What is LSTM (Long Short Term Memory)?

Рет қаралды 170,676

Күн бұрын

Learn about watsonx → ibm.biz/BdvxRB
Long Short Term Memory, also known as LSTMs, are a special kind of Recurrent Neural Network, or RNN, architecture capable of learning long-term dependencies as well as a solution to the vanishing gradient problem that can occur when training traditional RNNs.
In this lightboard video, Martin Keen with IBM, breaks down why we need LSTMs to address the problem of long-term dependencies, how the cell state and its various gates help transfer relative information in a sequence chain, and a few key LSTM use cases.
#LSTM #RNN #AI

Пікірлер: 83

@engr.inigoe.silvagalvan1161 Жыл бұрын

Very useful and helpul. This is a hard topic to understand as readily way, but you can did it in just 8 minutes. Thanks for that Mr. Martin and company. Greetings from Mexico!

@channel-xj7rp 2 жыл бұрын

So are going to ignore the fact he wrote everything backwards on a clear glass wall?

@IBMTechnology 2 жыл бұрын

That's likely the #1 comment! The presenter writes on the glass and we flip the image in post-production. Martin is actually right-handed. Some married presenters will switch their wedding ring to the other hand or not wear it so as not to confuse their spouse.

@gehallak660 2 жыл бұрын

@@IBMTechnology Smart ! I was wondering how you did this.

@abhishekrao6198 Жыл бұрын

@@IBMTechnology I don't understand how you guys made it look like he's writing backwards on a glass, could you explain it to me.

@chaoma8228 Жыл бұрын

@@abhishekrao6198 The presenter uses right hand to write. The camera is in front of him so that the original video should have all sentences reversed. Then flip the video left to right, you can get the video you watch now

@nitinkapoor4752 Жыл бұрын

@@IBMTechnology 😂😂…the most obvious observation. People who didn’t even figure that out 🤦‍♂️should not be watching LSTM/RNN or other such videos 🕺

@simonbax2002 Жыл бұрын

Martin you are a wonderful teacher! Thank you very much for the explanations.

@toenytv7946 2 жыл бұрын

Great advancement in time. Glad to have a better understanding. Thank you folks

@DieLazergurken 2 жыл бұрын

Very helpful lecture. Keep up the good work!

@amalkumar256 6 ай бұрын

I'm in love with his way of teaching!

@IsxaaqAcademy 26 күн бұрын

After watching Martin Keen's explanations, you don't need another short explanation

@vivekpujaravp Жыл бұрын

Fantastic explanation. Please keep making more.

@RishabKapadia Жыл бұрын

thank you martin and team. great work.

@theneumann7 Жыл бұрын

Great video and nice visual effects!

@fouziafathima6460 10 ай бұрын

Clear and concise explanation. 👍

@WangY-ip3sb 10 ай бұрын

Good lecture ! Thank you very much for the explanations.

@Gabi_09 Жыл бұрын

Very good teaching, thank you!

@phonethiriyadana 2 жыл бұрын

Thanks for the clear explanation.

@athmaneghidouche7746 Жыл бұрын

Very useful, plain, and concise 😀

@tammofrancksen5186 2 жыл бұрын

Really good video. Thank you!

@waleedt_trz 2 жыл бұрын

such an informative lecture, thank you so much

@siddhesh119369 Жыл бұрын

really good video, thanks you!

@kikichung2955 Жыл бұрын

Useful video. Thanks a lot

@balenkamal182 Жыл бұрын

Too many thanks, your lecture is very helpful, could you please explain all gates from LSTM (Forget Gate, Learn Gate, Remember Gate & Usegate(OutPut))?

@thomasplum9868 2 жыл бұрын

Super good lecture

@mastajigga13 2 жыл бұрын

Thank you for the lecture

@shilpadas1311 2 жыл бұрын

Great lecture, thank you.

@IBMTechnology 2 жыл бұрын

You're welcome and thanks for watching, Shilpa!

@akashthoriya 2 жыл бұрын

Good lecture, Please make video on "transformer based models" too, It will be very helpful

@IBMTechnology 2 жыл бұрын

Thanks for the suggestion, Akash! We'll see what we can do! 🙂

@massimothormann272 Жыл бұрын

Is there a way to use clustering and similarity measures to load relevant context? Say we have a Text corpus, cluster it and the lstm checks what the current topic probably belongs too and loads information about this topic into the State or as an extra input?

@jayasreechaganti9382 Жыл бұрын

Sir can you do a video of Rnn example by giving numerical values

@user-wq2sn4zb1d 7 ай бұрын

Hi Martin as you given an example of martin and jenifer, When questions asked about jenifer, martin is no more reelevant so iit'll be forgotten right, if anytime in the future question asked about martin does that relevant to the LSTM? I mean will it be able to remind about martin even after forgetting it?

@user-nz5tk9wm7f 2 ай бұрын

please teach us more content like this..

@mohammadrezarazavian9305 Жыл бұрын

That was great!

@kedarnandiwdekar20 Жыл бұрын

Thank you !

@bysedova Жыл бұрын

Tell me please, how is it regulated what we need to remember in sequence? How is it determined? That's how the model will determine what makes sense for the investigation?

@aimatters5600 Жыл бұрын

thank you

@rollopost Жыл бұрын

Great video - do you write on glass and then transpose/flip the video?

@IBMTechnology Жыл бұрын

See ibm.biz/write-backwards

@sathiraful 13 күн бұрын

are there any new algorithms more powerful and more efficiency work than traditional neural network ?

@colabwork1910 Жыл бұрын

Wao, Very helpful

@Grzeroli1 Ай бұрын

How does this trick work, you record normally and then mirror the video?

@trinity98767 2 ай бұрын

The video is mirrored image of the actual video. The writing appears straight to the audience.

@codeschool3964 6 ай бұрын

Thanks.

@ashleygillman3104 2 жыл бұрын

Wait the Homebrew Challenge dude does Deep Learning too?!

@TheHomebrewChallenge 2 жыл бұрын

LOL yes, Homebrew Challenge dude has a day job :)

@TaylorDeiaco 7 ай бұрын

Anyone else distracted by the fact that he's writing backwards? Great vids, keep it up

@IBMTechnology 7 ай бұрын

See ibm.biz/write-backwards

@wmxoae1237 Жыл бұрын

Who can please tell me what screen he wrote on, it is so cool!🤩

@IBMTechnology Жыл бұрын

See ibm.biz/write-backwards

@maloukemallouke9735 6 ай бұрын

tanks how you can apply lstm with times series?

@LakshmiDevi_jul31 10 ай бұрын

Cam we use rnn for ctr prediction

@gupsekobas2209 4 ай бұрын

how can you write backwards?

@matthewg7702 7 ай бұрын

awesome

@fatmamahmoud9148 Жыл бұрын

magnificent

@elevated_existence Жыл бұрын

How are you writing ?

@IconOfSeas 11 ай бұрын

saved my exam

@ivant_true 9 ай бұрын

what if Jennifer is they?

@willdrunkenstein5367 2 ай бұрын

Imagine if the model takes the context "My name " and predicts "J" as the next letter 🤣

@abdullahshaikh7409 Жыл бұрын

After first 10 seconds of this video me: Woah now I know the origins of knives out movie😂

@anirvinkandarpa5544 Ай бұрын

why echo though?

@amirarshiamirzaei710 2 ай бұрын

who would have guess that some algorithm can beat current state of art memory (eg me )

@petchpaitoon 2 жыл бұрын

The voice is too low

@akashthoriya 2 жыл бұрын

Yes needs to work on audio 🔉

@jagungkaka5977 2 жыл бұрын

Subtitle is a solution

@douglasespindola5185 2 жыл бұрын

Use earphones. I did and had no problem!

@BinodLaKshitha 8 ай бұрын

❤❤

@eyupozturk8586 Жыл бұрын

"always butler" is high bias :D

@willw4096 9 ай бұрын

5:00 5:42

@simplexination9837 Жыл бұрын

mirror writing🙃🙃🙃🙃🙃

@user-fp8jx7gr7v 6 ай бұрын

anyone here to know the murderer, its the butler🤣

@cybrhckr 2 жыл бұрын

I see the LSTM = low volume. This is the 3rd guide that is almost muted sound.

@jeverydk Жыл бұрын

Sound is just fine mate

@manmohanmahapatra6040 29 күн бұрын

VOICE too low!

@domnic7431 Жыл бұрын

Are u using mirror or can you actually write backwards sir 🤐🤐

@HITNUT Жыл бұрын

martin are you married or nah