Sensor Networks

Подписаться 770 тыс.

Просмотров 26 тыс.

50% 1

What is the best way to route data in a network of routers spread out across the globe? This 'internet of things'-based problem can be solved using reinforcement learning! In this video, i'll explain the 2 types of policies, the bellman equation, and the value function. All of these concepts are crucial in the RL pipeline and using animations + code, i'll break them down. Enjoy!
Code for this video:
github.com/llS...
Please Subscribe! And like. And comment. That's what keeps me going.
Want more education? Connect with me here:
Twitter: / sirajraval
Facebook: / sirajology
instagram: / sirajraval
Github Syllabus:
github.com/llS...
Take the full course at the School of AI:
www.theschool.ai
More learning resources:
becominghuman....
medium.freecod...
www.oreilly.co...
kvfrans.com/rei...
/ basics-of-computationa...
www.toptal.com...
www.wildml.com/...
Join us in the Wizards Slack channel:
wizards.herokua...
And please support me on Patreon:
www.patreon.co...
#SensorNetworks #SirajRaval
Signup for my newsletter for exciting updates in the field of AI:
goo.gl/FZzJ5w
Hit the Join button above to sign up to become a member of my channel for access to exclusive content! Join my AI community: chatgptschool.io/ Sign up for my AI Sports betting Bot, WagerGPT! (500 spots available):
www.wagergpt.co

Опубликовано:

19 сен 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 82

@manwuzi 6 лет назад

Siraj is buffing up for the when the machines take over.

@mulanszechuansauceisthemeaning 6 лет назад

hahahahaha

@alienkishorekumar 6 лет назад

Yeah

@SirajRaval 6 лет назад

haha. i am trying to keep my body as healthy as possible so i can continue to churn out content at the rate i am currrently

@malharjajoo7393 6 лет назад

Very nice approach.

@keepforever726 6 лет назад

And the award for best sentence in this video goes to: "As long as the agent learns an appropriate action response to any environment state that it can observe--we have a solution to our problem." I love the saying "standing on the shoulders of giants". It's dizzying to think about the succession of shoulders upon which that sentence (rather, the usefulness as you understand it to be fit for) finds it's footing. Centuries of knowledge cooked up in a rusty digital tablespoon called youtube and injected straight into the brainstream.

@SirajRaval 6 лет назад

Wow. Deep stuff and true

@yokonthesage 6 лет назад

What kind of fetish does this guy have

@SirajRaval 6 лет назад

everything

@aesthetic4866 6 лет назад

nice gainz, bro

@JewishSpy 6 лет назад

I learn better when aroused, thanks Siraj senpai

@alienkishorekumar 6 лет назад

You've been hitting the sugarwalls

@2252chelo 6 лет назад

Siraj, at 10:35 you are referring to the q-function (action-value function) but you are showing the state-value function in the video, it would be better to clarify that so no one gets confused

@devonk298 5 лет назад

Siraj shirtless and tied up. Very nice!

@tedp9146 6 лет назад

First ten seconds cringed me away for this day

@chinmaykunkikar 6 лет назад

First 3 seconds. I paused the playback since.

@SirajRaval 6 лет назад

Tarter sauce! well, noted. i'd really encourage you to skip it, theres some seriously good content in this video

@KunwarPratapSingh41951 6 лет назад

It's not cringy tho.

@KunwarPratapSingh41951 6 лет назад

Imagine that spy as reinforced terminator

@rajatkantibhattacharjee4599 6 лет назад

The difference between state action and value based optimization was really well put. I was trying to frame the difference for a while :-D ... Nice ... But still if you could explain with a bit more explanation regarding the mathematical foundation and formulation. Especially the discount factor affect in the stochastic environment.

@yitto4965 6 лет назад

There’s been a routing protocol called “OSPF” that has been doing this for decades. Reinforcement learning re invents the wheel

@magxtopher1332 6 лет назад

You the man and keep on educating us.I appreciate.

@Kingstanding23 6 лет назад

The two types of value functions are; state value = V(s), where the value is dependant on the state only, and state-action pair value = Q(s,a), where the value is dependant on both states and actions.

@SirajRaval 6 лет назад

correct

@wolfisraging 6 лет назад

It really touched my senses

@SethuIyer95 6 лет назад

1:00 that meme caught me off guard xD

@aradarbel4579 6 лет назад

That is an interesting subject, and an interesting intro... but still! I like the video! ;D

@xiubinzheng7 6 лет назад

came here to learn about machine learning, ended up looking for weight training tips.

@arcade-fighter 6 лет назад

Bellman Equation is based on dynamic programming which actually is backward induction and it is related with the stopping problem. e.g. how many job candiates do I have to interview to get the optimal one?. Another example would be the valuation of an American Style Stock Options. Can we use reinforment learning to valuate finantial instruments?

@larryteslaspacexboringlawr739 6 лет назад

thank you for sensor network video

@vipindube5439 6 лет назад

Nice work

@0L1M 6 лет назад

thanks for a great video! but in the end did you mean: "to learn an optimal policy, we need to learn an optimal value function, of which there are 2 kinds: state-VALUE (not state-action) and action-value" ?

@abeaumont10 6 лет назад

As always, a great video

@anonyme4778 6 лет назад

Greeting from algeria

@hardikaggarwal7426 6 лет назад

Awesome job bro

@AbhishekKumar-mq1tt 6 лет назад

Thank u for this awesome video

@RandomGuy-hi2jm 6 лет назад

Its 5:AM in India. 😃

@SirajRaval 6 лет назад

good morning india

@smartsniper268 6 лет назад

Siraj can you show the rate at which your subscribers are increasing... Through calculus😂???

@vincesanityyy 6 лет назад

Hello siraj. How do i improve my neural network accuracy? Can i use some algorithms to it?

@masterjiggle5728 6 лет назад

How is it beneficial over Djiktra algorithm? It also routes for the best cost which was reward in this case??

@dinnerplanner9381 5 лет назад

How is this different that Dijkstra's algorithm for shortest path?

@eigerultra6478 6 лет назад

Please explain ECMP.

@sandeepkrishnan3696 6 лет назад

Hey siraj! ❤

@melshakobyan8670 6 лет назад

daaaaam, gaaainzz

@cu7695 6 лет назад

Epic beginning !

@shanyaanand7336 6 лет назад

@Siraj Raval can you please explain thing in a more mathematical way, it is easy to understand in terms of the equation rather than just intuitive feeling, thank you

@SirajRaval 6 лет назад

sure

@Verrisin 6 лет назад

0:20 - um, there's already an algorithm for it. All networks, routers and PCs do it automatically... --EDIT: Ok, sorry. It's not about that. I haven't watched far enough. - Interesting video!

@Verrisin 6 лет назад

Geez, you learn this algorithm in first semester informatics class. It works a lot better than hops too. You compute how many ms to each peer, then it propagates to all your neighbours, and then you know from whom came the shortest to X, so when you want to route something to X, you send it there.

@Verrisin 6 лет назад

oh, yeah, it's called Distance Vector. - I guess there are other that are more dynamic, but this is not a problem that needs machine learning...

@Verrisin 6 лет назад

Ok, sorry. It's just an example to show the general idea.

@MrTk3435 6 лет назад

Siraj, I am curious, How would AI Winter play out? Thank :-)

@salmx__4243 6 лет назад

From lavendaire life style

@jaddasivakrishna5351 6 лет назад

Hiiii Siraj sir, I want to learn artificial intelligence with python.please suggest the best online course

@subschallenge-nh4xp 6 лет назад

i think that if your videos would be *2 speed you could get more viewers because i alweas do it and then i notice the diference

@CryptoRootz 6 лет назад

Dam Buff Guy ..

@nnslife 6 лет назад

You know, you should take this online Python course on Udemy

@bhuvaneshs.k638 6 лет назад

Bro u r truly amazing....!!! Siraj can u try to do a video on building an OCR(optical character recognition) Engine like Tesseract using Deep Convolution sequence

@ilyasmax4778 6 лет назад

why not use short path algorithem ez and not that computational complex

@gustavomartinez6892 6 лет назад

Siraj why don't we use the data set of the pornography sites, I mean they should know about dopamine, and the reinforcement learning more than any psychiatrist or any other doctor, why don't we ask them? Instead of JUST looking at the mathematics and formulas?

@navaneethkt4289 6 лет назад

first

@barath_ 6 лет назад

Wtf man!😂😂😂 I guess then everyone's gonna become a reinforcement learning agent.

@reyariass 6 лет назад

That’s hawt

@starinsky2873 6 лет назад

Siraj are you programmer I want to know that fact please!

@teegnas 6 лет назад

Anyone please comment, if you found out the reason behind Siraj winking during the video

@gustavomartinez6892 6 лет назад

Hahahaha secret agent...

@prajwalmishra1475 6 лет назад

theschool.ai is still not working

@teegnas 6 лет назад

Bro, use theschool.ai It didn't work for you earlier, cuz you were not using secured http

@prajwalmishra1475 6 лет назад

@@teegnas is it working for you

@teegnas 6 лет назад

@@prajwalmishra1475 yes, did you try the https protocol

@prajwalmishra1475 6 лет назад

@@teegnas ya

@SirajRaval 6 лет назад

now it works

@chrisBruner 6 лет назад

You are speaking slower. Thanks, that helps assimilate these concepts.

@sakyaris2458 6 лет назад

1st

@shobhitsundriyal4822 6 лет назад

RL agent 🤣🤣 0:03

@XRagnouX 6 лет назад

It's not even a hidden agent tho

@TheMultipower47 6 лет назад

Remove the text from the thumbnail and the video will take on a new tone.

@ashishsinha8893 6 лет назад

Hey bro please make a video on realtime project which can I show in resume

@cu7695 6 лет назад

You need to implement the code to show it on resume. Video might explain theoretical concept but tuning hyper parameters & making efficient tensors is where you shine as data scientist.