbing for tracking or otherwise augmenting the experience.

Time trials sound like a quick avenue to painful injury, though. If I were a gym, I'd be cautious in encouraging it. In my experience, climbing prowess is not a matter of speed but control -- the more statically you can complete a route the more control you demonstrate. A friend of mine once said his goal is to complete a route as silently as possible, avoiding accidental scuffs of his shoes or knocking of his knuckles on the wall. This is actually a great measure of control, too.

Oops, maybe I unwittingly had access by being on work VPN. Sorry bout that! Thanks for the open link, jacquesm.

Looks like Nature makes the whole paper readable online: http://www.nature.com/nphys/journal/vaop/ncurrent/full/nphys...

It's not as snazzy as in the article but this alias in my .gitconfig is an integral part of my git workflow:

  gr = !git \
    --no-pager \
    log -n 16 --graph --date-order --date=short --branches \
    --pretty=\
  '%C(yellow bold)%h%Creset\
   %Creset %C(blue bold)%d%Creset\
   %C(white bold blink)%s'

The 'gr' is short for 'graph'. You can tweak the -n to show more history. I actually have several of these, 'gr', 'grr', 'grrr', etc. to show more and more history.

They make slightly less than the median tax filer. The difference is potentially significant (I have no idea what the statistics are)

paper on arxiv: http://arxiv.org/abs/1412.1521

So now at traffic stops, "license and registration" can become "unlocked cellphone and registration"? No thanks!

I'm guessing they have established relationships with the card issuer and will use your card number only once to establish the link, after which the card info isn't needed.

Baez's google+ posts are always super interesting, informative and full of references for further reading. The comment threads are generally also very interesting and active. Highly recommend following him.

I may be off base here, but I think the distinction is between number of degrees of freedom (what you describe) and density of bits of information. Bekenstein's result showed that the information in a black hole is proportional to its surface area, not its volume. This seems unintuitive in the sense that if you, e.g., stuff more ram chips into a box, its information content seems to increase as a function of volume; the disparity lies in the fact that bits in ram are not as densely packed as in a black hole.

EDIT: misspelled Bekenstein

Buried deep inside the article is a link to this preprint: http://arxiv.org/abs/1403.4226

downvoter(s) -- this is a subtle, poignant, and (IMHO) poetically written joke about the prevalence of non-convexity in real-world optimization problems.

I always wanted to try this! I imagined having two antennae, though, positioned orthogonally to give a 2-d signal.

"A deep truth is one whose opposite is also a deep truth" -Niels Bohr