Perceptron 

Mans explaining with 3 markers and a chopstick and its clearer than any professor could explain with slides and computers and simulations Youre an amazing teacher Thank you so much

i wish my professor could explain this. idk what she does wasting hours of our time every Tuesday. i leave knowing less than this 13 minute video

Quit reading the comments, boys, because this video will be all you need if you just focus on it. Great explanation!

the only perceptron video I needed, btw I have exams tomorrow

I confess, I didn't understand any single word of explanations from my MIT professor about perceptron. How ever, after I saw this video and understood clearly what the idea is. Thanks.

Incredible how an explanation with a piece of paper can be more effective than any other digital methods. Thanks

yo this η is an eta, not a nu lol

You're really a talented teacher! I love the way you walk us through such abstract concepts. I was struggling with my HW for a whole day until I saw your videos. Keep producing more brilliant content! Subscribed!

Amazing clear, and concise explanation! Thank you so much, I'm trying to program a perceptron neural network in Go and this really helps

wow great video! Way easier than how my professor explains it! My prof uses terminologies, which he doesn't explain, and it gets confusing at times.

I’ve been trying to understand perceptron for a couple days now. Nothing really clicked until this video. 10/10 explanation, you are better than majority of the professors and teachers out there.

Hi, amazing video! Quick question: Is the learning rate represented by the Greek letter "eta" or "nu"? Because your drawing ressembles "eta", and you wrote "nu" beside it. Thanks

thank you! Well explained. A small question about the symbol of learning rate. Is it called 'Eta'? 'Nu' is the v shape one?

0*1 + 1*x1 + 5*x2 > 0 1*x1 + 5*x2 > 0 x2 > - 1*x1 x2 > -(1*x1)/5 x2 > -0.2*x1 Is that corrct or did I make a mistake?

random guy with a chopstick and marker better than most professors at my uni

07:29 By far the best use of a disposable chopstick in a ML video. Well done!

such a clear explanation! you're a natural at teaching and machine learning. thanks so much

How you get x1 = 2 ; and x2 = -2 ?

Thank you man , my machine learning class is a mess .

Thank you for the video! It is well explained. Some lecturers just have a passion for making easy things difficult.

You assigned the omega values for the question but how did you initially assign their values and why did you assign them as 0,1,.5 why not something else ?

Great video, you pretty much have explained the whole perceptron problem in 15 minutes while I was trying to figure it out after taking 4 hours of uni classes.

Thank you for taking the time to make these videos and share your knowledge, your other videos deserve more views.

great video. thanks !

It would be better for the beginners if u speak slowly always deliver ur lecture while speaking slowly

at 6:15, from where this equation derived X2 > -2X1

So what if 2 points are incorrectly seperated? Do I just pick one at random to update?

5:46 How is 1 + x1 + 5x2 > 0 simplified to x2 > -2x1 ?? They are two very different expressions.

one of the best ml video

Cool! Thank you! ❤

This video just solved one assignment I had from university. Good explanation, thanks

You could have made it better by differentiating as a symbol( sample) and axis ... Symbol could be a circle and triangle rather x and triangle ... the word x interfering with the word x1 and x2 . Otherwise consciously need to remind yourself x and (x1, x2) are different

Need teachers like this guy teaching like we're all bloody five instead of my lecturer who decides to fill the first half of the powerpoint slides with biological neuronal networks ffs

This is amazing, I've been scratching my head with lecture videos and notes.. going nuts.. but this explanation is simply super... I have been looking for this to get the concept and fundamentals. Love it. Please keep posting more and more..

Still wondering how you convert "0*1 + 1*X1 + 0,5*X2 >0" to "X2>-2X1". If I let ChatGPT draw the line it looks completely different.

8:47 but what about when you have more than 1 moitake and some should be in the upper and some in the lower?

REALLY GOOD VIDEO. Thank you so much, why can't our teachers at university explain like this.

Excellent video! But could somebody help me with my questions please? 1. At 5:32, how did you know the equation is bigger than zero since x1 and x2 are unknown values? 2. If let's say we have instead of 1, we have 2 misclassified points. So shall we calculate two w prime? If so, how do we use the two different w prime to get the hyperplane?

Hey Ritvik, thanks for this video! I was wondering how will the parameters be updated if we have multiple points that are wrongly classified?

Super helpful explanation. Thank you! Is the updating the same as back propagation, or is that only with multiple layers?

How is the initial w vector established? That is, why did you use w = ( 0, 1, 0.5) ?

Wow...never had anybody explain things so clear...just loved and learnt a lot about perceptron!! Thank you!!

So help me understand this, anyone. Lets suppose I want the perceptron to classify spam and not spam mail. Then, by a point (on the plot) do we mean a single email? Also, does the update of parameters happen after each mail is classified or after the entire training set of emails is classified? From the video, I get the impression that update happens only after all emails are classified. How do we update params if there are multiple wrongly classified points?

Took me a while to understand which greek letter "nu" is. It's the greek letter "eta". You may see in some programs that the learning rate is called eta instead.

Thank you very much for the insightful video, very well explained, just a nitpicking remark, the greek letter you showed at the beginning is eta and not nu. nu looks like a v

Maybe not chose one of the symbols on the graph as a cross so it doesn't further confuse someone with all the x. And x1 and x2 are harder to distinguish then x and y.

Thanks for the video. Two remarks: 1. It is not clear if the algorithm converges to a solution. 2. mistake at 10:15-10:34: the new line found by the algorithm does not pass throughout the origin...

thanks brah

1 million thanks

silly qn here, why you chose >0 as X and

Thank you for such a clarify video. I'm a bit confused and hope you might explain it. Does the final hyperplane go through the origin (0,0). If it does not, how could we make it go through the origin? Thank you.

"Linearly Seperable" is the most pretentious nerdy terminology I've heard in years.

something about the chopstick makes me understand your explanation better

bro its crazy how ive been learning this in class for the last 2 weeks and I didn't understood them until now

8:49 how do you define upper and lower? If the line is mirrored, is the upper region still the 'right' side of the line, or the 'left' side

i wanted to ask one question. isn't d= target value- expected value?

Is the change of coordinates from cartesian to polar related to the so called 'kernel trick'?

How would we know if the data is linearly separable or not before we use this algorithm. And also how would we know if we are supposed to use polar co-ordinates ? we should know beforehand then that the data can be separated through a circle. not a good video though

The learning rate is NOT a parameter. It is a hyperparameter. Please be careful in the terminology.

So how would you know to use a perceptron if you were working with a high dimensionality? Not like you can plot it all.

Would polar coordinates work if the circle isn't centered around 0,0?

Ritvik, I saw the thumbnail and title and thought I was heading into a Bill Wurtz video 😂

Great video! Small correction, η (the learning rate) is the letter "eta", not "nu"

What if we had two or more misclassified points? How would the w(i) be updated?

Very good video. The learning rate is the greek letter "eta", the letter "nu" looks more like a v.

Really great explanation! Loved it!😍

..........That "ν" you wrote at 2:30 looks very "η"-like 🤨

converting data to polar coordinates is just beautiful

tha goat

Thanks, i haven't seen this one before. If there are multiple lines that will separate the classes, is the line selected by perceptron optimal in anyway, or does it simply stop when any viable solution is found? If there is overlap between the classes, will this still work? Assuming there is no one line that separates the classes with 100% accuracy, will iterating get you a "better" solution?

most aesthetic handwriting i've ever seen

How you got x0 , x1,x2, values at (09:32)

So perceptron in machine is the same thing with neuroplasticity in humans?

how do we select the value of d ("upper or lower") in higher dimensions

Great video! I'd like to know how the addition of another element to the number of dimensions, gives us a constant number? How does just the no. of Dimensions refuse to provide a constant number?

What if dataset 8s separable only by a curve not a circle.

I think that's an Eta, not a nu, but that's a GREAT explanation!

why would you called it as weight zero if it's literally the bias

Very good video,thank u very much!

عالی بود. دمت گرم.

what if my data has a triangular shape? then what ML algorithm should I use? | | xxxx xxxx xxx | xxxxxxxxxxx | xxx xxxxxx | xxxxxxxx | xxxxxx | xxxx | xx |____________________________________________ the Xs' represents points on the graph

It helped me my friend, really appreciate that. :)

Great video; the nu at the beginning though... that's eta not nu

Great explanation but does bro not have a ruler lol

excellent video! thank you

you are so amazing you are reading 3 years into the future

Just amazing. Please keep up the good work, it is this kind of clarity that makes people like topics like this.

Thank you sir!

Great video straight to the point thanks

??? Writing an eta and reading it nu???

can please start uploading these pics

Thank you so much

Why we need to do draw the line to seperate different classes

Learning rate isn't a parameter for a perception.

Not the hero we deserved, but the hero we needed

This is one of the best best best channel for learning ML

Pretty sure learning rate variable is eta, not nu.

Very useful for my exams..... Thanks a lot sir

its called "eta" not "nu" I believe