Diving Deeper Into Deep Dream: Different Distortions

The moment you get Google Research's Deep Dream project set up, you can make it do incredible things. If you want to generate trippy zooms with lots of dogs and eyes, the code Google provides works more or less out-of-the-box.

For reference, here is a gist for a .py file containing all the code from the notebook. This is meant to be run through an interactive interpreter, e.g. by providing the -i flag at runtime or, in iPython, starting a new notebook and then executing e.g. "run demo.py".

And, here is a gist for a similar but extended piece of code which takes an input file on the command line and performs the iterative process described at the very end of the researchers' original iPython notebook. Both these files are almost entirely composed of code from that notebook; they're just provided as references, so we can have some well-established starting points from which to branch out.

Setting Up Deep Dream, Google Research's Hallucinatory Work of Genius

Google Research wrote recently about a technique they call "Inceptionism", which needs to be seen to be believed. Click through and check out their pictures. A full gallery of official images can be found here. The techniques used to generate these sorts of images were described in broad strokes in this blog post, but the level of detail stopped just short of what someone wanting to replicate their results might have wanted.

That is, until they published their source code.

In this repo (which was created just this morning, as of this post's writing) is an IPython notebook which contains everything necessary to duplicate the results you see in their blog post. That's so cool that it's honestly a little hard to believe. The dependencies are detailed in the repo. There are surprisingly few of them, and they're all surprisingly easy to get set up. Let me walk you through what I had to do on my Debian Stretch system to get everything up and running.

Evaluating the Alternating Harmonic Series

Here's a fun little note I just found while organizing my desk. I discovered this proof during a class on infinite series last summer. The professor mentioned the limit of the alternating harmonic series, then commented offhand that "this limit is true, but we cannot prove it yet." Never one to shy from a challenge, I decided to see whether he was right, and had roughly this proof scrawled on a piece of paper by the end of class. He suggested that I try to publish it in an undergrad journal, but that didn't pan out for various reasons. Nevertheless, it's a fun proof, and I wasn't able to find it anywhere online, so I thought I'd share it here.

The alternating harmonic series is defined as

which, of course, is just the harmonic series with the sign of every other term flipped.

It is well known that the harmonic series diverges. Slightly less well known is that the alternating harmonic series converges to a very clean limit -- strangely enough, it turns out to be ln 2. This fact, if it comes up, is usually proved just as a trivial consequence of the Euler-Mascheroni constant's most common identity. Not only is that proof beyond the reach of many undergrads, it's also so trivial as to border on inane. It's like using a sledgehammer to pound in a nail: of course it works, but is it really the best way? Someone pass me a hammer.

Consider the following series:

Finding Files with Multiple Substrings

I'd like to share with you a challenge I ran into last night, a problem that's really difficult unless you take to the command line.

Some background: I play Go. The GoGoD game database, which I recently bought and downloaded, has somewhere on the order of 80,000 Go game records in it. These games are all in .sgf format, which is a simple plaintext format. Records of games from professionals of many different ranks are included. The rating scale for professionals starts at 1 dan (or "shodan") and goes up to 9 dan, the highest rating a player can achieve. Games between professionals are always interesting, but games between 9dan professionals are, in particular, packed with subtlety and value for the ambitious student of Go.

Any game from this database is worth studying, but with over eighty thousand games to choose from, it helps to be able to pick highlights, and what better way than by focusing on games between 9dans?

If we decide to take this route, we're immediately faced with a second question: sure, we can just open games at random until we find one between 9dans, but this is boring and slow. Are we doomed to tedium, or is there some way we could search this database of more than eighty thousand games, pick out only the games between 9dans, and copy these over to a special "elite database"?

To the average user, this might sound impossible. On the command line, it's almost trivial.

In Defense of Honors Programs (or, Read a Fucking Book)

In planning the follow-up to a survival skills event for incoming Honors freshmen, the question came up of how best to articulate to new students the value of WWU's honors program. Especially for STEM students, the fact is that the liberal-arts-focused Honors courseload can be expected to overlap very little (if at all) with their major's courses. That's a non-negligible number of extra credits, and there's nothing scarier than a delay in graduation, and so lots of people in the program question, at one point or another, whether the program is worth seeing through to the end. It's our job to try and make the case that it is.

Of course, regardless of obligation, it'd be disingenuous of us to argue this point if we did not ourselves wholly believe it. For my part, I am lucky in this regard, because in spite of (because of?) being a Math/CS double major, I'm convinced that the honors classes I've taken have on average been far more valuable than anything I've taken in either of my majors. Ditto for the classes I'm planning on taking. That's not because I don't like my majors -- far from it, I'm crazy about them. But computer science is close being treated as a trade skill, and math's "beauty cold and austere, like that of sculpture" (Russell) nourishes only certain parts of the mind. Both majors are fantastic for what they are, but what they are is not a complete education. Similar things could be said for practically any other major: They leave gaps, and honors fills those gaps.

My first, informal attempt at articulating to my co-planners this argument for why to stick with honors went something along these lines:

Why stay with honors? Because we offer things you can't get anywhere else. We have classes you can take as a freshman that're on par with the upper-division classes in specialized majors. You can just be like, "I want to learn about Russian literature this quarter" or "I want to learn about film this quarter" and nobody's going to be like, what the fuck, you're a chem major. These kinds of classes are just there, and anyone in the program can take them. If you don't want to be an English major but you want to study great books, we're your best shot. If you want to be in rooms full of clever people talking about Nabokov or Robbe-Grillet or Bely or Dante or for that matter anyone else worth reading, we've got you covered. We've got some of the best professors in the university teaching our classes. Great classes, taught by great professors -- what more do you fucking want?

2015 iCTF Retrospective

After some delays, last Friday saw the 2015 UCSB iCTF come and go, and it was quite the experience. This is (to my knowledge) the first year that Western has joined the competition. We didn't exactly place first, but it was great fun and lots of lessons were learned. Now that it's over and we've had a few days to reflect, here are some thoughts on what we did right, what we could've done better, and what I didn't go in expecting.

There were two big surprises, for myself and (I think) the rest of the team: First, the sheer mind-boggling number of exposed vulnerable services -- upwards of 40, I'm pretty sure -- and second, the inability to directly attack other teams.

You know what?

It doesn't seem like there's any good way to handle armchair "philosophers" who think they have something deep to say about the idea of certainty. "See, the thing is, you can't actually know anything for sure! I mean, like, what if we're all just brains in vats?" The only compelling argument I see for this point proceeds by example: The claim is that nobody can truly know anything whatsoever. Anybody who can in all sincerity make this claim must know very little -- perhaps little enough to serve as supporting evidence.

Joking aside, this is one of the silliest ideas out there. Those who contract it generally do so within a year of their first serious exposure to the great philosophers, in that honeymoon phase where everything makes sense and the ideas of "discarding all assumptions" and "questioning everything" still seem novel. Socrates and Descartes are particularly severe offenders, as brief or unconsidered treatments of their works can easily give the impression that they advocated this total uncertainty. What, more precisely, is this misconstrued idea, though?

It's the idea that we just can't know anything for real, man. You know? Like, think about it -- everything we know comes from observing the world, via our senses, right, and how can we trust our senses? What if we all see, like, different colors? What if two plus two doesn't always equal four? What if it could equal five, but we don't believe that because we've never seen it? Or because we can't see it? Would we even know how to count if we were all blind? I, like... I dunno, dude. I dunno. Whoa. We don't even know that. We don't know anything. Not really.