Your videos are amazing! They remind me of the quote "anyone who tries to make a distinction between education and entertainment doesn't know the first thing about either"
Damn the way you've moved from Scalar-to-Vector by Algebra-to-Linear algebra, so Array-to-Matrix. You're a professional Educator, Siraj. Thank you very much.
Siraj, I jumped into Deep Learning about a year ago and am a huge fan of your videos; this would have been a fantastic resource when I first started out and I think it's so cool that you're taking the time to get through the foundational stuff so everyone can enjoy your other, awesome videos! Thanks :)
SIRAJ, i finished the Spanish subs in the video of the Intro, maybe i'll make the next tomorrow or today at the night. Please, accept it and update the Google Drive Document with the script of this video when you can (: Awesome vid bro!
Gracias Pinkie, i've accepted your subtitles. I've also added this script to that same document. I will continue adding every weeks script to that document and review any subtitle submissions you make. :)
Loved this video. Very interesting subject. It reminds me of stuff I have been thinking about myself independently for years. Thank you Siraj, this video was great.
let me also add, the more I go thru his mathematics videos the more I despise my past mathematics teacher. I flunked highschool algebra, yet now I find myself using linear algebra to optimize learning programs. Thank You for the teaching, Siraj.
Hey Siraj, please consider making a video on Object detection, the deep learning approach perharps with Fast-RCNN, YOLO, SSD or The tensorflow object detection API..please....
WOAH. I didn't expect you to understand Einstein Field Equations! Dammit you gotta do a series on them too :3. You inspire me Siraj. You really really inspire me to keep learning. Thank you for existing.
Hey Siraj Raval, how probable is it for abstract machine learning libraries to reduce demand for machine learning expertise to the point that the average person could not reasonably expect to gain employment in machine learning after investing the considerable effort it takes to understand the underlining math of machine learning? For example in a previous video you showed a abstraction of tensorflow that allowed you to simply take a labeled data set f(Vi) = Yi. All training/learning was then abstracted by the library and and you simply input a new vector that was not labeled and and the library output its best guess of y's value or label based upon the data set you feed it. Will machine learning EXPERTISE only be necessary for companies that develop machine learning libraries? How long is the window for wide demand, not counting grandiose entrepreneurial endeavors?
Please help, I don't understand what is tensor. Vector is invariant under coordinate system. I can use no coordinate system at all e.g. for moving particle under the known curve - I can measure the path length. Q1: Is tensor depend on coordinate system? Q2: I have rather good example of tensor of second order like intertia tensor of rigid body. I can use it to "exchange" non-free rotation vector "w" which I can move along a line into momemt of impulse(kinetic moment) via L=IW via using using usual rules of mulitplication. But what does it mean to be a more high dimensional tensor like 3? In video it have been said that this just an array of numbers but to which ojects I can apply this imagine tensor of degree 3 and how I should do it? How this numbers which are represented a tensor changes if use other coordinate system to describe my objects? Thanks
Awesome vid Siraj. One question - at 8:56 when you show the 4 types of norms, is there a norm that you recommend developers should use most often? Or does it depend on the type of ML model you're building?
Hey, I find your memes to be distracting. The one with spongbob at :50 didn't really carry the point you were making. Don't get me wrong, I love that you use memes, but I feel they should be included only to solidify the point rather than a quick joke - or after your point is made so it's not distracting.
Thank you for the explanation, but one thing I'm still not clear on is how does Word2Vec finds similar words from sentences. For eg. 1. Mercedes is a German legacy 2. Honda on the other hand is pretty new 3. But nothing beats Toyota reliability In this context, how will Word2Vec get to cluster {Mercedes, Honda, Toyota} as one group - can you please help explain this. I do get calculating distance, vectorizing the position of each word and selecting the dimension randomly but how does Word2Vec works. Thank you again Siraj from a software developer.
Zero norm at 08:59 is not a norm. Never was be, never will be. Some people which work in convex optimization called it caridnality. At lease because card([1 0]) = 1, but card(2*[1 0]) is also 1. It is not 2.
So how does matrix multiplication work programmaticly in numpy? I thought matrix multiplication is done the algebraic way in a loop like you showed in the algebra example and numpy is just faster because its doing its looping calculations in C. (4:52)
hi can you answer this for me sir? 😶 " given the line x+3y+5=0 and a circle x^2+y^2-6x-4y+3=0 find the equation of the tangent to the circle which are parallel."
HE IS MY 'RIVAL' WE HAVE EQUAL POPULARITY AND MY ADS ARE LIKE HIS WITH SIMILAR COVERAGE AND MONETARY RECOMPENSE, LIKE SCROUGE AND MCQUACK, IT'S A RACE TO THE FINISH WHERE WINNER TAKES ALL, THANK GOD
Would be great to see someone actually point out the mistakes in this video. I begin by pointing out 5:31. By reshaping your input you don't take it to a new vector space. An mXn dimensional vector space and a 1Xmn vector space are homeomorphic. Anyone want to add to this list?
Adding on as I watch, 6:00 word2vec was not a single neural net. It was an ensemble of models. And no it can't learn the embedding for a new word. There's no training/testing split there. There's only an embedding.
Crap you are convincing me to take the two vector calculus courses at my college. In addition to the multivariable course. Okay so which foundational papers in the field should I read and which major books for class, for amateurs and professionals.
Thanks! Also I will happily review any PRs to my syllabus if someone finds that the order could be made better. github.com/llSourcell/The_Math_of_Intelligence
I think the example compares an iteration pattern using algebra with processing arrays in say, Python. The %%time magic command will proved that list computation is faster.
Siraj Raval how can i build a machine learning model for blocking the attacks made by hackers.and how would be the combination of block chain and machine learning...n thanks for reply
09:54 - I think it false description in that sense that you're overfit and etc. what is real going on you solve bicreterion optimization problem. For more details what is regularization please take a look into convex optimization books, such as this web.stanford.edu/~boyd/cvxbook/ p.s. Thanks for video! It is still great!
Is the Interview with Elon musk happened or the bit shown in the video is a made up? I couldn't find any result in search. If it is real, someone please send the link. Thanks :))