One year ago I was watching your tutorials on how to draw squares on a canvas with code. One year later I'm trying to build machine learning models, also with the help of your tutorials. I'm not even a CS student, I'm a pianist!
I'm here because the coursera course instructor in the class I'm taking just can't explain it with as much joy and happiness as you. I feel like the information I'm getting here is paired with enthusiasm and that's the way it should first be introduced to my brain.
Taking it back to Boolean Algebra makes it very clear why MLPs are a natural solution to the XOR problem, thank you. Nobody's done that yet in anything I've seen, even though it's obvious in hindsight and maybe should have been obvious in advance.
Amazing teacher.I have my semester exam tomorrow and was searching a lot about multi layer perceptron on the internet and wasn't able to find good explanation.thank god I found your video.💙
Linearly separable and linearly not separable explanation is the best explanation! It's now logical, why there are multiple layers was required! GREAT! Thank you!
I have been trying for quite some time to figure out what the "hidden layer" is, how it works and what the purpose is. So many others either get right up to that subject and then stop posting, or talk about it as if I should already know. So for some time, I have only been able to do simple perceptrons. Now I finally understand that hidden layers are just layers of multiple perceptrons being pushed into other perceptrons, where each perceptron has been trained to complete a different task. Thank you!
I found you years ago when I needed to learn steering algorithms. You made the math and algorithm simple(r) to understand and your videos are a lot like a drop of sunshine in my day. It reminds me of Reading Rainbow when I was a youngster. Now I am back to continue my work on CNNs. And there you are again in my suggested videos :D
The way you teach is fun , it's like you're yourself enjoying teaching, which us students love ...one could fall in love with the knowledge presented here!!
If every black kid in the hood had a teacher like this they'd all succeed at understanding this easily; why? Because this guy's likability makes you want to learn. When you enjoy the person teaching you, you will usually enjoy 'what' they're teaching you. The 'capacity' to 'understand' has very little to do with 'achievability' in human affairs & 'thinking' certainly pertains to human affairs. I'm understanding concepts I've never encountered before, not because I'm 'smart', but because the instructor in this video is interesting, funny, has a charm individually his own and is not intimidating or threatening in anyway, least of all, neither is he boring. Every young person deserves a teacher like this.
Keep going man, I wish you were my teacher from college. Fun, smiles and learning together. Such a great experience to learn with you, 15 minutes passed like nothing but full of knoweledge. Love from Brazil! Keep going!
An excellent and exciting explanation! This is exactly what I was looking for in trying to understand the motive behind the multi-layer perceptron. Not to be taken for granted!
You should be given your own show on the science network. More educational, fun, engaging and entertaining than 99% of the crap we pay for. Better than most courses I have seen on programming. Bill Nye + Bob Ross + Mr. Rogers. 11/10
How can someone dislike his video .he seems to be a genuine happy man..exuding joy...let him be :) The kind of excitement he has towards his code is what I need towards my life ;)
I've seen a lot of videos about neural networks, both advanced ones (which go over my head) and beginner ones (which are too general). That XOR example in this video was an epiphany for me! Now I have an intuitive sense of what makes neural networks so special (vs., say, linear classifiers.) Now I feel like I'm finally ready to go deeper into this subject
@@TheCodingTrain Yah, that bit about how a single layer network can only solve linearly-separable problems, and how hidden layers fix this limitation, finally makes intuitive sense to me thanks to the XOR example. Thanks! Not sure if you cover this in subsequent videos, but I'd be interested to hear your take about why having multiple hidden layers can be useful, vs. just one hidden layer.
Outstanding explanation of linearly separable. You make it very easy to understand why multiple perceptrons are required. Plus I love Boolean logic. Thank you.
Interesting, so basically same analogy to electronics building Logic gates from transistors. You kind like add they together to get more complex operations. Very good material. Keep going, I'm really into this
Another way to see linearly separable problems: If it has a binary output, as in it either is or it isn't. With the dots on the canvas, they are either below the line, or they aren't. We just picked "aren't" to mean "above", but that's how we humans chose to read the output. We read it as "below" or "above", the computer reads it as "is" or "isn't". If you draw a line across your data and define a relationship between the data point and the line, the point either falls into that relationship, or it doesn't.
Love this video. Explained it really well. I have an exam on Wednesday which covers MLP and the functions of layers and neurones. This should help form my answer.
today i should study MLP but because of my some problems i could not concentrate.But after watching your tutorial you make me smile and forget about problems and understand the topic.Thanks a lot :)
Damn Dan you seem to be such a lovely person and I say it as a man! Keep doing these tutorials becouse I don't know if there is any other channel on yt explaining neural networks on code as good as you do it.
Great video as usual Dan, I'm looking forward to the sequel =) On a side note, I think everyone here understands !AND but the usual way is to call this gate NAND (for Not AND).
Hey there. I work as an artificial intelligence expert. I write state of the art neural networks libraries in C++ for a living. If you would like to talk about NNs with me in personal, or just ask me any questions, feel free to do so. I like your teaching style and I think knowledge about these kinds of things should be more universal!
