Tuesday, July 31, 2012

Tuesday: No, it didn't work

It basically just gives you a very complicated calculation of the median.  Crap.  I guess I'll need to think about it more.

What time is it?

Go, Team Red Panda!

  • I guess if owning 14 guns isn't going to make you not live in fear, shooting someone probably is going to make you think it's his fault.
  • Republicans are still trying to kill the post office.
  • Is it really "fanart" if you work on the TV show?  In any case, hilarious.
  • Fucking finally.  I suspect this will also magically kill the "$10 a day for wifi" scam hotels have been running for the last forever.

Monday, July 30, 2012

Monday: I wonder if that would work

Take the data vector, run a k-means scan on it to determine group means.  Then, calculate the probability that that mean is a good representation of the data as P = prod(1 - normcdf(abs((X - K_means) / sigma))).  That gives you weights for each of the input means telling you which is dominant.  Use that value, and then calculate weights proportional to the weight relative to that best K_mean times the weight from the data value variance.  This should shrink the weights of things that don't match the K_mean sharply, decreasing the effect of outliers.  Values close to the best K_mean (which, naively, should be close to the median) are also weighted differently, but are all roughly weighted the same.  This should give a weighted mean that is fairly resistant to outliers.

This should be pretty easy to calculate, so that's nice.  My only concern is that it breaks down as the fraction of outliers goes above 0.5.  Then the K_mean is hard to select, and you start weighting the components equally.  I guess that's the right way to do it, as when you have no clue about the answer, you should gradually return to a regular weighted mean.

Anyway, math.

Birthday Panda looks terrified.

Watch out, Bear! That's probably hot!
I had the idea to reheat my chicken saltimbocca en papillote, which worked out well.  It kept the moisture in, and basically steamed the chicken back up to temperature, which allowing the bottom to warm up directly from the cookie sheet.  Good idea, and I think I'll do it again tomorrow.  Another thing that was a good idea was spooning the sauce from yesterday into the bun, wrapping that in foil, and letting it steam as well.  I didn't bother separating the fat out, making it like buttering the bread (with a bit of chicken jello for flavor).  It helped keep the bun pliable when I was loading it with chicken slices.  I do think I'll need to double check the sage before cooking.  Today's chicken was lacking that bright sage flavor that I want to cut through.

Sunday, July 29, 2012

Sunday: I win at sticking to the menu plan

Even though it meant having to clean my kitchen.  Stupid manual clean kitchens.  Anyway, I've already done the recipe for chicken saltimbocca, so I can skip that bit.  I was lazy today, and didn't pound out the chicken, nor did I roll the chicken.  It worked out tasty anyway, and had it not been like a thousand degrees in my kitchen, I would have done the full thing.

Once done, I deglazed the pan with some vermouth, and let it reduce into a sauce.  Then, slice the chicken against the grain into ~centimeter strips, and put into hot dog buns with some Parmesan.  Boom. Chicken saltimbocca sandwiches.
I ate two because I didn't have lunch today.
This worked really well, even though I was disappointed to learn that my buns were already sliced.  I was hoping to cut it myself, so I could do a top cut instead of side, sort of like a lobster roll.  That was kind of what the idea behind the whole meal was, in any case: "like a lobster roll, but with one of the best chicken recipes ever."
The other two thighs and the sauce.  I'm hoping the fat separates out in the fridge, so I can skim it tomorrow.

So this is a time lapse 4-pi steradian photo?  Huh.

Damn, I knew my apartment was missing something.  I need to go find the "giant white sphere" 

Saturday, July 28, 2012

Saturday: Things that are not the same

First off, there are very few values for which (2 * u * F) == (1 - 2 * u * F).  Pretty much one value, I guess. (u * F = 0.25).  So yeah. Those aren't the same.

Secondly, it turns out that after spending an hour and a half trying to figure out what I wanted for late lunch/early dinner today, I discovered that the burrito place has summer hours now, and closes at 3.  Damn. However, the burrito place is close to the Mexican place, so I went there instead.
Chips! How totally worth taking a picture!

I apparently only get one thing at the Mexican place.
 It wasn't as good as it has been in the past, with meh service and gummy rice.  Disappointing after I had built up by brain for a burrito and chips and queso.

Bear time:
What are you doing bear? There is no lifeguard on duty!
I also came up with a menu plan for this week.  It is totally the best plan ever.

  • Sunday: I bought chicken, sage, and prosciutto.  Yep, chicken saltimbocca again.  My actual thought is to make this, and then make saltimbocca sandwiches.  I think it'll work deliciously.
  • Monday-Wednesday: I have four chicken thighs.
  • Thursday-Friday: I also bought two buffalo hot dogs, because they looked good and I didn't have a good menu plan.  Then I bought the hot dog buns.  Then I came up with the saltimbocca sandwich idea.

Friday, July 27, 2012

Friday: headaches

Lots of headaches.  Tomorrow I need to do more work, plus get groceries, plus clean out my fridge to get rid of the crap that's too old to eat, plus like a dozen other errands.

Thursday, July 26, 2012

Wednesday, July 25, 2012

Wednesday: N-dimensional logistic regression

I'd hoped to have it done today, but instead I spent the evening sorting out data structures.  Blah.  At least I should be able to get something working tomorrow.  In case you don't want to look it up, logistic regression allows you to define classes of data in the training set, and then comes up with the best probabilistic model that can be used to classify a set of unknown data.  Kind of cool, huh?  It does require some understanding of the input classes, so it's a supervised learning method.  Not quite as cool as a magic oracle algorithm that could figure it out for you, but that kind of borders on impossible.

Yeah, that ended up being my decision as well.
Are you cold, tiny kitten?


  • I have no links today?
  • How is that possible?
  • Now I want ice cream.

Tuesday, July 24, 2012

Tuesday: Why is it almost 11 and I haven't had dinner yet?

The answer: because I don't have a menu plan and nothing sounds good.  :-/  I probably should have just picked up something on the way home.  I have bagels. Let's eat a bagel.

Ok, Operation BagelTown didn't go off quite as easily as I thought.  Given that it took an hour.  On the plus side, I started some laundry, so I'll have clothes for tomorrow.  Always a good thing.

FFTs and me made friends today, even though I'm still not totally happy with the results that came out.  It's hard to actually have to think about this crap in the transformed space, as my usual treatment is convolution or something like that where I'm just using the FFT to make a hard problem a multiplication.  Then reverse transform, and you just cheated at math.  It's like multiplying numbers by adding their logarithm, that kind of cheat.


Ok, let's look at "Missiler" here. A) Stupid name. He "missiles" I guess?  I was unaware we'd verbed that word. B) I'm not really seeing any missiles anywhere on his body.  Maybe those two tubes on the chest that look like tennis ball tubes?  C) Maybe that rocket in the background just shot out of his left shoulder, perspective kind of works.  If so though, how useful is it to shoot a missile nearly straight up, but also kind of backwards to where you came from.  I think Missiler is going to have to answer some friendly-fire charges after Space War is over. D) How does he even walk?  No knees, and it doesn't look like he has any ankle level articulation.  Does he just scrape his foot up to some angle, tip over to fall down on it, and then straighten back up while scraping the other one to match?  That things going to need new feet like every step.  E) I'm pretty sure the guy on the car is just pointing at it while shouting, "Holy Shit! That is by far the ugliest and most useless robot ever.  Back up and ram it and see if we can knock it on it's back! I bet it'd be like a turtle!"

Missiler lives in a world of constant useless annoyance.  Those hands make it look like he was designed to hold trays, possibly as a butler.  But now, here he is, in the middle of Space War, trying to figure out exactly how he's supposed to missile.

Hopefully this one won't be as long as the last caption.  Bears. Winged Bears. Winged Bears with Laser Eyes.  Look in the background.  Smoke.  Clearly these WBwLEs have fucked somebody up already.  I will now simply point out that bears are carnivores, and have forward facing eyes to gain the benefit of stereopsis.  Since the lasers nicely point out where each eye is located, we can finally classify these as Winged Bears with Laser Derpy Eyes.  

Down on his luck squirrel has to go dumpster diving for food. :(
  • Squirrels!
  • I've read at least one of this guy's papers in the past, and I didn't buy it then.  A) his "fit" is pretty crappy. B) If the universe has a preferred axis, it's unlikely to exactly line up with RA. C) isn't the Sloan footprint that spidery thing that covers the north galactic cap pretty well, with spider arms down through the plane?  That doesn't seem like the best test sample in that case.  D) Finally, he "eliminates edge on spirals from consideration"?  Really?  If your theory is rotation direction A that way, and opposite rotation direction -A the other way, then somewhere in the middle you should see mostly flat things, don't you think?  Otherwise it's not really a "single universal axis" or whatever he's calling it.  Or, you know, spectral follow up to get line-of-sight velocity curves?
  • Boring but somewhat important economics article.
  • Sally Ride.  Honestly, the first two questions on this blog post were basically what I said when I heard.
  • I want to go to Book Camp.
  • I'm tired of the olympics already.  It's just a money grab where they build up athletes hopes while filling their pockets with money, leaving the host city worse off afterwards.
  • Some fucking common sense at last on this story.
  • Another peaking squirrel to get you in a good mood!
  • Just to have it crushed by this absolutely horrible story.  If I had Winged Bears with Laser Derpy Eyes, you know who I'd send them to find.
Ok, time to twiddle the date on this so it shows up around when it should have.

Monday, July 23, 2012

Monday: Stupid FFTs

Mostly because they really only want nice evenly sampled data.  I could zero pad the data until it is evenly sampled, but then it's like a billion samples long, >97% of it being zero.  Since my sampling is basically of period A and period B, I kind of feel like I should be able to take the FFT(A_sampled) + FFT(B_sampled) and get something reasonable, but I can't find a proof of that already online, and it's too late to do it now.  But really, it should just be an extension of the linearity principle, right?  If F{g(t) + h(t)} = F{g(t)} + F{h(t)}, then we should be fine.  We just cheat a bit and let g(t) be those samples sampled at rate A, and h(t) be those samples sampled at rate B, then the total signal is just g(t) + h(t), so the fft can be separated into those two components.  Blammo.

Except proofs that end in "Blammo" probably aren't as rigorous as I'd really like.  I guess in the worst case scenario, I do have a LS periodogram program already written, so that could probably give me something like a reasonable answer.  That is kind of a O(N^2) operation, though, so it sucks next to the O(N log N) FFT method.

Ok, enough math rambling for today.
This might end up like the recurrent bears.

What are you doing, bears? Sears doesn't have anything in your size!

What are you doing, sheep? You're not zombies!

What are you doing lady? You're not...oh. Wait. Nevermind then.

  • I totally had this problem in college with certain professors.  No matter how interesting the topic was, the voice was too soothing, so I'd always fall asleep. Even if I was in the front row. Like six feet away from where he was lecturing.
  • Little known fact: lots of people in the 80s had an eye replaced with a screw.  Just for fun.
  • I've kind of wanted to try this place, but it's in Waikiki, and parking sucks there.  Still, 25 layer katsu sounds interesting.
  • Umbreon shoes.

Sunday, July 22, 2012

Sunday: When you have three things you want to do, it's best to do none of them.

That way, none of the things feel like they've been cheated.  I am totally sure there is no logical flaw in this plan for the day.

I still don't have a good menu plan for this week, and that's probably not a good thing.  I think part of the problem is that everything I come up with seems to be something that I make too often, so I decide against that.  Any similarity to the problem brought up in the post title is purely coincidental.

Let's do today's random pictures:
I don't know why I like these captcha cartoons, but I do.

"Jim, look at this asshole. Hey, dumbass! Only bison are allowed to go this way in this lane!"

I can just imagine all the crap that would end up hidden in the overhang.

Wasn't this an episode of Pinky and the Brain?

It's like we're being attacked by z = (x^2 + y^2) * cos(4 * atan2(y,x))

Saturday, July 21, 2012

Saturday: Lazy Saturdays are becoming normal

I really need to sit down and actually make a meal plan.  Also eat dinner. It's after nine already.

I did do a bit of work today, so that's something.  Not a lot, so I maybe should do more tomorrow.
Moon + Jupiter + Jupiter's moons.

I have no idea what this is supposed to be, but it's kind of cool.  Basically illustrating how even in random noise, it's easy to make correlations between two data sets.

Well. That's a cool picture.
  • Man, I'm sure glad the NYTimes is doing hard hitting stories applicable to regular people.  I'm sure everyone I know is worried about how to talk to their kids about the millions of dollars those kids are going to inherit.  That totally seems like a common problem.
  • I told you it was a lazy day.

Friday, July 20, 2012

Friday: The writing part of science isn't as fun as the science part.

It's not even as fun as the math part.  I really need to be in the right frame of mind to accomplish much with writing.  The other stuff flows a lot more naturally.
"Seriously, I promise not to eat you."

He was riding high on his life of drugs and guns, but Pikachu knew that one day, it would all come to a horrible end.

Thursday, July 19, 2012

Thursday: Math, infinite math

So I started working on a problem, and then
This was a great episode of Adventure Time.  "I floop the pig."
I had working N-dimensional, K-mode Gaussian mixture model code.  I wrote a 1-dimensional, 2-mode version for my thesis, but this is clearly (N-1)*(K-2) times better.  Plus, as a side project, I wrote a k-means code that I can use to feed the GMM code with better than random initial positions.  I should really write these things as functions instead of full programs, so I can use calls to the functions in other programs.  I'll do a refactoring at some point that does this.

This gif is apparently 99,000 times larger than I thought.  Oops.

This is the Titanic.  I don't actually care about the boat, but I like cutouts.

"Ice cream eating motherfucker, that's what you are."

Today's bear picture.
I also decided that a bacon sandwich with tomato soup totally counts as a healthier BLT.  Plus: bacon for tomorrow's lunch sandwich!

Wednesday, July 18, 2012

Wednesday: Did you realize that "here" and "there" do not rhyme?

I didn't actually notice this until I discovered that only one of them rhymed with "bear".  This was important because of things.
I saw these yesterday.  I don't know what makes Hello Kitty macadamia nuts different. Probably just the packaging.  Still, Lilo and Stitch seem to be like 10 cents more expensive.

Me, all day long. Well, sans pikachu. I don't have one of those.


Bears come in baskets.

Tuesday, July 17, 2012

Tuesday: Pretty much all day smelled like dog buns

Of the six things I tried to accomplish today, all six turned out to be harder/more complicated/fundamentally broken/whatever than I expected.  Other than the experience of failing at six things in one day, I should have just stayed in bed.
Basically this was my day. In this analogy, I'm the bunny, and all of existence is the snow.


I believe this guy is the King of the Lions.

Have I posted this bear? Whatever. It's a bear.  I like bears.

This is a baby red panda.  I'm unclear as to the shape of the function cute(t) for red pandas. Constant?
No links today. Too much time bashing my head against things.  Or the internet sucked today.  Anyway, let's listen to FDR piling on the snark while discussing a political constant.  It's a minute long, so it's not like it's going to eat up a lot of your day.

Monday, July 16, 2012

Monday: Why am I so tired?

I almost fell asleep when I got home.  Not what I was planning on doing.  I did discover today that the inconsistent data I was looking at over the weekend is a known bug, so I have to calculate my own averages instead of relying on the averages I downloaded.  Crap.
That's what I have to say to you, data.
Today's bear:
"Hey, are you guys open?  Can I sit on the patio?"

I also discovered that if you other-click on the file browser in the upload thing, you can set "bookmarks" for certain directories.  Like the one where I download all the things I see online.  And the one that my phone auto-syncs to.  That's a useful feature I never knew about.