@creambuncreambun4511
Great video, great illustration, easy to understand and fun. Cosine similarity was well illustrated. If there's one thing I may suggest, it would be the scenarios where U1-U2 similarity is -0.5 but U1-U3 is +0.5, in which case the denominator (i.e. their sum) will become zero. Of course based on the ratings the similarity between any pair of users will at least be zero, so zero-denominator scenario would not happen, but that also means we need to "tweak" the cosine similarities to make 0 to be the minimum instead of -1, and in that case the suggested scores would be more predictive. Well done
@nv9369
Please make more content like this! Applied math is so cool
@prettyfaceanb
More content like this please. When it clicked that they werent just looking for similiar numbers but using math to determine what a user might rate a show...my mind was blown.
@StingerSecSol
1st video of yours I have seen. I subscribed because you used a real life example to explain a complex idea.
@Hidden_Box
This type of teaching. I didn't not see in my school life. But I learnt about vectors without knowing it's real life application 😕😕. Thanks for this Video. I can connect dots. It make much meaning to me understand better.
@heenakhandelwal8608
My brain is totally blown away 🤯🤯 This is just wow
@x.in_hype
Matrices 🔥🔥🔥
@rafalablamowicz7919
A very nice application is to use SVD to decompose and then reconstruct, step-by-step, a color digital photograph F. For example, a digital color HD photograph with 1280 x 720 pixels is stored in terms of three integer matrices of dimensions 1280 x 720: one matrix R for red, one matrix G for green, and one matrix B for blue with each integer entry between 0 and 255. Thus, each color pixel at the location (i,j) of F is a result of mixing red, green, and blue colors whose amounts are determined by the integer values R(i,j), G(i,j), and B(i,j). Now, apply SVD to each matrix R, G, and B. Each matrix has then at most 720 singular values arranged in descending order as say g1 > g2 > ... > g720 where the majority of the later singular values are close to zero or are equal to zero (To be precise, these zero values on the diagonal of Sigma are not singular values which from the definition are always positive). Let k be a positive integer between 1 and 720. By using say the first k=10 then k=20, k=30, etc. singular values of each matrix R, G, B, one gets an integer matrix Rk, Gk, Bk which approximates the original matrix R, G, B, respectively. By combining the colors determined by Rk, Gk, Bk for each k, one can create a photograph Fk which will be approximating the original photograph F. That is, one gets a sequence F1, F2, ..., F720 of these approximations such that : (a) the first photo F1 gives some an innacurate but pretty close approximation; (b) every next approxaimation is better; (c) the last one F720 gives the perfect image F. Using a free software now, make a movie that displays these approximations in the order F1, F2, ..., F720. The idea behind is this: since the singular values are arranged in the descending order, the first few approximations contain the majority of the features of the original photograph, and the remaining approximations fill in small, minute details which are hard to even notice with a naked eye. Thus, when a picture is taken in space or from the Earth orbit, it is enough to send to Earth a small information derived from the SVD of each color matrix R, G, B (say, the first 20-30 singular values from the matrix Sigma, and the first 20-30 columns in the matrices U and V as in R = U Sigma V^T and likewise in G and B), the image can then be restored pretty well. To restore it completely, all non-zero singular values would need to be sent and the corresponding columns in the matrices U and V for each matrix R, G, B. To perform SVD on such large (or even larger for larger resolutions) matrices, one needs a software that can do it. Then, a software is needed to decompose any color photograph into the three matrices R, G, B, and a software that can create a photograph from three appromations Rk, Gk, Bk. Maple, a Computer Algebra System, from Waterloo, Canada, can do it all. Then, one can find a free software that can make the movie. I have done such movies with my students when teaching them matrix algebra. This was a great fun. The above mentioned SVD is one of the most complex concepts taught in a matrix algebra class but it is pretty useful. A much easier application of matrices is to find the smallest number of moves to perform a task, e.g., moving missionaries and cannibals across a river in say a two-person boat (or, goats and cabbage, etc.) using some basic ideas from graph theory.
@navinsingh9133
One of the best videos on Matrices 🎉
@cheems_0007
Most of the time I don't have anything to comment but I comment on this channel so that algorithm things this is a good video good channel and promote this channel and this channel grows and continue making this type of banger videos
@gokulaashiq9372
Thank you. Very good explanation ❤️🌝
@IhsanPro999
Your Teaching style is incredible
@AshutoshPandeyGlobal
beautifully explained, thank you for making such an amazing video, it unlocked a different way of looking the world in my view now.
@evrenkutlu6300
Please continue, they are really helpful.
@shaikusman536
Awsome content brother.....keep educating....Respect from Bangalore, INDIA....
@JJ_TheGreat
8:14 Question: The imputed 4(.04) rating of U1 for M2 is based on the ratings for the user he has the 2nd most in common with (U4) and the user he/she has the least in common with (U2) - while the user he/she has the most in common with (U3) hasn’t rated it. On the contrary, the relatively low rating for M5 does come along with U3’s rating.
@_rishi_ranjan
This is just an small example how well he explained this. Recently I worked in one of my Ai project and used vector database. And for smiliarity we use different algorithms one of them is cosine similarity ❤
@AmberRathour366
Nicely explained. Nice video
@francishubertovasquez2139
Matrices on one side, π Pi on one side, then the bridge of guitar 4x4x4x4X of your favorite like staircase up or down
@webflyer035