I've been wanting to make a video about the Euclidean distance for the last 3 years, but visualizing higher dimensional spaces is hard and I didn't have the skills to do it back then. Now, with an extra push from the #SoME3 competition, I managed to do something I'm quite happy with. What do you guys think? :-)
There is a book I really like called "Drive your plow over the bones of the dead" by Olga Tocarczuk. There is a scene where two of the characters are translating William Blake's poetry into English. They both go away themselves and write how they think the passages should be translated into English, and then they compare their two versions of the poem to see the differences in their translations, to come up with a third version of the poem in polish that is truest to the the imagery, meter and meaning of the original. The catch is that the book is translated into English (which is my first language). They somehow manage to capture the nuance of translating into another language and provide two versions of the translations so that you can see the hiccups that a translator might face, all while using only English. It is a brilliant magic trick of translation and understanding. This video is on a par with that.
My god that was simply INCREDIBLE! The visualization from the car "learning" to drive made me understand how that type of AI classifies the sensors distance to decide which way to turn! Awesome video Radu!!!
This was fantastic. I used what I learn to make an application that matched products by their RGB color values, pairing the different products brands with the closest colored products from other brands. Not as fancy as the car application, but using it to automate data saved me a day of work!
Oh my god!!!! Your tutorial are like dream come true for those who want to understand things so clearly as you explain. I want to become a teacher like you. I'm also learning a lot from your videos. Thankyou for putting so much efforts in making these videos and making it available for free.
Love this video so much. Clear explanations and surprisingly high production value - with your help I managed to grok it in a way I never had before. You are a godsend. Ta
First step is to convert words into numbers (feature vectors). You can look up word2vec online (there are others as well). It's not the full story for getting something like GPT but it's the step that makes the problem more similar to those we studied.
bro your intro... omg.... you probably could be working at Google and earn so much money with your skills and brains... the intro amazed me holy fuck Edited after I watched it: Bruh.... I am speechless... If only I had such resources when I was at school, I would've probably finished it with much better grades. How do you make all those visualizations? Is that a program, or what? Like, when you showed the math, or when there were sensors showing when the car moving and on the right the whole "tree" was lighting up accordingly... There's a lot intertwined to Hypotenuse and pythagorean theorem, and squre root...
Thank you :-) I'm glad to hear you liked it so much. I coded everything you see here, pretty much... visualizing the math, the self-driving car with the sensors (I have a full course on that) I just used it here (easy). And the drawing recognizer is from my machine learning course (also on RU-vid). The Pythagorean theorem is from one of my other explainers (check out the math and code fundamentals playlist).
Hey Bro! So I have been practicing coding for many month's now but nowadays I feel like I am nothing learning/improving as much as I used to before. What can I do about it?
In general, learning curves look like what you describe. You quickly get the basics down, and then improvements naturally come slower. It might be just normal.
