Beta
×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

Genetic Algorithm Helps Identify Criminals

ScuttleMonkey posted more than 4 years ago | from the playing-to-your-strengths dept.

Software 84

Ponca City, We love you writes to tell us that a new software approach to police sketch artists is finding surprising success in a trial run of 15 police departments in the UK and a few other sites. The software borrows principles from evolution with an interactive genetic algorithm that progressively changes as witnesses try to remember specific details. Current field trials are reporting an increase in successful identification by as much as double conventional methods. A short video with a few working shots of the new "EFIT-V" system is also available on YouTube. "[Researcher Christopher Solomon]'s software generates its own faces that progressively evolve to match the witness' memories. The witness starts with a general description such as 'I remember a young white male with dark hair.' Nine different computer-generated faces that roughly fit the description are generated, and the witness identifies the best and worst matches. The software uses the best fit as a template to automatically generate nine new faces with slightly tweaked features, based on what it learned from the rejected faces. 'Over a number of generations, the computer can learn what face you're looking for,' says Solomon. The mathematics underlying the software is borrowed from Solomon's experience using optics to image turbulence in the atmosphere in the 1990s."

cancel ×

84 comments

Sorry! There are no comments related to the filter you selected.

What's genetic about that? (0, Offtopic)

mhajicek (1582795) | more than 4 years ago | (#30274056)

I understand it's an evolutionary algorithm, but it has nothing to do with DNA.

Re:What's genetic about that? (4, Informative)

Cyberax (705495) | more than 4 years ago | (#30274162)

Uhm....

http://en.wikipedia.org/wiki/Genetic_algorithm [wikipedia.org] ?

The word 'genetic' itself has nothing to do with DNA.

Re:What's genetic about that? (1)

CorporateSuit (1319461) | more than 4 years ago | (#30274708)

The word 'genetic' itself has nothing to do with DNA.

Yes it does. The usage of it, in this case, does not.

Re:What's genetic about that? (2, Informative)

Cyberax (705495) | more than 4 years ago | (#30275200)

The word 'genetic' predates the discovery of DNA: http://www.etymonline.com/index.php?term=genetic [etymonline.com]

Re:What's genetic about that? (0)

Anonymous Coward | more than 4 years ago | (#30280022)

pwnd!

Re:What's genetic about that? (1)

ascari (1400977) | more than 4 years ago | (#30277658)

Indeed, the word genetic is generic!

Re:What's genetic about that? (1)

Kjella (173770) | more than 4 years ago | (#30274280)

From wikipedia:

A genetic algorithm (GA) is a search technique used in computing to find exact or approximate solutions to optimization and search problems. Genetic algorithms are categorized as global search heuristics. Genetic algorithms are a particular class of evolutionary algorithms (EA) that use techniques inspired by evolutionary biology such as inheritance, mutation, selection, and crossover.

I'm guessing the main feature here is crossover. Basically you have the eyebrows that look like the suspect in one, the jawline in another and it'll probably show you a crossover with both. With a little bit of memory of what features you've already voted for and against in the past you can "assemble" a face by parts adjusting the different features as you see them coming together. It doesn't really sound too complex to me, but making it actually work is a lot of effort.

Impossible. (0, Troll)

Luke727 (547923) | more than 4 years ago | (#30274104)

Most criminals are young black men who all look the same.

GA vs. Hillclimbing (4, Informative)

jockeys (753885) | more than 4 years ago | (#30274176)

it seems to me that if you pick the best face from each "generation" and then randomly modify it and pick the best from the next generation, you are merely hillclimbing:
http://en.wikipedia.org/wiki/Hill_climbing and not using a proper GA. This seems to be something that the EigenFit package does.

TFA says that up to six faces may be "bred" together resulting in a new generation, which would indeed be genetic, so the EvoFit package seems to be genuinely genetic.

TFA is unsurprisingly short on details, but it seems to me that EigenFit is using hillclimbing (at least partially) while EvoFit is using shotgun-genetic.

Re:GA vs. Hillclimbing (2, Interesting)

Lemmeoutada Collecti (588075) | more than 4 years ago | (#30274316)

I also have to wonder how they are accounting for selection bias, where the witness selects the face that appears most like their internal image of the "bad guy" rather than the one closest to the actual suspect. I recall reading some studies a while back where they found that most witnesses are not that reliable when it comes to things like facial details.

Also, are they accounting for racial variances, such as the word white being used for anyone of light skin type?

Re:GA vs. Hillclimbing (1)

Brad Mace (624801) | more than 4 years ago | (#30274602)

I also have to wonder how they are accounting for selection bias

are they accounting for racial variances

I don't see how either of those issues are affected by this new technique. It's something the police have had to deal with long before this, so I assume they'll continue to do so.

Re:GA vs. Hillclimbing (2, Insightful)

Zerth (26112) | more than 4 years ago | (#30275462)

Instead of a sketch artist listening to a description and modifying based on feedback, the system will be "prompting" the witness.

Prompting has been shown to cause false memories of details, so I imagine it will be even worse when you consider the "the computer generated this, it must be right" phenomenon.

Re:GA vs. Hillclimbing (5, Informative)

eh2o (471262) | more than 4 years ago | (#30276578)

This method could be modified to avoid the prompting problem. Essentially the entire test can be buried in noise (i.e., random faces) so that the subject is never aware of the convergence process. It should also be possible to modify it to detect when the subject has insufficient information to identify the target. These sorts of techniques are quite common in experimental psychology when you need to suppress adaptation effects or do testing for medical purposes where the subject can't be trusted to be truthful.

Re:GA vs. Hillclimbing (1)

epine (68316) | more than 4 years ago | (#30281344)

Instead of a sketch artist listening to a description and modifying based on feedback, the system will be "prompting" the witness.

Many years ago, in a part of town I don't normally visit (touristy) in pursuit of a mother's day present, I ended up instead in pursuit of a man who assaulted a woman in an ocean-front parking lot.

I was standing there with my credit card about to purchase some crafty pottery thing, and then we heard a woman scream from across the road. The clerk, an attractive blond with big hair, spots this man emerging from a downward sloped parking lot, churning his short little legs toward the warehouse and trendy restaurant district like he's feeling guilty about something.

She screams at me, "stop him!" I guess I'm suggestible. Some almond shaped brain nugget takes control: "breasty blond, at your command". Off I go, pottery transaction dangling.

Now it happens I have Dustin Penner's body dimensions, as well as his (former) work ethic (not so good), and (former) fondness for wheat-based beverages. So I'm built like this Manitoba Mennonite, minus the muscles and masculinity. Intimidating--if you're afraid of hulking shadows. This was back in my unabomber, motorcycle makeover phase, with hockey hair and a bad shave. Hey, I'll be like Ted, that will get me the girls. Really, I'm still nerd to the core.

Anyway I must have learned something playing so much Quake: if you're in relentless, aggressive pursuit, the other player doesn't have a lot of time to figure out you've got no skills (or, in real life, no manhood). Usually when I played Quake II, I had the highest death total, but if the invincible monkey no one else could hit had only died once, I was guy who did it. Pistols at dawn was my primary style. I rarely got off the first shot, but if your first shot missed me, you were in trouble. Not that I was slow, far from it, but I could never bring myself to lay off on "damn the torpedoes". And since I died so much, I was often pursuing with knives or hand blasters. With any self-control, I might have been good, but that's not why I played.

So I'm off in pursuit of this low-life, armed mostly with the other fellow's ignorance about my wimpitude, and a motorcycle helmet swinging madly in my free hand, tight inside grip on the mouth guard with my dominant hand.

Unfortunately, the laws of physics dictate that this is a foot race I'm not going to lose. Big guys with giant loping strides often look slow, "as if the wind is blowing you in the right direction" (as Lowetide once wrote about Penner). This is an illusion. Giant loping strides quickly overtake furious chop. It's mid afternoon in a downtown tourist district. There's no one around. The guy turns up a deserted street, that two hours later would be restaurant central, ten witnesses to every parked car.

Perp has by now also figured out the law of invisible breezes, so he wheels around to confront me. We both stop, about 8 feet apart. He steps toward me. I step back with my right leg, keeping the Marushin in my swinging arm, tucked behind my hip. Damn, I don't want to crack the dumb thing on some low-life's noggin. I'm now standing there like Dan Marino protecting the football, about to go yard with my skid lid, waiting to see what this bulldog with short little legs is going to do next. He stops again, decides he doesn't want to fight through the first blow. Yet.

The problem for me taking the guy down (which I'm not man enough to do, in any case) is that I haven't actually seen the guy do anything wrong. We're three blocks from where the chase started, it's a little late to invoke "heat of the moment" assaulting someone for beating a rapid retreat. Maybe he doesn't like the sound of a woman screaming. This is Canada. We don't do heat of the moment. You get busted for that.

He quickly figures out I'm not going to initiate, says something inane I don't recall, then buggers off. I let him go. I got a great look at him. Too bad I wasn't wearing my glasses. Too bad about 10,000 hours spent reading books late at night.

I saunter back to the scene of the crime. Cop car shows up seconds later. Cop gets out of passenger side, rushes up to distressed woman (leopard tights, a little heavy on the bling) and wraps his arms around her in a giant embrace. Wow, I'm thinking, this is more compassion from the men in blue than I've been told about. Turns out, the leopard unit is the man's wife. The cop shop was only five minutes away, other side of china town. He was probably off duty, chewing the fat, then gets a call from his rattled spouse, causing him a testosterone surge that took ten years off his life. Ten minutes later, there's not a cop car in town doing more than ten miles an hour. There's a cop car crawling up every street in every direction in super slow motion. The hive is buzzing like I've never seen.

And I'm the only witness.

Two days later, I'm sitting in front of the book of noses. The chief detective has actually driven me to a bigger city with a first rate police artist. By the law of averages, the fellow I'm trying to describe is now a black man. For every white nose (cute button nose to angry beak), there are at least ten black noses, of every conceivable dimension: wide, wider, and widest. Until then, I'd never given black noses a moment's thought. Other facial features, not so unbalanced. The rest of him was only half black, a quarter Asian/hispanic, a quarter white.

So I make up my initial Mr White Potato Head from the less expansive buffet while the artist composes. The thing is, witnesses suck, and I was no exception. This was later explained to me: under stress, observational memory is suppressed. Your mind is concerned with something else: What is the character of threat? Is he packing something? If you spring a leak, are you going to bleed to death before help arrives? A zillion minor considerations like that, firing in a massive parallel shit storm of "what next?"

And that's what you actually remember. How did you interpreted the character of the person you've confronted? Which ADD future psychopath from your grade eight school year does the fellow most closely resemble? Does he even care about saving his own skin? Did he make it through school? How desperate is he not to get caught? Is he going to take your face to the local mob's composite artist and lay bounty on your ass?

With every sketch the artist drew, I knew instantly: no, that's wrong, if he looked like that, I would have thought he dropped out of school in grade three. This perp actually looked like a guy who drops out in grade eleven, not stupid, gets a job as mechanic's apprentice, maybe had some bad friends, some emotional management issues that sabotage his relationships, gets depressed, gets into drugs, not yet desperate enough to do something about it, but probably will, unless things turn from bad to worse, which he's presently bent on achieving.

So we kept sketching the guy until that's what I saw. But I couldn't really describe his features. As a witness, I sucked.

Part of the drill is they try to sway you, by suggesting things that aren't right, to see how you'll behave on the witness stand. Not going to work on me. I don't do sway. I have the tenacity of Hans Reiser. (Apparently, this works better for a witness than a defendant.) After stopping at a strip bar for lunch on the way home (this particular cop was a major cooch hound) he told me he was drooling to get me on the stand. Glandular fellow. His observation was that after each suggestion to sway me, I would pause for two seconds, furrow my temples, consider it deeply, shake my head dismissively, then flat out: no, that's not right, it was the way I told you. Cool. Manhood. He also told me, "How come with a lid in hand, you didn't take the guy down?" Not cool. No manhood.

It turns out the actual assault was that the policeman's spouse, in her leopard tights, was unlocking the driver side door, when the lowlife came up from behind and tried to push her into the car. She resisted and screamed. He chickened out. Never found out whether it was an attempted purse snatching, sexual assault, or kidnapping/badly conceived retaliation against the men in blue, which does sometimes happen.

I came away from the process with a fair amount of respect for the police artist. Maybe they can sway, if they wish to, but my experience was good.

Re:GA vs. Hillclimbing (1)

vlm (69642) | more than 4 years ago | (#30275264)

TFA says that up to six faces may be "bred" together resulting in a new generation, which would indeed be genetic, so the EvoFit package seems to be genuinely genetic.

Worst case scenario, by breeding faces together, they may only mean six simultaneous hillclimber algorithms, one for the chin, one for the eyes, snout, eyebrows, cheeks, lips, completely independent hillclimbers, one for each region...

Re:GA vs. Hillclimbing (1)

jockeys (753885) | more than 4 years ago | (#30275570)

that's a very good point, it didn't cross my mind that they could be doing localized hillclimbs, but that's definitely very possible. good point indeed.

Re:GA vs. Hillclimbing (1)

ascari (1400977) | more than 4 years ago | (#30277722)

After watching the simulation of six hillclimbers "breeding" for the fourth time the witness cried "Yes, it's him! It's him! Just get me out of here!"

Re:GA vs. Hillclimbing (1)

Tablizer (95088) | more than 4 years ago | (#30277262)

it seems to me that if you pick the best face from each "generation" and then randomly modify it and pick the best from the next generation, you are merely hillclimbing

How is natural selection *not* hill-climbing? I agree that the hopped valleys may be much deeper in real evolution, but it's merely a matter of degree, not existence of.
       

Re:GA vs. Hillclimbing (1)

jockeys (753885) | more than 4 years ago | (#30277448)

How is natural selection *not* hill-climbing?

if you are not "breeding" more than one species and adding a bit of randomizing, it's not a proper genetic algorithm.

by that same logic, irl asexual creatures DO reproduce and evolve by hillclimbing (with an added random factor for mutation). saying something is a genetic algorithm has specific connotations, I fully agree that the metaphor is not always consistent with its biological origins.

Re:GA vs. Hillclimbing (1)

Tablizer (95088) | more than 4 years ago | (#30277726)

They *are* randomizing it according to the original post I quoted. If you mean there's no chromosome cross-snipping to emulate sexual reproduction, my understanding is that GA's don't require that to be called GA's. It's merely an additional technique to add on to the mix if desired. (cross-snipping's usefulness depends on the subject matter, according to what I've read.)

History Lesson? (0)

Deflagro (187160) | more than 4 years ago | (#30274178)

Didn't they used to do this like 100 yrs ago? "See he looks like an animal, therefore he must be criminal" I vaguely remember seeing something along the lines of that being a prosecution's argument and it being accepted.

It's funny that they may have been onto something? :P

Re:History Lesson? (1)

Deflagro (187160) | more than 4 years ago | (#30274232)

Wow I totally went Slashdot on that and didn't RTFA. -2 to INT.

Re:History Lesson? (1)

snl2587 (1177409) | more than 4 years ago | (#30274282)

Yeah, you look like you might've done that.

Re:History Lesson? (0)

Anonymous Coward | more than 4 years ago | (#30274302)

+2 to WIS

Re:History Lesson? (0)

Anonymous Coward | more than 4 years ago | (#30276354)

+1 to CHA

Re:History Lesson? (2, Funny)

Eravau (12435) | more than 4 years ago | (#30274342)

Wow I totally went Slashdot on that and didn't RTFA.

You didn't even RTFSummary... let alone RTFA.

You're going to want to read the article (1)

Wrexs0ul (515885) | more than 4 years ago | (#30274268)

This is about evolving an image from generic features to a specific person by having the viewer rate a series of generated faces from best-to-worst matching.

Not that this is any better. At best it's leading a witness because it promotes guessing, at worst I feed source imagery of stereotypical "bad guys" and voila: every Snidley Whiplash lookalike in the country is running for the hills.

-Matt

Re:History Lesson? (1)

TheThiefMaster (992038) | more than 4 years ago | (#30274324)

Um, no?

This is about an eyewitness trying to create an image of the criminal for police to track down, by choosing between multiple different images based on their last choice. Instead of the traditional approach of choosing eyes, hair, lips, nose etc individually and then assembling the final image, the witness guides the system towards the result.

It has nothing to do with deciding that someone is a criminal because of their genetics. The title could be taken to mean that, but the summary is pretty clear.

Re:History Lesson? (1)

omnichad (1198475) | more than 4 years ago | (#30274370)

At least read the summary. Helps identify as in helps create a sketch.

Re:History Lesson? (0)

Anonymous Coward | more than 4 years ago | (#30274576)

Lol. You didn't even read the summary.

Vaguely remember? (1)

flyingfsck (986395) | more than 4 years ago | (#30277096)

Damn, you are that old? No wonder that you only vaguely remember it, but my goodness, you must be the oldest Slashdotter.

The scene opens with... (3, Funny)

Anonymous Coward | more than 4 years ago | (#30274198)

...a police artist sitting at a sketch pad drawing a helical structure. He glances back at a witness sitting across the desk. After drawing two intertwined double-helices, he begins filling in base pairs like the rungs of a ladder. He draws Guanine joining a Cytosine. And just as he finishes the Adenine joining a Thiamine the witness screams "That's the guy!"

Does it swim? (4, Interesting)

bugnuts (94678) | more than 4 years ago | (#30274242)

Yes: Is it a frog?
No: Please enter the type of animal.

This article reminds me of the old Animal game, where it does a binary search for whatever type of animal you're thinking. It's been expanded to handle all types of nouns, with a 15-questions interface that is uncanny.

For another computer-generated facial reconstruction test, take a look at the mona lisa. [rogeralsing.com]

Re:Does it swim? (3, Informative)

Kozz (7764) | more than 4 years ago | (#30274842)

Algorithm wonks please correct me if I've got it all wrong, but... I believe a binary search is only performed on a sorted list of items. What you're describing sounds more like a well-trained decision tree.

In a similar manner there's pages out there in the triple-dub that ask you questions in an attempt to guess what fictional tv/movie character you're thinking of. It is trained by the very people who are "playing" the game so that at the end, if the program did not guess correctly, you can enter your answer. And provided you haven't been giving bogus data, you're helping to provide training data which makes the decision tree even stronger.

Re:Does it swim? (1)

bugnuts (94678) | more than 4 years ago | (#30275240)

I believe a binary search is only performed on a sorted list of items. What you're describing sounds more like a well-trained decision tree.

If you think of each animal being reached by a traversal of yes/no questions, you can construct a "gene" consisting of 0s and 1s, which would tell you how to reach that particular animal. In that manner, the animals are actually sorted on the gene. The binary search is exactly like the tree with < and > leaves, but is instead a question with yes/no leaves. You'd still have balancing issues with the animal tree, similar to a sorted binary tree.

I was just saying that TFA's software interface reminded me of the animal game, even though it's considerably different under the hood. If I used the term "binary search" incorrectly, I'll go turn in my geek card, although I think it can be legitimately shoe-horned into such a tree.

Re:Does it swim? (1)

glwtta (532858) | more than 4 years ago | (#30275480)

Yeah, traditionally "binary search" refers to locating an element in a sorted list, even though here you are searching a (binary) decision tree, calling it a "binary search" is confusing.

If we are talking about 20Q [wikipedia.org] , it's actually a bit more clever: it's using a neural net, rather than a decision tree, so it can work with ambiguous and contradictory data (it's yes/no/maybe/not-relevant rather than just yes/no, and can deal with intentionally wrong answers). And yeah, even the hand-held version is just plain creepy in how well it works.

Re:Does it swim? (1)

jockeys (753885) | more than 4 years ago | (#30276404)

Alsing's program is very cool, and I've had a lot of fun playing with it, but it is NOT genetic... it's hillclimbing.

Perfect tool... to throw investigators off (1)

madfilipino (557839) | more than 4 years ago | (#30274340)

This is the perfect tool to throw someone off.

Commit a crime but "become a victim". Falsely describe who you saw and bam, they're in jail and you are "free".

"Yes ossifer, the man that robbed that liquor store was a black man, about 6-1/2 feet tall, dingy yellow/blonde short hair, lots of tattoos on his body, earrings. Just looking at him, I think he plays basketball.". A short time later after the call goes out, some cops arrest Dennis Rodman for a crime he didn't commit.

Re:Perfect tool... to throw investigators off (1, Funny)

Anonymous Coward | more than 4 years ago | (#30274420)

Today, still wanted by the government, he survives as a soldier of fortune.

Re:Perfect tool... to throw investigators off (1)

Sir_Lewk (967686) | more than 4 years ago | (#30274634)

And this is different from any other sort of eye witness accounts how exactly?

Re:Perfect tool... to throw investigators off (1)

madfilipino (557839) | more than 4 years ago | (#30274748)

Unlike now where the "victim" needs to think to remember what the "suspect" looks like, with this, it'll be easier to describe the "suspect" with an "instant" response (photo); just keep spewing crap till the new picture matches the first picture.

How valid does it turn out to be? (2, Insightful)

PatHMV (701344) | more than 4 years ago | (#30274386)

There's a fair amount of research on the performance of memory and how our recall of events and things is affected by the very act of being questioned about and actively recalling those memories. Before I relied on this for much of anything, I'd want to see some pretty well controlled studies on just how accurate it is. For example, they should put the test subjects under some kind of stress, have them look at the person they will have to describe and have sketched, then put them in front of the software (do a control group using traditional sketch-artist techniques, while you're at it. You should be able to do an objective evaluation of the accuracy of the sketch by mathematically comparing it (using existing algorithms developed for facial recognition) to determine just how close the resemblance is.

You got that right. (4, Insightful)

NoYob (1630681) | more than 4 years ago | (#30274560)

I was thinking of a test case for this - the picture of Solomon didn't impress me one bit. Now, you can't have folks mugging test subjects or other violent things BUT there is way.

The test case:

Get a group of test subjects - college students are always great for this. Have your "assailant" run up to the subject and Yell, "Hi!" and then hand the "victim" a flower and then run off. Right then and there, the "victim" goes a "files a police report" with the researchers following typical police procedure.

After about a thousand tests on different subjects with statistically significant positive results, then and only then, will I start to believe this "technology" and maybe with more tests will I think it should be allowed as evidence in a court of law.

Other than that it just a gimmick - we're talking about taking people's freedom here or sentencing them to death.

Re:You got that right. (1)

bws111 (1216812) | more than 4 years ago | (#30275012)

What are you talking about? This is no different than a regular old police sketch artist, just quicker. I highly doubt anyone has been convicted based solely on a sketch. This just provides the police with an image of someone they should be looking for. Once they find that person they can interview him, get an ID from witnesses, obtain fingerprints, get a search warrant, and all the other stuff that goes along with a criminal investigation. The sketch will never get to court because it is not evidence, it is just a tool.

Re:You got that right. (1)

NoYob (1630681) | more than 4 years ago | (#30275166)

That's how it works on TV.

Finger prints - another pseudo scientific technique used by the cops.

Look it, every time I read a story about someone who was wrongfully convicted and then spent years in jail, in every single case there was one piece of evidence - usually some sort of eye witness - that put the poor bastard in jail. Now factor in a computer? You know how the ignorant public (the ones usually sitting on a jury panel) trusts anything the "computer" says - especially with all those fictional accounts on TV with those CSI shows. Unfortunately, reality isn't that precise.

I guarantee you, if this goes live, it will have a near 100% conviction rate regardless of its accuracy.

Re:You got that right. (1)

mhajicek (1582795) | more than 4 years ago | (#30275268)

It will be even easier to convince a jury to convict someone when you show a CGI "photo" that looks like him. Lots of people trust computer output and ignore the human link.

Re:You got that right. (0)

Anonymous Coward | more than 4 years ago | (#30276236)

That's because there is no man behind the curtain. We didn't know it in 1900 [wikipedia.org] , and we still don't know it now.

- Pitabred

Re:You got that right. (1)

millennial (830897) | more than 4 years ago | (#30276202)

The problem is that people's memories are not permanent. If you show them a bunch of different (but similar) images of people, it may well completely change what they remember about the suspect. A police sketch artist doesn't show any faces until they're done.

Re:You got that right. (1)

swillden (191260) | more than 4 years ago | (#30276840)

Once they find that person they can interview him, get an ID from witnesses, obtain fingerprints, get a search warrant, and all the other stuff that goes along with a criminal investigation.

It's not that uncommon for someone to be convicted on nothing more than a witness identification, perhaps with a little weak circumstantial evidence along for the ride. The risk of something like this (and the risk also exists with police sketches) is that a vague memory will be "firmed up" by viewing the completed sketch. The police then find someone that looks like the sketch, but who maybe only looked vaguely like the actual criminal and put them in a lineup. Based on the original memory, perhaps the witness wouldn't be able to identify the suspect with much assurance, but having had their memory "strengthened" by looking at the sketch, however generated, they then become quite sure that the individual who looks like the sketch is the one they saw.

None of this is inherently bad, mind you. It often does find the right person, and even if it doesn't that's still more or less okay as long as the defense attorney does a good job of exploring this memory issue with the jury, to help them give the witness ID the weight it actually deserves. If the defense attorney sucks, though, this process can easily send the wrong man to prison. And even if the defender is good and the jury returns a not-guilty verdict it still sucks to have had to be tried for a crime you didn't commit, but what can we do?

Note that I'm not saying I think this technology changes the situation much. It's possible that the evolving sketch may allow the witness to settle on an image that is further from the criminal's real appearance than would be achieved with a police sketch artist. Or perhaps the evolution does a better job. Testing is needed to know for sure. My point is that the whole process, however done, tends to pollute the witness' already-fallible memory and is of less value than might be thought.

Re:You got that right. (1)

LUH 3418 (1429407) | more than 4 years ago | (#30275632)

There is at least one reason why this might work better than a regular police sketch. People's memories are bad, but I believe people have an easier time looking at something and identifying it than recalling specific details about that something on demand. Case in point: you may not be able to exactly remember the melody or the lyrics of a song you head only once, but if that song is played, you will be able to tell you heard it before, and it seems likely that you would be able to tell if the melody wasn't the same as before. If you had to pick between a few altered versions of that same song and the original song, you could probably tell which one you listened to before. It's similar for faces. We think we remember people's faces well because if we see someone we've met before, we can recognize them. However, if you met someone only once, and tried to draw their face the next day, you would probably have a hard time recalling what they looked like.

Re:You got that right. (0)

Anonymous Coward | more than 4 years ago | (#30276188)

It's still dicey. We don't remember exact things. Our brain flattens them into indexed patterns, using lossy compression.

The brain also sees patterns where none are present. One example is shapes in the clouds.

I've noticed that once you identify a song, you mind fills in the gaps, increasing the apparent fidelity so you seem to hear it much clearer than you really are.

If the song is a slight variation, and that the variation is beneath your sensory threshhold, you'll fill the gaps to the version you're familiar with and *swear* that's the version you heard. And you'd be right. Because consciously hearing is something happens after the brain has processed the sensory data and combined it with recall.

The problem would be that the version you heard is not the same as the version that played.

I've seen this happen with visual queues too. You can try any number of tests where letters in words are are transposed yet the people read (and remember) the "correct" words, not the ones that were written.

We don't have anything better than "sketch artists" but it does setup a feedback loop. Sometimes the process converges on the correct answer. Sometimes it converges on something else. Which is why we have a court system. Arguments are made, evidence is weighed, and the judge or jury settles the "facts". There is simply no better way to do it.

I agree with something someone said earlier. Sure, this is really just a Computer Aided version of something a sketch artist already does. But the public is much more trusting of computers than people. They would tend to weight this as more.

On the other hand, I wouldn't think police artist evidence, CG or otherwise, is allowed as "evidence" in a court proceeding. It shouldn't be. I could see someone arrested on that evidence. But not convicted.

Re:How valid does it turn out to be? (1)

vlm (69642) | more than 4 years ago | (#30275482)

Before I relied on this for much of anything, I'd want to see some pretty well controlled studies on just how accurate it is. For example, they should put the test subjects under some kind of stress,

That is an interesting, but tedious way to test the overall system on average across all users. I think I could test the hillclimber algorithm itself, on an individual user, by making the algorithm converge intentionally slowly, simulated annealing style. Then pester the person with lots of very slowly converging and even occasionally diverging sets of faces, and see how "consistent" the persons answers are compared to their final answer. Does the user always select the bushy eyebrows, or only about 50% of the time, etc? Perhaps even assign a numerical or verbal value to their consistency as part of the final report each time the system is used.

How do you measure success? (2, Insightful)

Hatta (162192) | more than 4 years ago | (#30274394)

How do they know if this thing actually works? If they're using the computer generated sketch to finger a suspect, and then presenting that sketch as evidence to a jury who convicts, and then using that conviction as evidence of the algorithms accuracy that's just circular reasoning.

The memory is not an immutable thing. It's quite possible that in the process of generating the sketch you are leading the witness on, even implanting memories. So what happens if you generate a sketch that doesn't look like the actual criminal, and present that to a jury and get a conviction. Is that going to be counted as a success?

Well - mine was a failure (1)

abbynormal brain (1637419) | more than 4 years ago | (#30274522)

My sketch ended up as a pink elephant ... go figure!

Re:How do you measure success? (2, Insightful)

jonbryce (703250) | more than 4 years ago | (#30274584)

They won't present the sketch as evidence to the jury. They will call the witness and ask him to identify the suspect. They will be able to do other things like take fingerprints and DNA samples from the scene and match them to the suspect.

Re:How do you measure success? (1)

swillden (191260) | more than 4 years ago | (#30276894)

The memory is not an immutable thing. It's quite possible that in the process of generating the sketch you are leading the witness on, even implanting memories. So what happens if you generate a sketch that doesn't look like the actual criminal, and present that to a jury and get a conviction. Is that going to be counted as a success?

They won't present the sketch as evidence to the jury. They will call the witness and ask him to identify the suspect.

... based on memories that have been partially implanted/modified by the sketching process.

There is some risk here.

They will be able to do other things like take fingerprints and DNA samples from the scene and match them to the suspect.

Assuming there is other evidence, great. If not... people do get convicted on the word of a single eyewitness.

Re:How do you measure success? (1)

swillden (191260) | more than 4 years ago | (#30276972)

There is some risk here

To be clear, I'm not claiming that the risk is greater than the risk of having a police sketch artist do the same thing. It may be greater, less or the same. It may even depend on the witness.

Re:How do you measure success? (0)

Anonymous Coward | more than 4 years ago | (#30277746)

There is some risk here

To be clear, I'm not claiming that the risk is greater than the risk of having a police sketch artist do the same thing. It may be greater, less or the same. It may even depend on the witness.

I agree. In some cases it will help because the software can be neutral where a sketch artist might have ulterior motives or predisposed to bias. Of course tampering with the software (or intentional design) could be used to tamper with memory more aggressively.

So really all this system does is to put some sketch artists out of a job.

Re:How do you measure success? (1)

CraftyJack (1031736) | more than 4 years ago | (#30275214)

How do they know if this thing actually works?

Especially if no one has actually *seen* Keyser Soze. And like that, poof. He's gone.

Re:How do you measure success? (1)

Tablizer (95088) | more than 4 years ago | (#30277328)

I imagine they use test cases, not (just) actual cases. For example, 40 people see the same candidate in person. Then the group is split up into 2 groups of 20. One group does the traditional method and the other does the GA method to derive 40 sketches total. Then an independent panel of judges who don't know about the GA software rate the sketch matches to the original candidate, and the total for each group's sketches is compared.

It's about time for GP (2, Interesting)

Kingrames (858416) | more than 4 years ago | (#30274502)

I got the opportunity to do a genetic algorithm at my university for one of my projects, and I'm surprised that only now is this tech becoming slightly popular.

You take a fistful of bad answers to a problem, throw 'em in a breeding pit, and let 'em go at it.
you essentially breathe life into binary data, becoming a God, and allowing 'your people' to evolve into a solution to your problem.
I suppose you could call yourself an 'Intelligent Designer', but that lacks panache.

Re:It's about time for GP (2, Interesting)

glwtta (532858) | more than 4 years ago | (#30275676)

Just to be pedantic for a second: Genetic Programming (GP) is a specific application of Genetic Algorithms (GA) where the solution space you are working with is executable programs (or algorithms). So GP is a subset of GA, the two are not interchangeable.

To answer your question, GA is not more popular because for most real-world problems it's difficult to come up with a good solution representation (one that lends itself well to "breeding"). Though they have been used successfully for a long time in several different niches.

Re:It's about time for GP (1)

Kingrames (858416) | more than 4 years ago | (#30275902)

being a programmer, I interchange the two all the time because of the overlap. Of course gp isn't ga, in the same sense that doing a traveling salesman problem by hand doesn't count as programming an answer, even if you are programmatically solving it.

and I am eager to begin a project I'd like to start which would make gp easier for anyone that wanted to pick it up and run with it. I think gp can be used in many more applications than it is now, and I'd like to prove it.

besides, wouldn't it be great if rampant sex could be used to solve all your problems?

Re:It's about time for GP (1)

glwtta (532858) | more than 4 years ago | (#30276542)

and I am eager to begin a project I'd like to start which would make gp easier for anyone that wanted to pick it up and run with it

Well, good luck with that, though I still suspect that GA (and especially GP) is inherently rather hard, meaning you are unlikely to find a technical solution that will substitute for thorough understanding of the problem domain and lots of practice.

It's not like there aren't already plenty of frameworks [wikipedia.org] that make it next to trivial to get a basic implementation up and running.

Re:It's about time for GP (1)

Kingrames (858416) | more than 4 years ago | (#30278598)

Well, that's true, but I think GP is a stepping stone to another kind of evolutionary programming style.

At some point I'd like to design and run an algorithm that uses multiple children at once as a team, instead of just testing each one individually. Sort of like a raid, in WoW, except that the 'characters' would be selected randomly. I think this might be the next step up for genetic programming, because the high-value children would be better performers across multiple groups, not just alone, and it would eventually develop into a system that creates highly specialized individual components, and might discover new truths inside the approach.

I think I'd call it Raid Programming, for lack of a better term. My project will probably be called "sup", short for "Supreme Ultimate Power."

Re:It's about time for GP (1)

glwtta (532858) | more than 4 years ago | (#30279832)

I may be wrong, probably am, but it sounds a little like you are just redefining existing terms with more awesomer names.

Your "raid group" is just a different way of describing the solution representation. There is no reason why a candidate solution can't be subdivided into parts that are tuned individually. You're still testing the fitness of the complete solution - I'm not seeing how that's different from classical GP.

Re:It's about time for GP (1)

randomsearch (1207102) | more than 4 years ago | (#30281230)

> At some point I'd like to design and run an algorithm that uses multiple children at once as a team, instead of just testing each one individually. Sort of like a raid, in WoW, except that the 'characters' would be selected randomly. I think this might be the next step up for genetic programming, because the high-value children would be better performers across multiple groups, not just alone, and it would eventually develop into a system that creates highly specialized individual components, and might discover new truths inside the approach.

Sounds like you're trying to design emergent behaviour amongst agents. A very well-established field, and a difficult thing to do. One approach is indeed to using evolutionary computation (Genetic Algorithms, Genetic Programming, Evolutionary Strategies etc.) to program the agents.

Lots of papers around on designing emergence, evolving agent behaviour etc. via Google.

RS

Re:It's about time for GP (1)

randomsearch (1207102) | more than 4 years ago | (#30281280)

> Just to be pedantic for a second: Genetic Programming (GP) is a specific application of Genetic Algorithms (GA) where the solution space you are working with is executable programs (or algorithms). So GP is a subset of GA, the two are not interchangeable.

If we're going to be pedantic, then GP is not necessarily a subset of GAs. It depends on how you define a GA: usually a GA is an algorithm that operates on binary strings, and GP is an algorithm that most often operates on expression trees. However, in truth the definition of both is so vague that trying to draw distinctions between them is meaningless. Often the GA crowd will claim GP is a specialised variant of a GA, and the GP people will claim they can simulate any GA with a tree, so GAs are a subset of GP.

They both share much in common, originating in related research communities and major conferences such as GECCO, CEC, etc. incorporate work on both.

> To answer your question, GA is not more popular because for most real-world problems it's difficult to come up with a good solution representation (one that lends itself well to "breeding"). Though they have been used successfully for a long time in several different niches.

The truth is, GAs _are_ popular, but indeed for niche applications. Academic research on evolutionary algorithms is still quite intensive, but industrial applications usually lag behind and a lot of work isn't "headline grabbing". This work for instance, is quite a simple and effective application of an interactive GA, but it could have been implemented in the 1970s from an EC point of view (apologies to my friend, who works on one of these projects :-D).

You find most applications in the area of operations research. Slashdotters short on cash might be interested to know that a lot of time and money has been spent investigating EC for financial investment strategies, and indeed these are applied and used to make money. Any such application is going to be kept quiet, though :).

Incidentally, if you're interested in GAs check out Estimation of Distribution Algorithms, which are a very elegant idea indeed.

RS

Re:It's about time for GP (0)

Anonymous Coward | more than 4 years ago | (#30276604)

Genetic Algorithms are for situations where you don't otherwise know how to optimize, because the problem is just too complex. And when that's the situation, then great, use it. But a lot of the time, people have a good handle on things -- or at least think they do. I see GA as an solution of last resort. If you have to use GA, maybe you don't know your problem well enough. Or maybe you're really working on something difficult, and let's face it: that's a situation most programmers aren't in.

Re:It's about time for GP (2, Insightful)

Delkster (820935) | more than 4 years ago | (#30276908)

I think GAs have definitely had a time when they were popular at least as an idea, mostly sometime in the early 90's or so, and there was quite a bit of research into applying them to various problems. They haven't always turned out to perform very well, though. Quite a few attempts have been made towards using GAs as a heuristic to traditional NP-hard combinatorial problems, for example, and while there has been some success, quite often other heuristics have beaten GAs.

My impression of the beauty of GAs in general isn't quite as positive as yours. The idea certainly is aesthetically pleasing, and you can, in theory, try to apply a GA to pretty much any optimization problem, but how well GAs work really depends a lot on the problem: the very nature of the problem (does it fulfill the building block hypothesis, or whatever magic is that makes GAs work for some problems?), what kind of a landscape the search space provides, what kinds of cases of the problem are more likely in your application, etc. That's not including all the nontrivial problem-specific tweaking that will be needed in a practical application of a GA, such as how to encode or represent the solutions (has a big effect on how much good genetic crossover does).

I'd rather say that GAs have worked well for some specific problems, and some new specific applications will probably still emerge, but I'm not sure they will ever become very generally applicable. They had a chance, but it turned out that they mostly work just for some particular problems, not others, and nobody seems to really know very well why.

Re:It's about time for GP (1)

Tablizer (95088) | more than 4 years ago | (#30277382)

essentially breathe life into binary data, becoming a God, and allowing 'your people' to evolve into a solution to your problem.

Sergeant: "Odd, why do all the people we arrest look like Jesus?"
   

This is actually very cool... (3, Insightful)

jarrowwx (775068) | more than 4 years ago | (#30274526)

This technology, at its core, is a little bit like PicBreeder [slashdot.org] . It doesn't include the complexification, but the principle is the same.

There is an argument about 'leading the witness' being bandied about as if that makes this thing useless. If you read the articles, they talk about that, and they show that it is no worse than any existing techniques, gets good results, and works for people that can't work with sketch artists.

The reality is, this technology has applications beyond what it is being used for.

  • Imagine, a site that you can go to and evolve the face of the woman of your dreams?
  • Or the face of a character in the book you are writing.
  • Or an avatar for the video game you are playing.
  • Or use the basic tech to create random faces for the crowd for an animated movie.

Personally, I would *LOVE* to be able to tinker with technology like this.

Re:This is actually very cool... (1)

Kingrames (858416) | more than 4 years ago | (#30275098)

but you already can!

get yourself some tools for working in the programming languages of your choice. I recommend 'Go' and 'D', since you'll be playing God.

then, just have fun with smaller projects, like figuring out the best way to fill a rectangle with circles, before tackling really complex problems. Genetic algorithms tend to 'cheat', so it's really best to just experiment with them, rather than reading info from a book.

Re:This is actually very cool... (1)

memristance (1285036) | more than 4 years ago | (#30277964)

figuring out the best way to fill a rectangle with circles

I don't know about best, but I think Mandelbrot has already figured out the prettiest.

Re:This is actually very cool... (1)

vlm (69642) | more than 4 years ago | (#30275600)

Imagine, a site that you can go to and evolve the face of the woman of your dreams?

Just the face? Thats all? Try again.

Your project has already been done, anyway.

http://en.wikipedia.org/wiki/Averageness [wikipedia.org]

Weirdly enough, the difference between the 1.0 and 10.0 womens face seems to be little more than body fat percentage. Actually, IRL for the whole body, isn't the difference between 1.0 and 10.0 little more than body fat percentage?

Finally... (1)

ground.zero.612 (1563557) | more than 4 years ago | (#30274644)

A wizard for designing a criminal suspect on the fly!

I understand the use of the term genetic in the case of this algorithm, but I can't help but wonder about obligatory Minority Report hypotheticals. In the USA your DNA is been stolen by the Government at birth; apparently they've been doing this since the 1970's. After the Feds work with it, it is then "anonymized" and sold to third parties such as medical research facilities and insurance agencies.

Intertwining these technologies leads this avid conspiracy theorist to fanciful visions of a future where one is not guilty because some mutant fortune tellers can see your future crimes, but instead because the computer simply says so.

Mii Channel (1)

FiloEleven (602040) | more than 4 years ago | (#30274912)

Nine different computer-generated faces that roughly fit the description are generated, and the witness identifies the best and worst matches. The software uses the best fit as a template to automatically generate nine new faces with slightly tweaked features, based on what it learned from the rejected faces.

I immediately thought of the Mii Channel on Wii when I read this. One of the ways to create a Mii is to start with a bunch of randomized faces and pick the one that looks the most like you (or whoever you're modeling). From there, it generates 9 variations of that face for you to choose from. This system is obviously more advanced, but the basic idea is the same.

Only 9 choices? (0)

Anonymous Coward | more than 4 years ago | (#30275044)

Nine different computer-generated faces that roughly fit the description are generated, and the witness identifies the best and worst matches. The software uses the best fit as a template to automatically generate nine new faces with slightly tweaked features, based on what it learned from the rejected faces. "Over a number of generations, the computer can learn what face you're looking for," says Solomon. The mathematics underlying the software is borrowed from Solomon's experience using optics to image turbulence in the atmosphere in the 1990s.

The wii gives me a lot more than 9 choices when I make a new Mii, and I don't have to use any math or borrowing ...

Yes, but is it better than... (1)

Smivs (1197859) | more than 4 years ago | (#30275392)

this e-fit [sky.com] that helped the Bolivian Police track down a murder suspect

Porky's (1)

t0qer (230538) | more than 4 years ago | (#30275514)

I couldn't help it, this story made me think of an epic scene from the 1982 movie porky's. In the movie a few young men are looking at girls through a peep hole in the girls locker room shower. One young man sticks his talleywhacker through the hole and almost has it torn off by the lesbian'ish PE teacher. Anyways, here's a synopsis of the following scene, courtesy of imdb.

Balbricker: Now, Mr. Carter. I know this is completely unorthodox. But I think this is the only way to find that boy. Now that penis had a mole on it - I'd recognize that penis anywhere. In spite of the juvenile snickers of some, this is a serious matter. That seducer and despoiler must be stopped; he's extremely dangerous. And, Mr. Carter, I'm certain that everyone in this room knows who that is. He's a contemptible little pervert who...
Mr. Carter: Miss Balbricker!
Balbricker: Well, I'm sorry, but I've got him now, and I'm not going to let him slip through my fingers again. Now, all I'm asking is that you give me five boys for a few minutes. The coaches can be present - Tommy Turner and any four boys you see fit to choose and we... and we... can put a stop to this menace. And it is a menace.
[pause]
Balbricker: Well, what are you gonna do about it?
Mr. Carter: Five young boys in the nude, a police line-up so that you can identify his tallywhacker. Please, please can we call it a "tallywhacker"? Penis is so ppp... penis is so personal.
Balbricker: We can put hoods over their heads to avoid embarrassment. Now listen: we have got to do it, as distasteful as it is. I know it's him. That
[pause]
Balbricker: tallywhacker had a mole on it. And that mole is the key to it.
Mr. Carter: Miss Balbricker, do you realize the difficulty of your request? Now, I would be very happy to, uh, to apprehend the young man myself. But can you imagine what the board of education would say if you were granted a line-up in order to examine their private pa... their private parts for an incriminating mole?
Balbricker: But Mr. Carter.
Coach Brakett: Mr. Carter, I think I have a way out of this. We, uh, call the police, and we have 'em send over one of their sketch artists. And Miss Balbricker can give a description. We can put up "Wanted" posters all over school..."Have you seen this prick? Report immediately to Beulah Balbricker. Do not attempt to apprehend this prick, as it is armed and dangerous. It was last seen hanging out in the girls' locker room at Angel Beach High School."

Oh great, here we go... (0)

Anonymous Coward | more than 4 years ago | (#30276574)

Welcome to the digital age of racial profiling.

EFIT-V (1)

mayhem79 (891695) | more than 4 years ago | (#30280882)

My name is Dr Matthew Maylin, I developed this software at the university of Kent during my PhD, and continue to develope it as the sole software engineer for the company. The algorithm used is an evolutionary algorithm, implementing random mutations, but no cross-over or mutation. Although the user does have the option to 'bred'/combine certain faces within the software. The method uses a statistical model of the human face (Cootes et el 2001), and at anyone time is restricted to a sample to a single ethnicity and gender. The are however, many databases of face statistcs that are used. Here is a movie of the process: http://www.visionmetric.com/images/stories/EFIT-V_demo.htm [visionmetric.com] I believe here is the original slashdot post: http://slashdot.org/articles/04/05/17/1042231.shtml?tid=133&tid=152&tid=185&tid=186 [slashdot.org] There have been many controlled studies on this system, it has been trialled by the forces in the UK over 3 years ago, and now is actively sold across the world. Psychological studies have been made by Dr Graham Pike (seen in the video) at the Open University. The quality of the images vary and are ultimately limited by the users ability to recall the face - some users are better than others - but generally composites are produced more quickly - to a higher quality - than 'jigsaw' based methods.
Check for New Comments
Slashdot Login

Need an Account?

Forgot your password?

Submission Text Formatting Tips

We support a small subset of HTML, namely these tags:

  • b
  • i
  • p
  • br
  • a
  • ol
  • ul
  • li
  • dl
  • dt
  • dd
  • em
  • strong
  • tt
  • blockquote
  • div
  • quote
  • ecode

"ecode" can be used for code snippets, for example:

<ecode>    while(1) { do_something(); } </ecode>