Numenta Live Streams, 12 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: BHTMS - Potential Pool and Permanences TDD

Description

Broadcasted live on Twitch -- Watch live at https://www.twitch.tv/rhyolight_

A

B

Open up a chat session again.

B

B

She'd helped this morning a bit, but it wasn't great I.

A

B

Sort of alternating hot cold this morning for a little while, okay.

B

So River rat is, we were writing tests combined.

A

With the exercise.

B

A

B

Just hurts not much getting around it alright, so we are writing tests, because what we want to do a quick overview of what what.

A

I want to do today if.

B

Possible may get a look at the right windows up here.

B

All right so in this application we're not even doing encoding numbers.

B

We're going to actually create a real spatial, cooler and simple HTM.

B

A

Going to we're going to use it for these visualizations.

B

A

This one, this is just about a coating, but this one potential pools in.

B

This one, the permanence, is we're going to use a spatial cooler instance to populate those. Instead of just the data structures we're currently using and.

A

We're going to do that in this simple htm'.

A

B

A

B

This with that's the code base that we're going to be working in today. First.

B

So, if you'd like to follow along this is where we're at so let me check out a brand space. Spatial bling and I'm gonna. Add the algorithm here. Tests.

B

A

I added these two.

B

New tests offline a little Obama's not applying on us. This wasn't officially doing building HTM systems, but I just added this test and.

A

B

Added a spatial Euler. So now we have a spatial cooler tests and we have a spatial pool or class. It's better cooler classes. Empty can I, have tests, tests, say.

A

A spatial cooler.

B

Instantiation, when creating potential pools with 100%.

A

Connectivity but.

B

A

Shenzhen here with 100%.

B

Activity, each mini column is fully connected to the input space and.

A

Contains potential for cool indices: that's that's what we're gonna so we're going to get running first and.

B

Then we'll come down.

A

Then we'll make sure this works, I.

B

Think this will just work, but.

A

So, first of all, first failure.

B

That we get, if you do test-driven development.

A

B

Know you just run the tests and fix the failures, and it's.

A

That we need a buffer pools function.

B

And there's no manually failing this one.

A

B

It's not implemented, but first of all, we'll fix this one. So we're.

A

Going to create a pretend, get a potential pools.

B

A

B

If you remember I, don't know if you.

A

Remember or not, but the way I do test-driven development.

B

Is stupid not stupid? It's you. You.

A

Write the simplest code. I know some people hate this, but you write the simplest code to get the test to pass.

B

I enjoy working like this yeah.

A

I'm gonna build my own basic.

B

Sp and if I, if I run into trouble, I'll go look at it Paul a spatial or implementation for inspiration. That sounds good.

B

B

A

The easiest thing.

B

A

B

This pass would be, and this is stupid- easy is essential, pools shouldn't, create.

A

Should return a.

B

A

One of those things is another list, so.

B

A list of lists and each list basically a list of indices, so I guess fluke. We should actually what's actually store some things from the constructor.

B

They were sending in an options.

B

A

So what I'm getting is.

B

B

Size and connected percent and.

A

So I can create exactly.

B

What we want here, using that so Gibson and, like I, said it I'm not going to take a ton of shortcuts. The outside video feed is block I'm.

A

Sorry, Dave I'm afraid I can't do that.

B

So this is the code we're gonna write here, it's not gonna be pretty. This is going to pass the test. That's the idea, so.

A

B

Call give it to the pools we're expecting it to have the length. Oh, that's the same as the size that we sent in okay, so I.

A

B

Let's say: wet pools, equal new array.

A

B

Table stop size so.

A

That'll pass that test.

B

B

Now so let's do this like sort of step at a time, I just want.

A

To see that tests are passing, this is a very incremental approach.

B

To programming but tend to make the least amount of mistakes, we do this right, let's see if this pet has passed okay, so that test passed because we're pet we're just passing in the array or the right size now.

A

If I put, if I change this.

B

A

One I also would like to see this fail.

B

A

It doesn't so something's wrong because we're not returning anything, so it shouldn't even pass.

B

A

B

A

I, don't I if I didn't. Oh.

B

A

Don't know I'll.

B

B

A

B

This isn't even like my test: isn't even working because the I'm asserting.

A

B

Why have? Why doesn't this work? I mean this should fail, because this isn't returning anything all right. Give potential.

A

Pools is not returning.

B

Anything with each mini column so am I not oh, it's probably because I'm, not stealing spatial.

A

Or instantiation of this creates a potential.

B

Pool for each mini column, it's pretty common, fully connected, see.

A

These are just because I commented the power and.

B

Contains a potential pool indices, it's I know what it is. It's because this it it shouldn't be within this other it and I've.

A

Since moved it, it.

B

A

So that's fine! Okay, so now it should fail because we're not even returning pools is on the phone okay. So we return pools all.

B

Right so now this touch to pass well.

B

A

Right, if I do this, it should pass.

B

A

B

I need to set this I'm gonna. Do this here, all right, we'll.

A

Deal with that later.

B

It contains certain technical indices.

A

B

Doesn't really matter the connectivity here.

A

B

Root will restructure this I'm server. I just want to see this test pass and then go then.

A

Now we're going to now, we should want to fail if.

B

I add like 1 to.

A

It good just want to make sure that I'm testing the right thing, yeah I'm, checking the tools for.

B

Sure all right, okay,.

A

So yeah 201 is not same as 200, so that.

B

Is working so now I need to expand my test and it will do that by saying for each pool, let's assert the length this crack so.

A

This is gonna fail, because these own families.

B

They're not they're, not Oh, for each. Oh, that's because the for each doesn't work, because this is has an undefined amount of things David. You pointed me out into this. It pointed me to this last time so, instead of dealing with.

A

This I'm just going to actually create.

B

I'm just gonna create.

A

What I want here? What I want is first I'm.

B

Gonna do score explicitly like I said this ops dot size. This is going to be for each mini column, for evil and I call this mini column index for each mini column.

B

All right, a four-digit mini column, we're going to say, pools, dot, append.

A

Yeah I know it's pretty full push.

B

We're going to send it.

B

Sighs we're gonna, send it an array now.

A

Now we should have a test that fails.

B

That's the idea, you're.

A

Hoping it's fails. Okay,.

B

Good, but not for the reason, I thought sighs. Oh yeah yeah you! You were trying to point it out. It's not for each it's time.

A

B

Essentially, what I'm looking for is times and we.

A

Did this by doing this silly silly erasing.

B

Right array, this table size right.

A

Was this something like this for.

B

Each or does it do I've.

A

Already forgotten, however,.

B

Even doing this, because it doesn't.

A

Make any sense there should be a much better way like lodash has times.

B

Which is exactly what I want, but I'm close to you're a little close here see David the.

A

Kotel coming in to fix things, thank you for doing that.

B

Why doesn't this work maybe.

A

I wasn't right.

B

Sorry, just core yeah I have myself muted.

B

All right Davey, can you hear me yeah.

A

Can hear me I can hear you.

B

Can you guys hear David I think actually here's some Paco in French, hopefully.

A

There's no echo now that.

B

I've got my headphones loud. Oh I can fix a sorry. Here's now, there's no echo okay!

B

So, what's going on here.

B

Isn't this the way we had done it earlier? This is for each deal.

B

Yeah, it should be doesn't like it.

A

B

A

B

Javascript they're just using hopefully the same version.

B

Javascript project configuration- oh, this is not part of a JavaScript project.

B

Commonjs es2016 I guess that's I, don't know where this even came from.

A

B

There is a je s: config file somewhere, yes, Code.

B

All right, I saved it. Maybe that will fix it, but that's that should not affect my run time. I mean I'm running with I'm, not running through running note, eleven point: seven: okay! There's got to be a better way to do this.

B

Well so, like I said, I'm just gonna do a simple way, we'll say: I wish servers like a range pythons got nice range auctions.

B

Why don't you just pull unload? No I'm, not gonna pull into the third-party library. For this.

B

When's, the last time you did this I know it'll work I got pools, it's actually faster yeah using like for each okay boss,.

B

Okay, so that now I'm hoping we'll get it like a legitimate failure. Yeah.

A

Okay, yeah, so that's the failure.

B

I wanted to get, which is this should be the size of the input, not the SP. This should be the size of the SP. This should be the length of the input so and it's saying respective depth of a length 200 contains potential bull indices. It's even the one that I expected so I'm doing it wrong.

B

This is silly. What am I doing? This is a this shouldn't be count on I'm like screwing everything up here. This stops dot sighs, so this is every mini column for each mini column. We're going to push this and we're not there yet. So this should give me a different error, not defined.

B

This is P, so now we should get this assertion, failure. It says a hundred and it's actually zero. Alright. Finally, we get to the point where we have the right failure, and so at this point we want to this- is now we're gonna have one pool.

B

This is a pool for a specific mini column and that's what we're going to turn back and now we're going to do the same thing essentially for let's use I and J, because that's like the quintessential count numbers they're always do it for 2d array, movement and j j less than this dot, ops, dot input count. So what I called it input count.

B

J plus plus and Poole Bush, so here's the tricky bit first of all, I'm just gonna push I I'm, just gonna push and let's just push this I know that's not right, but it'll pass the test. That means we have to write better tests. Okay, so that's certainly not what we want to do now.

B

We have to actually look through and make sure that each index, so this is a hundred percent connected, so every induction should match so now this is going to fail because I doesn't equal zero, but now I'm gonna do this wrong first, and because, if I put a J here, that's um that's not now. That is what I want. If I put an eye here, it should fail actually I think both of them will work.

B

A

B

J is the one that works: okay, okay, so this is not going to work for everything all right, so we have a passing test here where we're getting a potential pool with a hundred percent connectivity.

B

So let's do one with less than a hundred per second activity and and for tests. I do not mind a little replication. In fact, I think it's almost better to have some replication of test code, because then you can go straight to one test and understand the whole thing. Okay. So let's get rid of this one.

B

So at 50% connectivity um this is going to be tricky because it's like random, so some semi random, we don't know which indexes are going to be included in which art, because for each one it's a 50% chance. So there may be I mean there should be there's a it'd, be a distribution around 50% of those values. So first.

A

B

See, let's see how to retest this, we don't know exactly. We don't want to be exact, but we know there should be around 50% indexes so for each length of this input count the.

A

First of all, let's watch it fail.

B

Because it will fail, it won't fail. That's because I need to create a new spatial pool or with a different.

A

B

I'm gonna do this I'm gonna I'm gonna, put my spirit s time, I'm describing so the input count connected percent. Now it's going to be 50 so out of four each one of these columns. We should be connected to about fifty input. This I don't know what the error is going to be, but I think what I'm going to assert on is that there's somewhere? First of all that every every index is within the range?

B

Okay, there's a couple things that also huh Janesville indices I'm having some ideas now, all and C's are in range of input, space, okay and we're going to do one that says no repeating and put space and disease, something.

A

B

So those are, those are, two tests will will need to write.

B

So it contains potential pool indices, so I'm not going to do this anymore. So for each one.

A

B

Pools we're going to.

B

I kind of don't want to kind of want to.

B

I'm not gonna suffer for each one of these pools. Now I want to do basically these these two checks.

B

Perhaps it's not going to work if I do it this way, but these are basically the things I'm going to be checking so for so, for the pool I need to go through all the indices and make sure pull dot for each and we'll call this pool index.

B

But one is that I could just make this comment.

B

All indices are within the range of the input, so that's one check will do and then no repeating input spaces that will do after I'll do that down here, it's easier to do outside the loop.

B

So all this is is assert.

B

B

But firmly with the assertions is at least now at most, the pool index I just have to make sure is not bigger than the input space this below.

B

This dot, oh whatever I, whatever the input, counted, yeah imp account.

B

Okay, so I think this should pass.

B

Like this will just pass, no argument to below must be in number: oh yeah yeah and put count its below. So it's pool index pool index below and count.

B

Okay, so that's that is what I expected now, if we, if we programmed this decree to return some crazy count like this, should this should fail more than one test, but it should definitely fail the one we just wrote there should be two yeah detains touchable indices.

B

I've got both of them, there's a close to assertion to like that close to, and you can give it oh great great, so I'll use that when we're doing the fuzzy thing, but for this this is an exact and what I'll do is I. Think there's a way. Can you give any of these will index out of input range I think you should be able to specify that and I'll give you an error message: yeah pull the index out of a big range purpose: okay,.

A

So that's the error. I wanted to get now.

B

If I take this off the.

A

Other error: what.

B

Did it say, expected zero to be one, so let's try and fix that one. So this is linked, though so yeah. So this this error is pool. Count is oh, no, that's in a different way.

B

This that's 100%! Okay, so this should say was for.

B

A fully connected pairs should hit sync fully connected layer fully fully connected pool should the same size as input space. Okay, so I don't want to see those errors.

B

It's not even know one, that's tripping, which assertion is stripping. Hands prospectus one to be equal to zero, so it must be this.

B

This is testing how oh yeah, so this is the wrong one. Well, this should be the one that goes in here.

B

Oh boy, right because this is the one I'm actually checking the size of the pool. This one is checking how many pools there should be or how.

B

This should always be the same size as the McCollum's pool should be same size as.

B

Okay, where's this.

A

B

Three, so this is what I hate about this. It's not I, don't know the line number.

B

It doesn't die you head for the top Aires.

B

I don't see it I mean it says six like six to eight, but that's not right, I, think it. It's compressing it or something.

B

But it's definitely test one spatial or instantiation 100 per second activity contains potential indices, contains petition for sees it.

A

It's got to be this.

B

One, the only one that I have in a surgeon or haven't given a description for and.

A

What am I doing.

B

Here sure the same size input space, okay by.

A

Doing this, oh, this is just.

B

Checking the index right, I'm so say each full index should match input. Space index was called like that. I just want to see there. We go he's pulling this metric so now at least I know of a better idea. What's failing, and that's only failing, because I did this so now.

A

B

Should work okay,.

A

And I've got my.

B

Butter test failure, descriptions and I'm going to add the same thing here, a certain length.

B

And I'm gonna just have to change this I think.

A

B

No, that's fine! That's right! That's always true pools should always do the same, submit a comp count, but it's this one. It's going to be different. These are going to be different because we have 50% connection. This is the only one I'm going to do inside.

A

B

So, for each be the same size for each all, indices are within the range of the input space did I not do this. One I should have done this one up here. Yeah.

A

B

Isn't this has nothing to do with the 50% connectivity? So why don't I do it here.

A

A

B

A

B

No, this is right, input index cool index yeah, but that's just cuz. It's a hundred percent connection right, so you'll have every.

A

Single yeah input.

B

Thank you and I should also I. Don't need to do this here, because I'm basically checking everyone so I'm not gonna check everyone like this, but I am going to so. I did have this in the right place, but I am going to check.

B

I'd have to do the same thing here for for each look at the end, no I don't need the pool. This is fine for you to pull index, make sure it's below the input count. Okay, so that should should work. I think I might have already run this okay, and then we do this part outside of the loop.

B

Oh I already did that, so this would be the is like close to sir. It was it close close to so I'm. Assuming we're going to say the coupie length is close to emphasize divided by 2 or x, the connected percent or something input input count times. Let's say connected percent that way we can change it later connected percent times.

B

So we're gonna move the decimal point so I, don't so I, don't.

A

Know you got yeah.

B

I think that's an argument of close to. You could tell what, like a delta, to make sure it's within a range okay. Let me look it up. That's so quite a link in the Twitter.

A

B

Expected Delta message actually expected: Delta, that's okay, so we will do actual PDF length expected.

B

Is well an input count? Well, it's exactly what I was doing input count I'm expecting it to be whatever the connected percent is. No, let's make this a let's make this a decimal to make things simpler.

B

It's gonna break things, but no I won't break anything yet so then we could just say connected percent here we don't have to and then so. The Delta is also so I'm gonna make a delta shoot. I, don't know it's something input count. Ten percent I, don't know and foot count times point one. Let's see what happens?

B

Let's see what happens? It's not gonna work right because I haven't actually set it up. Connected percent is not defined.

B

Respected it 100 to be close to 50 right. So it's getting a full. You fully connected I think that's a reasonable go 10 to 10 percent. So that's a legitimate failure. We expected a hundred to be close to 50, but it certainly is hot. Okay.

A

So now we have to do.

B

Something legitimately do something like real code, because this is a stupid function. That's so that's always fully connected.

B

So we're only gonna push this under a condition. I'm gonna say if, after andum greater than this cops dot connected percent, then we'll push.

B

So I don't want to do that. David I want to keep this really really explicit, so even if, if maybe we'll come back and refactor later, but especially during the construction of all of this logic, I want to be really explicit and and I think this. This is pretty explicit, so this is exactly why we would create a potential pool we're above the connected percent, so I'm.

A

Gonna run about a less than sign.

B

That would make a difference with it expected 61. Are this interesting we're getting quite a bit of deviation? Then 60.

A

B

Sixty-Two, okay, so some.

B

Why, with that.

B

B

Want to make sure I'm.

B

Sumari thing.

B

One point: okay, so that's that's right! Math, dot! Random is.

B

A

Is this happening.

B

It's just freaking me out now it's 38. Maybe it is. That was weird, because I had sixty something three times in a row, and now it's down I, just expected it to be like closer to 50. Okay,.

A

I got it better. I got.

B

Idea, let's, let's just increase the size of all these accounts, so you can have a bigger like sampling space. So, let's just add one hundred I, don't need to add 100 to that. So let's look multiply this by 100 for the infant count. We should get more reasonable numbers yeah. So.

A

B

I think I just needed to increase the number of samples. I was doing to get something closer, which makes sense because it was deviating to much fewer samples. That's.

A

B

Sample size is important, but this is still gonna fail, occasionally, which sucks I mean. If I ran this enough, it will eventually fail. So it's not the best.

B

I figure I make it, though, as.

A

B

It doesn't take too long I take that takes a long time. It takes me too long time out, so we're not going to do ten thousand one.

B

A

Like I said, we're not gonna, think much about.

B

Optimization or anything for the time being we're just going to make the functional tests and make them work so where we're at is we've got a different connectivity. I am also.

B

We could turn this into a test that goes through several different connectivity's and tests. It probably might be a bad idea. Well, let's do the repeating input test.

B

So basically, this is a.

B

So so I'm just gonna take a I'm gonna. Take the.

B

Pee, are we in the pool most it yeah, so look so for this pool. Isn't there a function.

B

But there's an includes, but that's not gonna tell you that's a unique I, so I swear there was a way of death might have been Python. I think this is super easy in Python. You.

A

Can see videos that dope.

B

You could filter and then check the length of the filter, as is set a top-level object in JavaScript yeah. So the idea being so, that's basically unique so.

A

B

Say new set and give it the P and pick that length that should be equal to P. Don't like that's testing, weather.

B

Contains more than contains non-unique indices? Okay, this is probably gonna pass, but I'd like to see it fail too. So, let's inspect it undefined to equal 502, whatever set, might not have a link.

B

It's it's probably see it.

A

Maybe ties yeah.

B

Blank rocket science and.

A

Is it a function.

A

It's a poverty.

B

Just run the testing okay, so that passed now, let's make it fail, so make sure that we're actually testing something okay. So what I'm gonna do. This is gonna feel more than one thing, but I want so I wanted to solo. I wanted to feel the right thing so.

A

I'm gonna push.

B

A

Is a little crazy, math.

B

And less than 10% of the time, we'll add we'll add something twice.

B

A

So this should fail several.

B

Tests on hope, yeah pool contains non unique indices. That's the one and.

A

B

Length test because it was Scott way too many things like that. So that's good! That's good! Alright! So let's not do that and watch.

A

B

All is good, ok, alright. So just to sum up where we're at well you've got a spatial Pooler when creating potential pools. 100% connectivity contains all the pool indices contain it contains such a pool and the Seas will same so that's pretty calm count fully capable of insane times in its face. Each pool index should match input space index.

B

Now, let's do this.

B

A

B

Should we should we should be able to do something like this?

B

So let's say, Const connection, connected percent test cases and we'll say 0.10.

B

Mm, let's just do point 0.25, 0.5 0.75 and then.

A

And then 100, but.

B

This should work for all of them, and so so, though, I'm just gonna loop over this.

B

See what I'm gonna do I'm gonna basically take this test. I'm gonna loop it and do this test for every one of these, so that we can verify. You know it's some type of spectrum which it should work exactly the same for each one of these, so connected percent test cases for each.

B

Target connected percent.

B

And now I can pop all of the strong I think into this okay, and then this doesn't exist. Connected percent is going to equal, targeted target factors.

A

B

This should work.

B

This should work I, wonder.

A

Okay, this will be.

B

Great hold on I got idea: I should be able to add another described here, described.

B

With and then specifically say with.

B

As a target, effective percent connectivity I, this works.

B

So now I can put all of this inside this describe and then in our test. Readout we'll get a full description of every single one that we tested. I hope here we go 25 50, 75, one perfect, okay, so we we are testing I mean this takes a while, but I don't want.

B

The focus has the quotes and the.

A

Describe are different.

A

B

Yeah, these are a strict attempt, a string, template yeah. That's so I can do this substitution here. Okay,.

A

B

Good, this is good. What else do we need to test so I mean we're writing way more code than our test code. Then we are a real code, which is what which is good, which is what we should be doing. So we've got one the case with 100% connectivity. We've got the case with less and then we go through all of those cases.

B

We've got it. This is sort of a special case, but because we're going to specifically check every indexes included- and all of these were just checking that it is close to 10%- and you know we could. If we wanted this to run less time, we could increase our very. Are our wiggle room here right.

B

So, let's, let's change this to tolerance. Let's say test tolerance, I'm going to put this up top here.

B

Equals I'm gonna make it zero point two.

B

A

B

We're gonna put 20% we're using the impact of a hundred, so I bet this will pass and if I put it back down to one, it might not I think one of them.

A

Yeah they all passed and then it.

B

Ran a lot faster: oh it's like increasing in speed, so I've got a member leox somewhere or increasing in size, but maybe in the test. It's probably in the test.

B

So I put this down to one.

A

It'll, it might fail.

B

Yeah, so let's find the happy middle here, I want to get close. That's still failing!

B

That's it's just too random. Let's make this like 200 at least: let's double it. That might be enough.

B

You want your test to be idempotent, of course, or is it stochastic I? Don't remember, I'm gonna leave this for now. If there is a chance that it might fail occasionally.

B

Okay, so, okay, what are we missing here? I think this is perfectly fine, as a matter of fact, I think.

B

As far as potential pools goes, this is probably fun.

B

So one thing: okay, so I know I know if something we need to do, and this is an API thing when we create the spatial cooler, the potential pool should already excuse me should already be created. We want to make sure there's another test. Here's another ties to secure one we want to it, doesn't matter the connectivity triple first money, column, I, think, I. Think.

B

Get potential pools are always the same.

B

All right, so this should be easy. It doesn't matter which now I'm gonna use one with less connectivity, because that will because that, if they're 100 percent, all they will always be the same. But but let's say central pools, let's call it less than then fully connected. Be very explicit. Are always the same. So.

A

B

Fail which certainly so we're gonna do this again we're gonna we're gonna make this 0.5 doesn't really that's point. 85 does really matter, and then we are going to say, pool 1, equal SP, get central pool constable 2.

B

Couples cert I know, there's like a deep equal, not deep one rule to they should never. They should oh I'm. Sorry people not-not-not, they should always beat you fools.

B

Is it that what word am I looking for stochastic or idempotent when you call this function, is always returning at the same thing, I think it's Idaho.

B

Let's just say there are as soon organism.

B

Okay, two pools: this should be the same. Let's watch this fail.

B

That's not I feel like that's an error. What'd I do that's assertion. Karen!

B

It's just can't read children of undefined it's in the middle of an it of the test.

B

I expected to I'm not getting a good test report out of this, so I mean this. Is it's here? Here's the error expected one or a two, deeply equal another RA. So I guess that's! Okay! That's definitely the problem! Okay! So the way to fix this is.

B

So I'm gonna change, this function, I'm gonna call it I'm gonna, make it an underscore function, I'm gonna say, create potential pools and then I'm gonna say this kind of show, pools and again I. Don't know what the way to do this, but this is the way I always do it as JavaScript still loosey-goosey. If there's.

A

A way to truly.

B

Make a private member that's great, but in otherwise I'm going to do this so I'm going to create the potential pools once.

B

Why can I call this Oh and so now I need to add this function and get special pools turn.

B

Here we go all right, so this will ensure that upon initialization we create the potential pools and they are never recreated.

B

Item potent denoting an element of a set which is unchanged in value and multiplied or otherwise operated on by itself. That's the yeah! That's the mathematical explanation, I think when it's using a programmatic term, it's like you have a especially in the service-oriented architecture, side of software. If you have a service, someone calls it in that's idempotent. It means it's gonna. It's not accumulating anything. It's going to return you the same thing every every time you call it it's not retaining a state and deciding to do something different.

B

If you call it at a later time, it's always going to do the exact same thing. I think it's the right way. I want to make this good potential pool is essentially an idempotent function.

B

Okay, so it's like Maggie today that I'm to school or looks like it's for my private member variables, yeah, that's the only thing, I'm trust Ryan. Did you know that if anybody is looking at this code, they shouldn't call this from outside, there's a better like Eknath scripts way to do this, you can declare it as like this private, okay.

A

How do I declare.

B

It as a private- they just were private, now I think like this, but is it still of this or do I? Do it up here.

B

B

Didn't know there was a private key word in JavaScript.

B

New privates class fields and how to use.

B

Etc, oh okay and do this I'll do a setter and getter that'd be good.

B

But there's no.

B

Favorite class oh I, see so if I don't declare anything and say that is okay.

B

Let me try this: let's do this, let's say we're only getting because you're gonna like get got potential pools.

B

And I think you could use that hash hash tag to clear this private wait, the hash tag. Oh.

B

So it's that's static and private, okay, so hash tag, something in it and I can do that within. Oh, so then I say this stock. Okay, so hash potential pools first they're going to be undefined.

B

B

I need the constructor.

B

That's that's like a future. Yes,.

B

I'm just gonna: do it the way three?

B

We can we come back and make it better later right now, let's just make the tests to run and they.

A

B

Hundred percent connectivity, lots of activity lesson fully connected and we just did the make sure that they're always the same lesson fully connected central pools are always the same, creates an initial premise value for each cell yeah. So we haven't done this, so we could start doing. Does.

A

Anyone think of.

B

Anything else that we should test we're, not testing right now. As far as potential pools, I.

B

Think we're about covered it, we're going through a bunch of activity range here.

B

All right, so, let's move on to okay when you're creating potential pools. So now we need to.

B

When establishing initial permanency services.

B

All right go inside here: okay, so we're gonna. Do we're gonna, repeat some code here, we're gonna again, I'll just use a fully connected one, because we're done essentially we're done kind of.

B

Yeah we're done testing the potential pool. We do need to assert that. However, we get the permanence this matches. The potential pool some house, like the size of the permanence, is like, if I ask for the potential pools, it's going to give me an array of indices to an input space which many column, if I ask for the permanence is going to give me an array for each mini column of arrays, which is weights okay. So it's.

A

Really similar to.

B

Potential pool, but instead of an index, I'm gonna get a weight back and.

B

So I think we're gonna do a lot of the same, but I'm not yeah a lot of the same testing here.

B

I think we might be able to like take this almost exactly and just change from from looking at the indices to looking at the actual weights.

B

Let's, let's do this.

B

Create a new one here: okay, so again, with 100% connectivity whatsoever that contains potential pool indices contains.

B

Any column urban Ince's.

B

I will just say: firms should be the same size as a mini column. Count. That's still true. Fully connected pool should be same size as the input space. That's also true. These are going to be pools anymore. This is going to be home, perms and we're going to say, get permanent suits.

B

Okay, so arm should be the same size mini column, count for each one, fully connected perms.

B

Should be the same size, the input space and then now we're getting so this is not an input index. This is a permanence.

B

Yeah, this is a tricky bit because it's not linked to the index here. It's linked to the pool and you have to get the pool to get the input index this permanence is associated with. So now we need to decide what exactly do we want this? To look like it's almost like permanence is or excuse me. Potential pools is a subset of what you want to do when you get the permanence is what you have to establish the potential pool before you can populate the permanent sleep relative. All points.

B

A

Just from a testing.

B

Standpoint, though, when I call get permanence, --is I am expecting one per minute column and each in each array form any column is just hold a list of scalar values that are associated with the same indices that are coming from the potential pools.

B

A

What I'm actually going to do I'm.

B

Gonna get pools to.

B

Perms and pools, and now for each let's call this perm we've got both perm and.

B

For each permit and now for each permanence, we can look in and get the input index by going to the pools and now I have to go.

B

Sube pulls up a Vinick column index.

B

Okay, this should be the input index.

B

This is not so.

A

This is actually true.

B

This is she stole.

B

It move this off one. If you want to test it that should still work and then actually yeah. You pull induction, execute this space index only.

A

For fully connected and.

B

So then our input index I guess all we need to test well, we need to test.

A

B

There's not duplicates okay so again,.

B

B

So I've got my input index I'm, going to add it to the infant indices that I've seen books before.

B

Index, okay and at the end.

B

The I should put this here so for each permanence. Yes, so put it here so forget your permanence I have a different yeah different I put into the next and then we're going to do my mark round. Thank you. Ibuprofen is kicking out a little bit so.

B

We should have at this point a list of indices, which should be the same length as the input, because each should be represented exactly once. So we can do this.

B

We're already doing this here. So I know this is going to be right.

B

And what I? What I want to do is know that there's.

B

Every one of these there should be the same amount at this point. Okay assert. So here's our first assertion equal.

B

There should be the same amount of perms as input indices. Wait since we're a hundred percent connected fully connected herbs should same way as input space.

B

Okay, let's see if that passes, I'm hoping I think.

B

This is again unexpected thing to pin. 537 I would actually give me a so I kind of miss effects or somewhere in life.

B

There's missin taxi: it's probably something to do with all these invested functions.

B

I just forget to something like this.

B

B

A

B

Give touch, pull snap function.

B

Now get permanence, is it's not a function? Why does it tell me that? Oh that's because I changed I was gonna change it to a Saturday night.

B

Okay, I'm, gonna, I'm gonna set this back because I don't I, don't want to change it right now there we go, get permanent, it's not a function. That's that's the one! So now we have to write, the permanence is get permanence is function, okay, so I'm gonna do the same thing. Basically that I did before.

B

And I'm gonna see, the thing is: if you call get permanence is from step to step, they will change so I can't write the same. The same test, compare about the same test assertion because, as learning occurs, will change. If no learning is occurring, they won't change. I, don't know if I want to test that right now, but I will I'll definitely add get permanence. This here will just make the test pass so far.

B

So if you have more relevant function, perm should be the same size as minigolf count. Okay, so let's basically I'm going to steal this and we're gonna go through and do something super similar.

B

B

Running into some term issues- and he confirms hold on let's- let's change this- to total of the spa. All firms hold this one firm you're like perms for this mini columns, permanence get all permanence a--'s for these size count, and so now we're not actually this is. This is stupid. This is not what I want to do. I have to use, I have to get the potentials, get back potential pores.

B

And then basically, I need to go I want it to match exactly the potential polls, because if I did it this other way, they might not, they might not bad so I happened so I've already done. The calculation I've basically established a pinch of bull's eye out for each one of them. I have to create this value so.

B

Just for skip touch bras.

B

B

All firms peoples that so for pools or each.

B

Cool and I also I, don't think I need the index. I don't need this index right. I do need.

B

Now for pool now I need this. This is nice. This is going to be input index and index.

B

To many pools, oh thank you.

B

Okay, so I'm using my pools and I'm going to create my perm so.

B

A

B

Going to be a terms for.

A

B

For each one of these I'm going to push a value and for right now, we'll just say: math, dot, random and then we'll just test that it expects parents, and then we should be able to return any fools. Not fools hurt all perms.

B

Okay, and here we say all perms stop push and we send in the perm. So each mini column is getting an array of perms which.

A

B

Through the pools, we don't need this pool index, we don't need any of this stuff all we need. We don't even need this.

B

Where can I do that?

B

A

B

Okay, so now we're just gonna get a random fool. So, let's, let's see I should have written a test for this. Probably before I wrote the code, but I wasn't quite sure when I was kid, I was gonna, do so so we're doing the others, but I'm not actually getting the permanence same size.

B

So I should check that the permanence value is a number between 0 & 1.

B

Number between.

B

All those numbers, an integer.

B

So I've done this before.

B

B

B

Stefan less than so, I could say his below is at least if I could say is below and is about harvester bad. Ok, so is the low permanence.

B

Its, but actually what I wanted is less than or equal to is that most is at most permanence is at most one.

B

Will say permanence value and not be large? One both do the same thing is at least zero.

B

Terminus value cannot be smaller than zero.

B

How long am I gonna stream I don't know, I, don't know, I need to take a lunch break, but I brought a sandwich so I'd like to try and get some of this hooked up into the diagrams. So not a couple hours. Probably, if I can, let's see what tests.

B

And put indexes not defined.

B

Right, I, don't know they put index I've got.

B

So many comment he's pullin up should match them in space and us well. Let me think this. Our input indices.

B

For each permanence.

B

Should be the central pool index, are you a declaring input in there before you know? It should be.

B

All right fully connected perm should be same length. This input space.

B

Which I confirm the same.

B

So this one's failing is my perms are 200 and there should be 100, I. Think I, just hard-coded a comparison.

B

Perms should be not the input indices.

B

We're inside so for one mini column. Its permanent lengths should be the size of the input array party asserted the other length you do length of.

B

B

I, don't think that's gonna make a pass, but it's better! So okay! So why isn't this working this.

A

Is a valid leave, a valid failing test.

B

So it's something the way I'm doing this, so I get the pools for each pool. I create a permanent, so this pool is one mini column. I create a list of permanence is for each.

B

I've already this, this pool should go through every input space for each one, I push a random one.

B

This is where we're feeling here this is at the top level, nine.

B

I think this should be called curb.

B

This is the actual value.

B

Okay hold on a second okay.

A

B

To change some name, so they gave some naming confusion here. I'm gonna call this all terms: okay, I'm gonna call this perms.

A

B

This is like one mini columns of evidences.

B

So so I'm asserting the wrong perms here. So this should be perms, not all perms I've, already already doing the all perms comparison here. This I'm making sure that individual many commas permanence is are mapped to the input that should have to see him as a boy I. Think that's right have some naming issues.

B

So and this seems to be working so what happens if, let's make this fail, let's add one to all of our numbers and we should get a failure on the upper bound. Yeah can't be larger than one. So that's.

B

Okay, so we are not checking the distribution of the permanence is yet right now they're, completely random, and if you remember the way I created, the permanence is last time what it needs to be in a subtype of distribution around a center point. At least that's that's what we do and before.

A

B

Let's write, let's write it, let's write a note and commit some changes and then I'm going to take a lunch break.

B

And we'll probably come back and do this whatever test we come up with here, maybe we'll come back and do a diversion of it for less connectivity. You know basically the same thing we did with sintra pools, hopefully be able to just copy and paste a lot of stuff.

B

A

It contains many common premises and I.

B

Need to ensure that many column, permanence --is.

B

Permanence permanence --is, our.

B

Normally distributed around a center point, so this would be the tricky bit here. We'll come back and do this.

B

B

A lot changes: okay, I, don't need this I, don't know. Sir. We've got category tests I. Remove that a spinal just hanging up special puller tests is all new stuff.

B

Special puller is all new.

A

We don't have a lot of code in here, yep.

B

B

Let's do them all the show Espie tests for potential moons.

B

Market I went to do them all.

B

Or pools and arms.

B

We'll have to commit so long.

B

A good push origins, facial pooling, so that should be out there. I haven't merged it. Yet we're gonna do a release. Once we get this branch fixed up, we'll do a release on simple the HTM and then we'll go to building HTM systems and.

A

B

Then we'll go to building HTM systems and require the latest version of the release and put this in place, but first we need to get that potential permanent start. One thing before I go because we're going to have to look at this soon enough is that Bates distribution that I used in building HTM systems and.

B

Where are we on that? The potential pools and permits misters appearances, so whether we did this the other day and we had right, render permanence --is? There was the Bates Bates.

B

In spades, random bait setup initial connection strings: this is the bit we did so we're gonna be pulling this out and using it somehow in spacial pulling I could I don't know if I want to write my own random baits thing or use D threes, I, don't know I'm trying to create simple HTM with as few dependencies as possible, but we're gonna basically do this inside the spatial polar when we're doing this permanence this so that we get them distributed properly.

B

So we'll do that next right now take a break. Do some more stretches eat some lunch feel free to chat. I'm gonna take headphones off, so I can hear you. You can walk around the room.

B

She was wanna, see the guitar of God coming I'm, so excited about this.

B

I pushed it to a branch, David called spatial pooling. You should be able to see it on my fork.

B

This way my guitar I, don't have it yet eventually soon I.

A

Should have it by Friday.

B

Look how pretty that is.

B

I've been waiting for a long time to get this to get a nice electric guitar and my saxophone finally sold and it's shipped already holy I love Sweetwater. If you're ever into anything, avy music, stuff, Sweetwater, calm it has, it is ready to be picked up by the APS. That's so exciting,.

B

Wonder with the chat of statuses, hey.

A

B

Think it's! Maybe it's back! Hey, hey! Look at that!

B

Okay! So I! Guess I don't need the second anymore.

B

That was weird.

B

That's a lot better I, don't like the other chat. I still have no idea. What was wrong.

B

Just magically.

A

Fixed itself so.

B

Works for me welcome back to chatroom.

B

I'm taking a quick break and we are building HTM systems, we're doing the spatial pooling right now, we're just creating the initialization for a spatial cooler.

B

We're not even using d3.js or react today yet or might use d3 a bit later. We're just hooking up the actual initialization functionality into a spatial clipping school loop of a spatial Pooler class and in our simple HTM project and then we'll integrate it into our diagrams.

B

All right I want to try and get this done.

B

So I wonder how hard it is: simplement, a random Bates function, random Bates.

B

Bates distribution.

B

The mean of the numbers disagree: independent, uniform, clean distributed random variables.

A

B

The interview no excuse me.

B

Continuous probability, distribution of the mean of n, independently uniform, independent uniformly distributed random variables on the unit interval.

B

This is defining.

A

The probability density, oh.

B

A

Know what I'll do I'll.

B

Just look at d3.

B

I'll look at the source code 43. Let's.

A

See exactly how they doing.

B

All right, random, bates, sorry, but this goes straight to the code right. Oh.

B

So it's going somewhere else, erwin hall! Oh my gosh! You.

A

Guys see her one.

B

Whole random urban hall hall.

A

B

Random interval, urban hall. So that's the thing this is it song. Zero. Zero is zero to the end times, plus one some go to the source source as the function we're sitting in.

B

So hmm let me just support this is exporting and calling.

B

So we take a source go back even further here when we call random base, we give it n, which is the number of the recall, what.

A

I'm calling a connection.

B

B

Yeah or spread I think I, maybe call it a spread.

B

What does it do we reckon so? This is the end is how many independent variables this is a math thing. So it's this we're doing this bit we're calculating a random Bates distribution. Every time you change this 25 is the N or 20 the think. I'm changing right now.

A

B

The that's n, or what they're.

A

What they call.

B

It as independent variables how many independent variables so.

A

B

Want to sort of just convert this.

B

This code, so I'm gonna, send it in which we are doing here or we're already sending it end which I'm calling spread and then there's something else that's outside of this. So all we're sending into this is in so I want to basically take this function. I could just use d3 I mean honestly, but I'm trying to prevent going into other libraries for things so just because I want it to be really clear. What's what's happening in the simple HTML page, so I want to so I could just like.

A

Steal this you know and.

B

What is why do we need default source? What is that I'm confused by default source? Oh that's! Just! Oh! That's! Just a random!

B

That's weird default source! There's a random number! I!

B

Don't know why.

B

So the that function gets called when it's exported, that's like wrapped, so it's just needs a visa.

A

And it generates a function.

B

A

B

The API is this: returns a function that basically runs your random beads for generating numbers.

B

So I could take this probably just put in my utilities.

B

A

To understand the seed.

B

Alright, what it's not like a seed, it seems like it.

A

A

B

Urban hall thing also takes a source and here's the end like here's, the function you're actually calling with the end when you call it I think, and it returns the function. This is what does it, so what do I have to go through this hoop like we're, jumping twice you're here? What is going on in here? That's not going on in there. It's divided, it's calling it and dividing by hand random urban hall.

A

B

B

So maybe I will I mean there's one easy way to do this I think. Let me check my let me check and see if I'm using d3 anywhere I might already be using d3 in this code base. If I am I mean I might as well continue using it bounded category I think I actually did use it for some of the scales scaler yeah.

A

Yeah, okay, so let's just use it, let's just use it.

B

And then we won't worry about.

B

I won't worry about I'll, just say we're using the random Bates distribution.

B

And that's what we'll.

A

B

So but I still need this bit.

B

So we'll take this. This is the main main code that we're going to need to make this work. Of course, I want to write a test so.

B

Let's just Park this for a moment: let's just Park this right here and and start writing a little bit of a test. So how am I going to test this but tape? So here's we're testing a fully connected. Tentacles are the same. Let's get down to the permits! This contains many column, permanence, it's okay, so the okay one thing I could do to test. This is the average of all of the connection weights should be.

B

The center of this should be point five, essentially the center of the distribution. That's going to be the correct! That's going to be. That should be correct, no matter what, unless I start twiddling with the center of that distribution, of course, but as long as.

A

I'm not doing that the.

B

Average should be should be 1/2. You know we're halfway between 0 and 1 I.

B

Think that's true! Yeah, ok provinces are normally distributed around the center point. So now we need to do some of this again.

B

I'll take some of this and copy paste, so I'm going to get my permanence ha's and get my pools not that assertion for all perms many columns index I. Don't need to keep track of this anymore.

B

To do that, I'm going to do that! Want me to do that. I'm going to do that! Ok, all I need to do is is test that each many columns average of older permanence --is should be close to 0.5.

B

Unless I specifically tell it not to be like I should be able to change that average easily by moving the center of the distribution, whatever amount I moved, the center of distributions I should add to that point 5 and it that should be the new average.

B

So the center point I'm going to put by default and that's gonna be 0.5 by default. Okay, so now I've got all my perms I. Don't need this anymore. All I need to do is say, firms dot.

B

Okay, so now now I'm gonna, look at Xmas script again, java script array average an easy way to do this. No, no. There should be damn it push it deep.

B

Okay, I'm sure d3 has like an average yeah.

B

D3 or a do I have to create a d3 array.

B

What they might call it me, and they do d3 mean mean all right: let's try that and.

B

We're gonna go grab, it.

B

B

Iii mean of my powers: okay, assert is closed, or is it wasn't close to okay? So this should be I'm shooting for point five.

B

That's the average and we're gonna, give it plus or minus 0.01 I.

B

Guess point one: average urban, it's across many column should be Center at.

B

So this is gonna.

A

B

I think cuz I mean we're doing math dot random, which is should have the same property.

B

A

I think this should work yeah.

B

I wonder how close we can get still five percent Delta yeah. So it's a little it's a little iffy, so that does work. At least we know that works, but that doesn't really test our. It doesn't certainly doesn't test that we have a normal distribution. So now what we can do to test there's a normal distribution as we can sort the permanence --is and then assert that the midpoint is after we've done the story.

B

We should. We should go through them and find where we hit point five, and it should be very early on after resorting well that depends on the what we would call that's the spread or the independent variables. That's, but.

B

So, first of all, let's set up normal random beats with independent variables being one which should be basically math dot random, and this test should still pass and, as we set up the random beats, it should continue to pass.

B

Okay, so let's, let's put the random baits in place.

B

All right, so our baits function we will create when we initialize.

B

I'll just say this baits function d3, random baits, and this is connection distribution. I, don't have a parameter for this. Yet.

B

A

What I really want to do connection.

B

Distribution, I guess I'll call it connection distribution, but this is really not a good name. It's it's. The number of independent variables for the base calculation, but lo is really really spread out. Hi is like super peaked, so somewhere between like 0 and 30, is what I typically been seeing here.

B

Let's call it beats and the let's call exactly what it is: independent variables, okay or.

B

B

So if I send it one and then I and then I put in place the if I don't send it anything and I put in place, those now I can use my Bates function. I have to have a center again distribution center. Let's, let's set this this dot Bates center equals this table, stop Bates center for 0.5, because we want it to be the center of the weights by default. I.

B

Reserve the right to change all the stuff I just want to I just want to see the functionality working and the test passing. Okay, so I have a center and then essentially for each mini column. We're.

B

Going to get so I don't have to do this anymore. This connected percent thinks they're all connected the ones that I'm going through already connected. All I'm doing is calling Bates function plus a center I've already got the center, but I'm, not gonna I'm, not gonna. Do the center, yet I'm not going to do the center yet so so all we have to do is instead of math dot random. We just call this dot beats.

B

Okay and that works, it's essentially random, because even if we change the distribution, it should still work until we start moving the center point. So.

B

Now so, if it's completely random, there should be essentially the same amount of numbers in like the first 10% of the range as the minute of the middle of 10 percent of the range. If it's not random, you're gonna have a bomb you know, so we want to get to the point where have a ball when we call it with with no independent variable and it uses one. Let's look at this again, let's be explicit about what we're doing here.

B

So if we make this one 180 one just one, that's that's the random spread. We make this two, it's already not random, so there should be a difference between one and two and certainly difference between one and twenty.

B

A

And zero is just flat.

B

B

All right so how to test this, so if I, if I use one, it should be essentially random.

B

If I used to there's, there needs to be distribution, so I need to check.

A

So here's what I could do I.

B

Get around the center point: that's assuming it's default! 2.5 I should be able to take an average and make sure that that's larger than that an average there, the on either side of it. Okay, so does that sound like a reasonable test just to ensure that there is a normal distribution or some distribution going on that's centered around a center point, I found welcome back I.

B

Think that's what I'm gonna try I'm gonna take. So let's go back to the test. Is all the happening within the test. All permanent sir, is normally just speed-read Centerpoint default. Okay,.

B

A

B

The good name of the test harnesses are normally distributed, no, let's say randomly randomly distributed by default, because.

A

B

What we're doing here, we're randomly distributing them by default.

B

And so let's assert that as we go through, let's go through like 10% 10% 10%, let's just let's window through this, take an average and make sure that they're all very similar I. Think that's the right thing to do here and.

A

B

A

I could just grab like.

B

The first 10% the middle 10% and the last 10%, and ensure that the average of them are essentially the same I, don't mind being very explicit about about this, like that. I don't need these pools.

B

So let's try that let's call this first 10% and we're gonna say: firms zero, two perms math.floor.

B

Firms dot blank /, 10.

A

B

A

No way, what am I doing her.

B

I, don't need that math dot floor I'm, just trying to get a slice and I'm not doing it right hold on curves dot slice, okay, zero! Two! There we go zero to whatever. That is the first length divided by 10 to the first ten percent Const last 10 percent. Basically, the same thing except backwards.

B

B

Stuck 10% up arm, stop Thanks, actually I. Think I can just do this yeah.

B

It's not splice. It's slice, slice slice from zero to something or from some length to the end.

B

Yes, place will give me a copy of your right right. That's perfect, okay and then I'm gonna. Do the middle middle 10% equals slice and.

B

A

B

A

Doesn't have to be perfect, okay,.

B

Starting at this /.

B

Hold on hold on before.

A

I, do this let.

B

Me finish: stop.

B

Sample width is going to be 10 percent of the firm's length, so math dot for firms per length divided by let's say times, 0.1, okay,.

A

So there's our sample.

B

Width so now I don't have to use this anymore. Now I can say sample with it's this clear now the sample, if we're going to slice from the sample, with an onward back from the link right, so the last 10, and so this one will be diverted this starting.

A

B

Let's call it mid index equals.

B

A

B

Then we're gonna slice from the mid index.

B

Okay, math dot floor sample width divided by 2, comma.

A

B

Index +, actually, let's, let's pull it out.

B

Squat 1/2 wrist this. Why not.

B

Or call it half simple, apps.

B

A

B

This or this all right I do I, do I will soon. Thanks for the reminder.

B

Okay, it's gonna get some of the minutiae of these calculations on their way. So, yes, I do want to sort them.

B

A

I'm, just thinking.

B

I'm thinking I want to.

B

There's no apology wait a minute now I'm a little confused I may have just confused myself.

B

Do this, how do they make this histogram I'm I'm having a brain, fart I need to go, look and see how I made this histogram, because I forgot me how I ordered it draw a histogram. The data is mini column, selected, mini column.

B

B

Freeze really column you're, creating another size of the input space.

B

If it's connected plus this.

B

This is hard to do without looking at.

B

Distribution center, it's.

A

Hard to do without visualizing it honestly.

B

Asian Center.

B

You think what I, when I called this.

B

Just manually telling it 0.5, so this is only no no okay yeah the the random beats will give you a distribution centered on centered on 0.5 by default. Yeah.

B

So the point of that is what the D, what your connection threshold, that.

A

Would have Center.

B

It well that line is the red line. Is the connection threshold? It's independent from the base distribution. We can have it wherever we want, wait, a minute, here's the line. This is the line. So when you change the connection threshold, it sort of operates on top of the on top of the permanence it's so you can have that threshold wherever you want.

B

That's beside the point, when we talk about activating many columns, that's going to be important, but all I'm all I keep thinking is.

B

B

And I just need to I need to my brain needs to think for a minute, because I'm confusing myself, this histogram is account. I keep forgetting like the histograms got yeah yeah.

B

I think I'm doing this wrong. I think.

A

I'm I think I've got I'm thinking.

B

Wrong in this test, I'm not doing this right. I definitely need to sort okay, okay, okay, yeah, so I definitely need to sort if I sort and there's a normal distribution. Definitely.

A

B

Person, the first ten is going to be way different than the last ten and then the middle 10, so I think that's what I'm that's what I need to test so I think I'm sort of on the right track, but the but I gotta remember that the histogram is a representation of the permanence. Is that it's not there? Not! The permanence is its account of the permanence. Is that that fit into the right? Bins: okay, like so I'm close I, think I'm close half sample index.

B

B

Okay, so let's, let's sort this first, let's say.

B

B

Sort copies the array right, I, hope, yeah.

A

B

I should be able to.

B

Slice slice, slice, I, better use the same thing: okay,.

B

Okay, so they should off by default, without creating, without specifically identifying the number of independent variables for the base distribution. They should all be basically averaged the same. They should all be pretty much the same, so assert close close to so when I say D, 3 dot mean first 10% or it isn't it didn't. Copy are good thanks.

B

This should be close to d3 mean last 10% by.

A

B

Of 0.1 I guess: let's see, if that, let's see if that path even runs okay, so it's not okay, great, so they're, totally different, totally different. Oh I guess this would always happen. Wouldn't it because, even if you have a random distribution, I mean you if you sort them, of course, the high values are going to be way different than the low values right.

A

Before sword, we had the burden of for sorting.

B

So this should this shouldn't be done on the sword. This should be done on just the NE.

B

Yeah I think you're right, I think we're gonna remove the sword, but it's still not sure I'm actually testing something I want to be testing if I have a random district and distribution. Oh wait a minute.

B

A

B

I got it I just had a thought go back to this, so we take this back to random and then, if I change this oh it does break it's not exactly what I wanted it.

A

B

The entire distribution.

B

I, don't know you're gonna to think I. Think on this.

B

If I specify that I want, the permanence is initially 85% connected and.

B

No 85% of potential pool that's different each one of those each one of those permanence is then I want to most likely be yeah. Okay, so here's the thing I want them to be it. If I give.

A

It an independent.

B

Variables is like 20, then most of those should be around 0.5. That's the thing, that's what okay! So, if they're all random and I sort them, it should be a straight line. Okay, that's that's! This is where I'm getting I'm.

A

B

Draw here, okay,.

B

So if everything is random and.

B

A

B

Plot them all and I sort them that should be like digital digital. Basically, I mean there's gonna be a lot of wiggle room here, but they should be sorted from you know one well. Actually it's gonna go the other direction right if I sit depends on how I saw it, but it's gonna go like this up to zero, but basically a linear plot.

B

This is if independent bars is one now, if it's not, if I give it a different, independent bars, because I want a normal distribution, if I give it, for example, 20, then this is this is out the window and what I should see you is.

B

B

I should see.

A

B

So this is just a distribution sorted by sorted by weight. It's going to go up.

B

You know I I, think I'm gonna bow out today. I'm really distracted by this neck pain and I, don't feel like I'm being very efficient right now.

A

B

I'm stuck on something I shouldn't be stuck on this. This isn't something I should be stuck on and I'm starting to think. It's probably because I'm having a hard time focusing so I hope you guys don't mind if I come back on another day and try and finish this up. It's only there's only three people watching anyway.

B

So I think I'm gonna do that sorry, David David I think I'm gonna, I'm gonna shut this stream down I'm gonna. Think of a little bit on this offline I'm gonna go to the chiropractor and and I'll be back. Then I'll be back on lines, probably definitely Thursday or Friday. Then we'll finish this out at least the permanence of potential pools and put them into diagram, but I think we're pretty close. It's just I'm having a hard time sealing the deal right now and, like I, said, I.

B

Think it's because I'm distracted by physical pain, all right so see you guys later. If you don't mind, I'm gonna take off and lay down or something appreciate it. Thanks for watching, I'll see you guys later close the thing. Oh here's.

A

My credits, I think I had a couple follows.

B

There be rousing and fixable thanks for following that beer brands in it yeah I will it's not in a good state, but I'll go ahead and push.

A

B

See you guys later.