Transcript

the following content is provided under a Creative Commons license your support will help MIT OpenCourseWare continue to offer high quality educational resources for free to make a donation or view additional materials from hundreds of MIT courses visit MIT opencourseware at ocw.mit.edu here we have a near-miss Learning Tree it's a little bit different but a little bit similar we've got different types of we're trying to learn about different light sources so well we're sort about different lamps flashlights all sorts of things so these guys can have types of support a bass or wires the base can be flat clamped or have legs the legs can be flat bottom two-wheeled the light source can be incandescent fluorescent or sodium vapor and the energy source that powers it can be electric battery oil or gas so there's also some other things but these are the trees that we might be climbing up with the climb tree here a stick our starting model is we have an incandescent bulb the height is equal to 24 inches exactly they have a flat base there's electricity to power it and it has a shade so our starting our starting model is basically like a table lamp that you set on the table it's got you know the shade over top of it you know the standard broom table what is our starting model so recall in arch learning which is what we are using here that there are several heuristics that we might want to use let's see well the heuristics are um let's see we've got um require link forbid link we've got climb-tree extend set closed interval and we also have we also have drop link forgot about that one that's served with the link so I can put it here okay so if you want to do well on these questions if that if they appear on the final first of all you better know what those six heuristics do so basically require link is a heuristic that it goes into your model and it as a requirement where before your model just didn't care so let's say there was something about color for your light for your lamp's base or for the lamp and right now it doesn't care about color at all but let's say that you wound up having an example that was for instance blue or white or whatever these colors are and it lets it and and it wound up get it wound up getting that blue was okay for a for idiots model wound up getting okay blue and white are okay and then it or actually know it even more basic example let's say that it just kept getting blue lamps as examples and all of them were blue and it just didn't care about the fence they were blue but they did turn out to be blue and then eventually it found a red one that was exactly the same as its model but the red one was a near-miss then it might require blue which was um already an acceptable permissible thing in his model but it could be required now another thing you can do is forbid some people ask me why would you bother saying forbid read say if you could already just require blue like together they cover they cover the same ground the answer is because this system this ARS learning system remember Patrick's doctoral thesis is an old system very because it was not built on old systems and because also it's generally a good idea it was very parsimonious if you let's say you had a set of colors red blue yellow green pink purple orange and you found that only orange was giving you problems you could try to require that it be in this or you could even use extend set making to make it make a giant set of red blue yellow green purple orange but oh sorry purple pink but not orange but if you just forbid orange you've just saved them a whole bunch of space particularly if there's a large possible as soon as the set you want to have the ability to forbid because it meet well it needs to have the smallest bottle that it can that covers all of its samples so we also have climb tree climb tree takes one of the elements in our model and moves up the tree one level to be you know so let's say we only we said only incandescent lights are good but we found that fluorescent lights are also good whoop we move up two leg stores climb tree extend set we don't have any sets here but let's go back to red blue orange yellow extend set is when we say only red is good as an example but we see another positive example the yellow so they say okay I'll extend the set red and yellow or both good all right close interval closed interval is really obvious when you need to use it because not only is it only used for intervals like height equals 24 inches or some kind of numbers but it is the only thing you can use for for when I equals some kind of numbers close interval will let you fuss around with the intervals and say oh well I guess one with a 20-inch bate 20 inch height is okay too so we'll make the whole interval 20 to 24 inches good because it doesn't make any sense for it be like 20 is good 24 is go but nothing in between is good no those are awful so that closed interval covers the entire interval last but not least is drop link um the thing about drop link is that drop link is again only really used due to the fact that the system wants to be as parsimonious as possible with its model drop link is let's say that you have a color and you're like okay only red is acceptable and then you see a blue one it's go okay and let's say red blue yellow your only colors only red and blue except wait a minute no we could just stop say that only yellow is not acceptable or something like that all right but then you see yellow as a positive example actually you can't even switch to the yellow being not acceptable cuz you've only seen positive renders you say only red and blue are acceptable lesson for us as indicated in the climb tree to light sources we don't see one that's Soviet of a percent stream do we have to go back down the tree or do we just add up for bid one is we high florescent to light source when we see sodium vapour what do we do the answer is our system unlike with lattice learning in arch learning our system is memoryless of its previous examples it's incapable of going back down the tree and so if you've got a incandescent fluorescent climb up the light source your only recourse if you see the subbing vapor doesn't work is to forbid link sodium vapor that's a good question it shows good understanding of how we work so drop link is you see red and blue they both are okay you don't know about yellow so you can't just switch that to a forbidden link to yellow but then you see yellow is also okay what do you do you can drop the link altogether save yourself space you don't if all the colors that you have in the entire world are fine you just not have part of your model be the color and just say any color is fine color must not be really part of what defines a lamp and I'd say it's not I mean it's what defines a tacky lamp I guess but it's not what defines a lamp so I mean that might be why color isn't here to begin with so to make yourselves do better on this more knowledge equals less search let's have more knowledge there's only some of these are used to generalize to make our model more accepting of new examples and those are ones we would use after we've seen a positive hit sort of generalize learn more things in the model some of these are used to specialize and make our model refine it make our model more specific you'd only use those after you saw a near-miss so let's actually separate those that way you guys will never make the mistake of using one of these in the wrong situation and more knowledge less search and less search means faster quiz time so require link what do you guys think is that a specialized ER or generalizer specialized er that's right we'd only use it when we saw him you're miss forbid links specialized or a generalizer specialized er we only forbid something if we saw miss climb-tree specialize or generalizer generalize it that's right we'd only climb up to a more generic thing in the tree if we saw a positive example extend set generalizer we'd only extend to the things that are set if we saw a positive example close interval that's a generalizer we'd only make the interval bigger if we saw a if we saw a positive example someone asks what do you do if you have like 10 to 30 and then you find a negative example in 20 it's pretty it's generally pretty annoying for the system to have to deal with that there's a variety of things you could do based on your implementation I've never seen us ask it in a quiz but one thing you could do is forbid 20 exactly just 20 and just just have a little hole at at exactly 20 so then drop links specialize or generalizer generalizer this one is an easy one to mess up if you don't understand the system because dropping you think Oh getting rid of something that's special it actually no you're generalizing the entire area it's saying we can forget about it because they're all good all right so given that we are set to do this problem it we have pretty reasonable amount of time considering that we don't have to do a sobs it's great so our first example is a sort of one of those stem reading lights that you sort of have it on the stand and you can like adjust the little light bulb up or down it is a in it's an incandescent bulb it has a height of 11 inches and it's got a flat base it's electric it's got a shade and it's a hit it's a hit a positive example plus it's good so right away we know that we can only use the esses or the G's the G's we can only use the G's because it's a hit so we're not gonna be requiring we're not gonna be forbidding today we're going to be happy and using in generalizer so I'll help do the first one you guys will help me do the next one or someone can say we didn't cover this and I'll do all of them but I'll do the first one and we'll stop but so we've got an incandescent light that's the same as our model where we've got a height of 11 inches oh I'm not only covers 24 flat base that's the same as our model electricity same as our model shade same as our models so I say the only difference is the height I say we need to close the interval closed interval and I say that our model is model is same as before except for that height it is an element of 11 to 24 all right simple enough I picked the easy one for myself next one so this one is going to be a yeah this one is going to be the picture here didn't turn out but it's basically a lamp that doesn't have a shade it just sort of has the light sort of shining shining on you so it is a positive example its incandescent height equals 11.5 flat base and electricity all right so first of all which kind would we use stresses on the specializing or generalizing generalizing and particularly anyone want to take a look at our model which is the starting model except for that the height can go 11 to 24 what which heuristic do we need to use here all right so people are all saying drop-link yes that's right someone said extend set well I suppose we could extend the set to shade it or not shade but that's everything so drop-link is the answer we can drop the shade shade or no shade or not this is still pretty much a lamp so that's correct we drop link so the model alright there's been enough changes to it I think I'll write it out again the model is incandescent incandescent height equals an element from 11 to 24 and flat face an electricity good stop question so you're the same thing except for it's tough for that there's one that's the same with no shades and isn't izzy is a mere miss or something so that you need to or you're on site you're saying what if there's one that has a shade now that's a near miss and so it would probably try to forbid link shades that's sort of the question or you're right the system is that's an inconsistency in the data the system is very fragile it's very fragile too ordering in particular and um I mean you'd be surprised how awesomeness this does uh considering that it was made in like the sixties like it's pretty impressive stuff but is old we're not saying that this is a way that you should do all of your learning nowadays because it has some serious issues but um it's it's old but it's pretty damn good for what it did a newer style of this kind of learning made by me and another friend who's now working at Google is a lattice learning which has his own share of issues one of which is that since it's trying to act like a little kid at first it tries to claim that everything is okay until you show it some good negative examples but very particularly one way gets around this problem is it's not memoryless it in fact stores all the examples it's ever seen and compares and contrasts them to what it sees in the new in the new example and this allows it to say something like let's say you can teach it what can fly in lattice learning you say all right well bluej can fly it's like and and then if you ask it can a caliph I sure Kyle jumped over the moon everything can fly so that's its problem if you say actually cow that was a nursery rhyme a cow cannot in fact fly it will then say only birds can fly because it remembers that the bluej can fly and then you can eventually say well actually bats can fly uproot back and fly oh okay so chiroptera pnes can fly and burn to fly but that's it and you can give it some other examples but there are other spells of learning you're absolutely right in fact if there's a multiple choice asking you about like the strengths and weaknesses of arch learning one of the weaknesses it's very vulnerable to ordering of what you teach it it can look but think about it if our first example was one that's exactly the same as this but with a shape with a shade and it was a Miss the system would just yell at you like you're kind of being an to it to give it to things that are inconsistent you might say in the real world that happens well arts learning not great with real world messy data it just it goes ballistic it's very OCD wants everything to match up if it gets to things that are inconsistent it'll just yell at you it's like you're wrong this is this can't be true because you said it's okay with the shade so that's a very good question another question to the model and it's in dear Miss and I'm sorry to cut you off we based on timing I'm just going to say whatever you said afterwards doesn't apply due to the fact at that point it's not a near miss it's very important it's only in your miss if there's only one change otherwise it's just a miss and if it's a miss with more than one change you're probably gonna say how do you know what to change the answer is you can't change anything because any of them could have been what was wrong that's why for hit oh it can have as many differences as you want you'll just generalize them all from Miss it can only have one thing different in fact there might be one here that's not in your miss and you just say oh we don't do anything we'll keep going we'll night see it if we have time so question ordering is a factor for this your thing and you encounter a item which has two discrepancies in values but is eventually but if the ordering were different such that that would be accountable like deliver be turned back so today's lesson how you not hear this can you hold onto it and use it later when it would be in your mince the answer is nope if we're going here for a system that doesn't remember things you can put in some clutches that definitely makes it smarter it uses odds data but we want Patrick in the 60s making this up strove for elegance and the elegant solution is let's just be memory less completely the idea being well little babies don't can't tell you ever experience they had with playing with blocks with an arch so let's remember nothing with lattice learning our idea is that people sort of somewhere in the subconscious maybe do store all the examples or at least a lot more than you give them credit for so why not store them but with arch learning it's little babies can't tell you oh yeah there's at one time I played with a block or I played with a lamp and it didn't have a shade and so it wasn't a lamp or something like that they're not going to be able to store it they're not gonna be able to save it for later it really really in sort of installed Turing who believed a human would be a teacher to a computer arch learning really focuses on the fact that the human is a kind and good teacher who offers examples that are exactly appropriate at a time and that puts a lot of pressure on you as the trainer of a large learning system good questions all let's continue working this guy out so the next example is basically another one of these scan lamps it's got a light shining down which is fluorescent it looks like this it's fluorescent height equals 13 inches flat base electric shade okay and yeah as I put here it's a it's a hit so as you guys have told me countless times so I believe you remember we're going to generalize for the hit so what should we do here so some people are staying extend set the people who are saying climb tree are correct why do we climb to extend extend set because of the fact that it's a tree not a set and it is more parsimonious to climb the tree than it is to treat it as a set and extend the set sets don't have a hierarchy it's just like red comma blue color yellow or shade common not shade whenever you have a tree it's a fact that it's more parsimonious to just climb it than it is to treat it as a set and extend it even though you couldn't do that I suppose and arts learning its parsimonious its elegant it's simple and the elegant thing to do just climb up so yes we're going to climb up the tree so our model is same as last time same as last time except that this time it has a light source instead of as light stores instead of saying incandescent it just planted the lights are sure all right so the next one is pretty cool it's another one of those fluorescent shine down lamps but in space it has like a tripod with three legs and those legs all have wheels so it is also a hit so it is fluorescent height equals 14 inches we old legs and electric so obviously we're generalizing what heuristic we'll use this time climb tree again that's right we're going to climb from flat base to base support itself because wheeled legs and flat base base support is an ancestor of both so our model is the same as before but this point has changed enough that I might as well write it out so we'll see that it is light source height between 11 and 24 inches base support some sort of support and a letter awesome now come on parody ah no back up you there we go so that down there don't get confused that down there's our most recent model we now have Oh oops Kiki um okay I actually missed one of the examples however we were actually completely correct which gives away quite a lot we had a miss incandescent bulb back in the old days when incandescent was the only kind of light we had um height equals 60 inches flat back in the old days when flat was only when we had electric back in the old days when we had electric and we also have shade after we've dropped the link for shade so we know that shade or not shade are both okay so the question is what do we do first of all do we specialize or generalize with this miss so obviously we're only going to be able to require for bid so standardly traditionally the answer would be nothing how are we accepted as sane the possibility that you might put a require onto the height some students did that like say oh gosh well that's it that's a hard set 11 to 24 but that's not what you would normally do so just to show that we try to understand that there might be multiple ideas in a weird situation we did get except that that one time but generally there's not anything you have to do this is the system at the time and still now set 11 to 24 this is spookily exactly the correct place where the example would go but you don't have to do anything to the model you don't have to use any heuristics ok so last step the question is a question ok it only the first one no first by a lot it actually do beer as we did it it would be any recipe no it doesn't in different than light source height is different than height flat is different than base support so uh variety gave a huge number and ok here's a question if even though on the tape on the tree incandescent is a child it is a child branch off the branching off hook light source if we instead encountered a because our current model says like come a look not okay instead we got a negative thing that said fluorescent again would that be considered about a near-miss because for us at the first lectures um so listen would it not be considered a near-miss if you had say fluorescent and rather than light source because fluorescent differs from light source so the answer there is um well sure it is memoryless it doesn't know that fluorescent used to be a positive example before unfortunately let's say we did have an example that was fluorescent height is an element of 11 to 20 for base support electric but actually let's say we have on a sodium vapour height is 13 sodium vapour height is 13 base support for some reason it was base it was called base apart and electric that's possible right we haven't seen a positive the sodium-vapor let's say the sodium-vapor was not allowed if we don't say that that's in your miss and then forbid link sodium-vapor how are we ever going to get rid of studying vapor unfortunately because of that we lose some we lose some amount of expressiveness it's actually um your what you're saying is very logical and make sense that you would want to have the ability to say okay this is a subset this should be okay but unfortunately because of that we climb the tree even when we haven't seen a positive example for all things in the bottom on that tree we actually lose that ability it's a trade-off it's a it is most certainly a weakness of Irish learning that you have pointed out here lattice learning fixes that but it has its own it has its own problems so you're right that you'd want to be able to do that but you can't because it's memoryless so last but not least which model would we present in order to teach the system right now from its final system that a lamp requires a base support so we need to be required link for a base for the base support so really fast incandescent bulb with a height of six a flat base and electricity would that do it incandescent bulb with a height of eight and a battery incandescent bulb with a height of 12 wire support and electricity oh yeah by the way this will have to be amiss to teach you require fluorescent bulb height of twenty one clamp base and electricity or incandescent bulb height of twelve flat bottom legs and electricity so did anyone pick that out over the fasted miss of me saying it because Ottawa is gonna want to come in question or answer hmm yeah so I know fast you can look through quiz 2007 final because Saturday is going to come in a minute but it would be a miss on the one that was correct on everything except for it was a wire support then the question you have is wait a minute it says incandescent ball that is now a subset of light source you have his question again the answer to that is they must have decided they must have decided here that that was okay but on the other hand how would it learn sodium vapor if that if that wasn't decided to be okay that's going to have to be an implementation detail so one thing you can do is try to treat subsets as okay in the logical way because questions may do that and then I guess you've lost the ability to forbid say sodium vapor so you have to lose one or the other I think you're definitely okay to ask TAS or someone who's proctoring if a situation like that comes up which implementation detail is chosen because you could choose either when you're running arch learning

Seek backward 5s ←
Longer seek (10s) Shift ←
Play Space Seek forward 5s →
Longer seek (10s) Shift →
0:00 Auto-scroll A Copy link at current time S

Mega-R7. Near Misses, Arch Learning