+Tyler Waite Same here and I agree with you. It is the best. Takes you from abstract idea to deeper concepts for better and easier understanding. What other courses did you take btw?...I could use some if there is any I didnt come across on my own.
Besides this one, I mainly looked at the Stanford and MIT OpenCourseWare on TH-cam. It also looks like there is one my mathematicalmonk (looks similar to Khan Academy) that I would definitely look into if I had more time.
Great thing is i can learn Machine Learning from one of the best professor in the world and bad luck is even not single professor aware about ML in my collage.
If you dont understand everything watch it again I some how missed all the important parts the first time and was confused but then I watched it a second time and it was amazing. I also like how he is smiling if he gets a question he can answer. :D Great lecture.
It's good to hear this. Yeah I do the same. I'm sick of all the people on here saying they watch it at 1.25 speed. I guarantee I am getting a better job than those people.
This kind of learning resource is useful for providing professionals with the right intuition and conceptual framework in the technology of Machine Learning. More power to California Institute of Technology and Professor Yaser Abu-Mostafa.
Great video! I like the puzzle part (53:00) This is the reason why I don't like most IQ tests. They show you just 5 sometimes even unordered pictures and want you to choose the one which belongs to them. There are many solutions usually.
who on earth disliked and downvoted this video?..the professor's teaching from just the first lesson has inspired me to research into other ML ANNs beside just the perceptron like the Hopfield NN and gain a megatonne of info...from just the first damned lesson..and i was simply clicking on recommendations goddamit...had no clue about ML..those damned lucky students
Seriously, this is the best lecture on #ML available on web. Brilliant explanation and the examples at the beginning (specially the Credit approval one) is well explained and fits exactly into introduction to ML.
Straight to the point with practical examples. After 10 minutes you already know what the course is about and if it is interesting to you. The best Lecture 1 of a course I have seen on any subject :)
The perceptron function hypothesis is actually a logistic regression, that guy makes it quite simple to understands the difference between the hypothesis/target function and the learning algorithm, the later is the true innovation for statisticians and econometricians.
Que obra! Por sua naturalidade de conhecer profundamento o que ministra, com generosidade e vontade ao ensinar, que acaba por nos revelar que o mestre é insubistituivel !!! Por qquer processo racional que seja, como eu depreendo daquilo que disseram Gödel, Newton da Costa, Fco. Dória e ai vai ... mestres ... E faz de um autor e professor como este um daqueles que a gente sente que nos liberta pela informação e não o oposto!
Examples of (h, a) where h belongs to H and a belongs to A: (Perceptron Model, Perceptron Learning Algorithm) (Neural Network, Back Propagation) (Support Vector Machine, Quadratic Programming)
At 21:36, it seems to suggest the only changeable parts of the ML flow are 'learning algorithm' and 'hypothesis set' which are to be decided by the analyst. But what about feature extractions? Two analysts can be given the same data set but one can extract a better feature set based on a given X, for example.
Yes and if you define the hypothesis set even if it just includes all possible solutions it will only work in a formal system ie a chess game or a game of go. Is it even learning at all? It is certainly not intelligence. In fact it is merely automation of actions defined by humans. Intelligence is the ability to find form a formerly undefined system. Humans are very good at it and do so very quickly. The difference between humans and machines is 4 billion years of intelligent (biology finds form out of a formless system) evolutionary learning that resulted in receptors and effectors that interact with the world. My research points to the fact that the actual human learning algorithms are extremely simple. The complexity comes from wiring 15 billion sensory and motor neurons into the brain. This simplicity is inevitable because neural intelligence emerged out of evolution so has to be based on simple units. The eye is a pin hole camera. The brain is a network of weighted balances provide more, less, the same information in continuous fashion. www.legonomics.org
Just started taking the online course and it exceeded my expectations. Love your teaching method and the "practical" topics included. Also thanks for including the QA session.
The sound cuts out briefly at 46:10 and at 46:13. At first I thought it was a buffering glitch on my phone. I re-observed the glitch on both my laptop and kindle.
At 31:14, what does he mean by the "angle" of the two vectors, x and w? w is an array of weights, and Xn is an array of specific values for the "customer n". How does that translate to vectors in a space that have an angle between them?
solidox cinci you can treat bold_w = (w1, w2, .... , wN) (you call it array of weights) as an N-dimensional vector. Values of components of bold_w will be coordinates of your vector in corresponding dimensions. The same applies to the array Xn. So, you will have two vectors in N-dimensional space. And with them you can define an angle between them, count dot product and so on... I hope this was helpful.
But bold_w and Xn are N and M dimensional vectors, respectively, of completely different sizes and dimensionality. It feels like computing the angle between apples and oranges.
solidox cinci In earlier moments he defines the model as SUM(xi times wi) where i = from 1 to d. This means that and bold_xi and bold_w are in the same d-dimensional space.
@32:02 are the diagrams flipped for y=+1 and y=-1. if the dot product of w and x was positive the y would be positive not negative? but in the top image w and x are not facing each other
This is what I call "A real Professor" , congratulations.
Can you imagine this material was on TH-cam since 2012! Thanks Caltech and professor Yaser for making it free.
Prof. Yaser is a great teacher!!
Yaser, you are not just a scientist but also an amazing teacher -- your students must be lucky to have you!
I have gone through many courses attempting to learn machine learning (no pun intended), and this one is definitely the best. 10/10
+Tyler Waite Same here and I agree with you. It is the best. Takes you from abstract idea to deeper concepts for better and easier understanding. What other courses did you take btw?...I could use some if there is any I didnt come across on my own.
Besides this one, I mainly looked at the Stanford and MIT OpenCourseWare on TH-cam. It also looks like there is one my mathematicalmonk (looks similar to Khan Academy) that I would definitely look into if I had more time.
Thanks man...I had never heard of mathematical monk before. Glad you mentioned it. Surely gonna follow his videos.
one by Ali ghodsi and Nando de freitas are also good
My teacher is apparently using this exact course as a basis. The power points are super similar and all!
Prof. Yaser is a great Teacher. I'm seeing this for the first time (October 2019). Still the best.
After Gilbert Strang's Linear Algebra course, probably the best course on the web.
Definitely agree
High praise; so far it seems so
Very proud of my country Egypt and of the brilliant prof: Yaser
مش متخيلة انا قاعد مبسوط منه ازاي طول المحاضرة
جودة المحتوي مذهلة ما شاء الله
One of the best teachers I have ever seen! Congratulations, Professor Mustafa!
2021, and still it's a diamond. Kudos to the teacher.
this guy knows a lot about a lot, you can tell by the confidence with which he navigates the subject and how he is aware of the limits of the theory
Great thing is i can learn Machine Learning from one of the best professor in the world and bad luck is even not single professor aware about ML in my collage.
You should try a college instead of a collage, you may have better luck.
If you dont understand everything watch it again I some how missed all the important parts the first time and was confused but then I watched it a second time and it was amazing. I also like how he is smiling if he gets a question he can answer. :D
Great lecture.
I THINK all machine learning lectures have to be watched twice :D
It's good to hear this. Yeah I do the same. I'm sick of all the people on here saying they watch it at 1.25 speed. I guarantee I am getting a better job than those people.
48:05 It's some kind of revelation. I'm enjoying this lecture so much! Thank you!
And Yaser is so intelligent and pleasant person.
For a beginner this is the course I find the best to start with. Thank You Professor.
Am I the only one who's addicted to Prof. Abu-Mostafa's adorable accent?
he is egyptian BTW
I am bro
No, actually it's hard to understand. The reason I endure his accent is his math talent, which is the most important factor.
He sounds like king julian from madagascar
Nope. It's a really nice accent. Makes everything sound friendly ^.^
This professor is awesome and funny at the same time.
In 2022 still I find his way of teaching way more informative rather than any recourses online or at Uni . great job
I had to come to comments section and search the man behind this lecture.
A wonderful lecturer wish all teachers could teach like him
00:00:27 - Outline of course
00:02:37 - Lecture 1 starts here
00:03:14 - Outline of The Learning Problem
My Learning Algorithm predicts that the professor loves his tea.
This is what I call "A real Professor" , congratulations.
Crystal clear exposition! This is the best intro to ML I have seen on YT.
Perfect way of giving lesson.. I have never understood ML course this way
l came here by chance and i surprised that this prof is Egyptian ,i am really proud of you sir you have a great style in teaching
Excellent lecture. Way better than stanford because of low resolution and poor audibility.
Thank you Professor Yaser Abu-Mostafa and to all you collogues and contributors. Most helpful and enjoyable. I'm impressed with Caltech.
This kind of learning resource is useful for providing professionals with the right intuition and conceptual framework in the technology of Machine Learning. More power to California Institute of Technology and Professor Yaser Abu-Mostafa.
The professor explains everything so clearly! The video and audio have high quality! The questions asked by the students have high quality!
Great video! I like the puzzle part (53:00) This is the reason why I don't like most IQ tests. They show you just 5 sometimes even unordered pictures and want you to choose the one which belongs to them. There are many solutions usually.
I'm just getting started learning ML. This lecture provides an excellent overview of the field.
This is the best Machine Learning course. 11 out of 10.
who on earth disliked and downvoted this video?..the professor's teaching from just the first lesson has inspired me to research into other ML ANNs beside just the perceptron like the Hopfield NN and gain a megatonne of info...from just the first damned lesson..and i was simply clicking on recommendations goddamit...had no clue about ML..those damned lucky students
Meus sinceros agradecimentos aos tradutores, ao caltech e ao professor. Conteúdo de qualidade e gratuito.
Seriously, this is the best lecture on #ML available on web. Brilliant explanation and the examples at the beginning (specially the Credit approval one) is well explained and fits exactly into introduction to ML.
It was great presentation. Kudos to Prof Yaser Abu-Mostafa for the same. It brought clarity on the fundamentals which were not clear from the books.
Wow, I could not find any other resource that explained the intuition behind using the inner product. What a great teacher.
My Learning Algorithm predicts that the professor loves his tea.
I think it rather "prefers tea to other drinks"
He might be the mathematical equivalent of King Henry, for Professor is one of the best AI teachers ever and King Henry loved his tea
Awesome teacher, teaching from realization...everything he makes so easy...kind of idol to me, added with his smile and ascent ...yes Sensational..
Anyone watching in 2024?
Yes, please add more videos on Al
Yeah
yup
👍
Hello
Big appreciation for whoever uploaded this video, It's such a huge help.
Straight to the point with practical examples. After 10 minutes you already know what the course is about and if it is interesting to you. The best Lecture 1 of a course I have seen on any subject :)
I think it is the best machine learning introduction I have watched so far
The perceptron function hypothesis is actually a logistic regression, that guy makes it quite simple to understands the difference between the hypothesis/target function and the learning algorithm, the later is the true innovation for statisticians and econometricians.
One of the most wholesome lectures I have ever experienced, and especially useful for me and my particular situation. Thank you very much indeed!
Love this. Thanks for the lectures CalTech! Perfect for the stay at home, wannabe grad student. (Like me!)
Best lectures of machine learning with clear voice :) Thanks Professor
Enlightening analogy on unsupervised learning (learning Portugal language). Great teacher!
what a nice Professor : ) Keep going forward to deliver such an amazing lecture.
Que obra! Por sua naturalidade de conhecer profundamento o que ministra, com generosidade e vontade ao ensinar, que acaba por nos revelar que o mestre é insubistituivel !!! Por qquer processo racional que seja, como eu depreendo daquilo que disseram Gödel, Newton da Costa, Fco. Dória e ai vai ... mestres ... E faz de um autor e professor como este um daqueles que a gente sente que nos liberta pela informação e não o oposto!
Good lesson, really can learn a lot. The best part is easy to understand and very interesting
Amazing, best introduction to ML I have heard so far
جهد رائع و عمل عظيم .. مشكور دكتور ع الشرح الواضح
Wow, superb lecture. Decided to watch the whole series. Thanks for uploading it.
Ray Romano knows a lot about machine learning.
+Matt McConaha : Raymond, Raymond, Raymond. Everybody talks about Raymond. Only Raymond. E....EEverybody loves Raymond.
+Matt McConaha Na, that's a young Netanyahu...
Don't compare this great teacher with a genocidal war criminal @@putinscat1208
This lecturer is incredible
Very well done. Obviously Prof.Yaser has an in-depth knowledge and has put in lot of thought in creating this class. Good work, Sir.
Many thanks for making this treasure online. Clear and straight forward.
Great first lesson. I'm trying to get better at "skipping the introduction", and this is one instance I'm glad I did. Thanks! :)
god has a new name and he has research & teaching interests in machine learning- simply wow Professor Yaser Abu-Mostafa. thanks and best regards
I like the examples given by the professor! Vivid and thoughts provoking!
I encourage people to watch this lecture at 1.25x speed.
Jim Gorman I stepped it up to 1.5x :)
Jim Gorman Very useful tip ! I recommend it too. :)
Iqbal Nouyed Gunna one up you at 2x :P
Jim Gorman As if these people don't talk fast enough.
+Jim Gorman I kind of like this guy's voice though, and it doesn't sound as good at 1.5x speed.
Examples of (h, a) where h belongs to H and a belongs to A:
(Perceptron Model, Perceptron Learning Algorithm)
(Neural Network, Back Propagation)
(Support Vector Machine, Quadratic Programming)
Greatest professor ever seen. Thank you!
At 21:36, it seems to suggest the only changeable parts of the ML flow are 'learning algorithm' and 'hypothesis set' which are to be decided by the analyst. But what about feature extractions? Two analysts can be given the same data set but one can extract a better feature set based on a given X, for example.
+Yellowknight888 Then that would be a different learning set (xn,yn).
Yes and if you define the hypothesis set even if it just includes all possible solutions it will only work in a formal system ie a chess game or a game of go. Is it even learning at all? It is certainly not intelligence. In fact it is merely automation of actions defined by humans. Intelligence is the ability to find form a formerly undefined system. Humans are very good at it and do so very quickly. The difference between humans and machines is 4 billion years of intelligent (biology finds form out of a formless system) evolutionary learning that resulted in receptors and effectors that interact with the world. My research points to the fact that the actual human learning algorithms are extremely simple. The complexity comes from wiring 15 billion sensory and motor neurons into the brain. This simplicity is inevitable because neural intelligence emerged out of evolution so has to be based on simple units. The eye is a pin hole camera. The brain is a network of weighted balances provide more, less, the same information in continuous fashion. www.legonomics.org
Excellent. Thank you Caltech and Prof Yaser Abu.
What a great teaching..!!
Salute you Sir..
Very nicely explained and taught..
Thank you..
The best lecture series I have seen so far.
Just started taking the online course and it exceeded my expectations. Love your teaching method and the "practical" topics included. Also thanks for including the QA session.
This was a brilliant introduction to machine learning !
Thank you Prof. Yaser, very useful lecture. thanks for sharing
Sensacional!
Neural Nets ?
Ohh tu aí, mano. Sou seu fã, me ajudou em muitos mapas de Karnout na faculdade hehe
Man... They had a better teacher in 2012 than ive got in 2020
Very good lecture!!! He is also answering questions from the online audience!
2021 , you are still the best lecturer
Top Class Lecture. Great Professor. Brilliant.
now this is a great teacher !
Breathe of fresh air. Thank you sir
Awesome Video.. A must watch for any person interested in Machine Learning OR any person who wants to get interested in Machine Learning :)
The sound cuts out briefly at 46:10 and at 46:13. At first I thought it was a buffering glitch on my phone. I re-observed the glitch on both my laptop and kindle.
i think one of the best introduction videos to machine learning! :) thanks a lot
Exceptional professor indeed !!!
Wonderful! Many thanks to the channel for providing subtitles in Portuguese.
Very simple, practical and understandable. Thanks
I like his pronunciation. I'm even able to listen to him without subtitle.
Excellent Introductory Lesson Professor :)
Thank you for uploading it online :D
Absolutely well done and definitely keep it up!!! 👍👍👍👍👍
thnk u caltech n honourable professor for this noble work
The Great Lecture Sir ,You have put in lot of thought in creating this Presentation
. Brilliant work !!
Excelente la introduccion al curso. Muy buen profesor.
What an Excellent Professor!, Caltech Please upload more courses.
Thank you for this great lecture.
This man is great! Hats off!
At 31:14, what does he mean by the "angle" of the two vectors, x and w?
w is an array of weights, and Xn is an array of specific values for the "customer n".
How does that translate to vectors in a space that have an angle between them?
solidox cinci you can treat bold_w = (w1, w2, .... , wN) (you call it array of weights) as an N-dimensional vector. Values of components of bold_w will be coordinates of your vector in corresponding dimensions. The same applies to the array Xn. So, you will have two vectors in N-dimensional space. And with them you can define an angle between them, count dot product and so on...
I hope this was helpful.
But bold_w and Xn are N and M dimensional vectors, respectively, of completely different sizes and dimensionality. It feels like computing the angle between apples and oranges.
solidox cinci In earlier moments he defines the model as SUM(xi times wi) where i = from 1 to d. This means that and bold_xi and bold_w are in the same d-dimensional space.
Vahe Hakobyan And since they are of the same dimension, the inner product is basically just "Are they facing the same way with respect to the x-axis?"
Very interesting and practical, a good start today!
Excelent, best lecture on internet 10/10
I finally got the book and I'm glad I did.
Great lectures，wish more videos from the professor
Great course, in depth coverage, and wonderful presentation.
thanks sir,this lecture is very helpful for every beginers
@32:02 are the diagrams flipped for y=+1 and y=-1. if the dot product of w and x was positive the y would be positive not negative? but in the top image w and x are not facing each other