I found a cool Android app that fits a trend line to weight measurements and predicts meeting a goal.

Getting my weight down is something my doctors have been bugging me about.  I’m using the same technique I used before, John Walker’s Hacker diet, but this time I’m tracking it with another Android app linked to the fatsecret.com site.  What neither the site nor the app had was an easy graphing/trendline capability.  Now I have that, too.

Looks like I’ll get down to my goal just in time to ruin it over the holidays.

I like the idea of getting something regifted.  It appeals to both my sense of irony and my sense of utility.

My wishlist, if anybody cares, is here.

But, if you have something you’re not using anyway, particularly still in the packaging, that you think I’d find useful or amusing, by all means regift it.  Got a strange kitchen tool you don’t use?  I’m trying all sorts of new things in my kitchen.  Have a set of cool wineglasses, but you quit drinking years ago?  I have a nice Pinot that needs a glass.  Did you misunderstand a DVD title, and it turned out to be some weird science-fiction soft-core porn?  Well.. I’ll give it a look, anway.

It just seems to me that having a gift whose only purpose is to demonstrate that gravity is still working is of zero, perhaps negative, value.  If nothing else, I’ll find it amusing.

tc>

(Renny Gleason’s TED Talk – 3 minutes.)

tc>

Then I was surprised by a failure to converge!

But then I got it in one with my next attempt.

Mary found a nice one in an old favorite.

She thought lyrics might give good results, but not Penny Lane.

A base slander led to a three-part loop.

Some things just translate well!

We had a romance novel fail.

And this reminded me it’s time for bed.

tc>

We have some simulated data, a cutout of which is shown here, that gives you the general idea of what the data will look like if it’s pretty good.  Jay Anderson (STScI) did this simulation.

Simulated NIRCam Data with 1% bad pixels

That picture is 1 percent bad pixels, which is pretty good.  For the roughly 4 megapixel NIRCam, you’d expect to get 40,000 bad pixels, before you add any cosmic rays.  That means one in every hundred pixels is a bad one.  It could be one in 50.  And still, that’s pretty good.

This is after we’ve done all the standard calibration:  Flat fielding, dark correction, geometric distortion and whatever else we can figure out.  Most of the cosmic rays will get removed, but maybe not all of them.

Most people look at that picture, and they are shocked.  The pictures you see on the front page of the newspapers, or in glorious color magazine covers, or in big wall-sized displays don’t look like that.  But to me, they look fine.  Good, even:  There are no ghosts, no big artifacts, no column bleeds like you see on CCDs.  It’s very nice.  I didn’t get it.

Mary explained it to me.

Most people’s experience is of normal optical cameras with very short, almost noiseless exposures against very bright targets, and those targets are “extended sources.”  Things like people, or cars, or trees under a bright sky.  These exposures are measured in hundredths of a second, and there are plenty of photons falling on the detector, and the exact efficiency of the detector isn’t all that important.  Defects are very rare.

For displays, people are even more demanding.  Dell’s specification for a defective monitor is 0.0005% bad pixels. That is, about six on a typical 1.3 megapixel display.  Even that number is controversial:  Some users demand replacements of monitors with a single bad pixel, and if they game the system correctly, they get it.

Even if you do have some bad pixels, say from noise in a low-light situation, you probably don’t notice them.  If you get a hot pixel from a reflection off somebody’s eye, it probably gets corrected away when you do red-eye correction.  If you have a few dead pixels in the corner of an image, you never look at them, because you’re looking at the whole picture, not the individual pixels in the corner.

So it’s likely that most consumers never see any bad pixels, so if you say images have some bad pixels, but they’re pretty good, they figure you’re talking about a handful, and if they are there, you won’t see them.  And they don’t understand the numbers, because you’re talking about 98% or 99% good pixels, and 98% is an A in any grading scheme.

For the HST detectors we currently use, and for the JWST detectors we’ll have in a few years, the underlying assumptions are different.  Exposures aren’t hundredths of a second, they are hundreds of seconds.  We spend literally minutes, sometimes hours, trying to hold the camera still on a target that probably isn’t visible at all from Earth, because it’s too faint.

The scene we’re looking at isn’t bright, it’s mostly dark.  (If you’re reading this  at night, try this:  Take your cellphone camera outside, point it at the sky and take a picture.  Unless you manage to snag the moon, I expect it’s completely dark.)

Finally, we don’t look at the whole picture anyway.  Even the HST observations of Jupiter don’t fill the whole camera.  We zoom way in, as far as we can and still see something meaningful.  We may be looking at a section of the image only a few hundred pixels across, so a single pixel is two tenths of a percent of the width of the image.  That’s 400 times the size of the defect Dell says they’ll replace.

The detectors are very efficient, and very susceptible to noise.  They’re in space, so they live in a nasty radiation environment.  They start with lots of bad (either dead or hot) pixels, and they get worse (to varying degrees) as time passes.  Most of the bad pixels will be in the corners, but some will be scattered all over the place.  If only one in a hundred is bad, we’ll be happy.  But for JWST, that means more than 40,000 pixels will be bad.

That’s fine, though.  We deal with that by taking pictures in pairs, with very small offset on the sky between them.  We stack the second picture on top of the first, and when we find a bad pixel, we replace it with the data from the other.  Often, we average the “good” pixels to improve the overall image quality.  Unless you get terribly unlucky (which can happen) there is a good pixel for any part of the sky on at least one of those images, and when you add the two pictures together, fewer than 100 remain.  If you use multiple pairs, say from three or four colors, and combine all those images, even if a bad pixel remains you’ll never notice it.

The insight Mary gave me was that while “1% bad pixels” means tens of thousands to me, for most people “1% bad pixels is pretty good” actually doesn’t mean anything to most people.  They hear “pretty good” and think of their camera, or their monitor.  Understanding that I needed to unpack the numbers, and provide a picture like the one above, helped an awful lot.  I appreciate that, and want to publicly say “thanks.”

tc>

The book started as a series of essays, some of which I’ve read before. There are three sections, and each has a theme around improved performance.

Diligence is about how to do something well, and it boils down to just that: be diligent. Planning things, making sure the plan is complete, and executing the plan; washing your hands every time; tracking your own performance and being honest about failures.

Doing Right is about understanding what will improve your performance, and then actually doing it. The Right Thing isn’t always obvious, and deciding on right conduct can require serious soul searching. Having decided on the right thing, however, means accepting the obligation to do right.

Ingenuity is about metrics, of all things, and about finding good metrics and applying them to your work, and about finding innovative ways to improve performance.

In an afterword Gawande makes five Suggestions for Becoming a Positive Deviant.
1. “Ask an unscripted question.” Ask about things that matter to the other person, and actually listen to the answer. This may make other people seem more like real people, and less like machines.

2. “Don’t complain. … It’s boring, it doesn’t solve anything, and it will get you down.”

3. “Count something. …be a scientist in the world.” Some metrics are better than others, but some metric is better than no metrics.

4. “Write something.” This is the same advice Cliff Stoll gives everybody. Publish something, somewhere, in the hope of contributing to a larger world.

5. “Change.” This doesn’t mean chasing every fad, it means recognizing that you are not doing everything you might as well as you might, and you need to try new solutions.

Despite being 250+ pages, I found it a very fast read. It is at times inspiring, at times disheartening and occasionally terrifying, but Better is well worth the read.

tc>

The first fifth of the book is about the process of moving from a world powered largely by muscles or water, and lit only by fire, to industry powered and lit by electrical utilities. No single invention — the electric generator, electric motors, electric lights or power transmission lines — was the key to this second industrial revolution. Rather, it was the construction of systems of generating, delivering and using electricity that led to a complete transformation of industrial production, and laid the foundation (with Hollerith and Watson) for the next great change: digital computing.

Computing is in the process of being transformed as well, this time by the interconnectedness of computing systems and the synergies that networks — local, corporate, national, global and finally home — that are generated by combining computing and networking.

The transformation is not always positive. Carr describes the effect of open communication on intelligence gathering. The United States, which purports to be a defender of liberty and based on individual civil rights, uses the interconnectedness of networks to undertake the most intrusive domestic surveillance ever seen, mostly out of sight and secret even from the Congress and the Courts. At the same time, Iraqi insurgents were using Google Maps to find and attack British troops near Basra. The “communication” networks have led not to better, richer communication, but better intelligence tools for warfighting.

For me, the most surprising part of the book (though it should not have been a surprise) is the discovery that blogs aligned along political viewpoints are making online communities increasingly polarized. The disparate bloggers rarely even read each other’s stuff, and when they do link to each other, it devolves into namecalling. That leaves centrists like me — a card carrying member of both the Republican Party and the ACLU — with nowhere to go.

Part of this polarization is clearly because the nets unite us. People formerly separated by geography find like minds around the world, and quit talking to the people next door or across the street. Since people across the street are real people, and you have to at least listen to them to accomplish things around the block/neighborhood/school/community, you end up having conversations with people who disagree with you.

On the net, you can find people who agree with you, and work together to either enjoy the online community, or organize to change society at large, or some combination. That’s very powerful, and may lead to a better world. If you’re the only sex-positive non-Christian in the PTA, it’s hard to lobby for change. If you’re one of a few thousand (or maybe few million) on the net, you’re not quite so alone, and you have the courage to Do The Right Thing for your daughter and women everywhere.

However, on the net, you tend to stay away from people who don’t agree with you, and groupthink can quickly set in. Suddenly the very nice person who always helps out with school dances and bakes desserts for the teachers on conference nights is nothing but a raving right-wing lunatic who wants to ban all television over a wardrobe malfunction.

The electric lightbulb, Carr opines, started the breakdown of family cohesion. We were no longer huddled around the fire, but instead we were increasingly free of the darkness, and increasingly in separate rooms, with separate interests.

The web may be the informational light bulb that breaks apart communities and builds new ones. To be a positive thing, this new tool leading us out of informational darkness must also get us to talk to each other, and get us beyond shouting and namecalling.

tc>

Platform     Proposals
Mac OS X        492
Linux           297
Windows         115
Sparc Sun        46
Not Given        11

Five years ago, less than ten percent of proposals were done on Macs. This year it’s more than half.

Of course, the more interesting numbers would be the percentage on each operating system for accepted proposals, but we won’t know who is accepted until the end of May.

(Put another way, what I’d like to know is whether the platform used for submission predicts, even weakly, proposal acceptance, and whether the proportion of platforms for accepted proposals predicts future submission platforms.)