This channel is criminally underrated, watching the video i was baffled by how well explained and how WELL EDITED THIS IS, I cannot fathom the amount of work that went into this especially the visualisations ! I thought this channel had millions of subscribers and i was wondering how i missed it, only 9K ?!!! I'm sorry this is outrageous. PS: I think 3Blue1Brown would definitely be proud.
You just brought the whole linear algebra of quintals of books of one full academic year less than an hour of enjoyable stuff to the senses of even common people. Kudos. Thank you very much.
Dude unfortunately we are only told to calculate them without any meaning. Now that I know the meaning it all makes sense. Thank you pouring so much information into one single video
In these 2 parts I could say I've learnt more about matrices than (sadly) during the high school and even university courses on linear algebra. It gave me so much needed intuition behind all those theoretical concepts. I could only wish our educational system was so engaging and demonstrative like these videos! Thank you very much! This channel deserves much more attention! ❤
I deeply appreciate your videos on matrix visualization, bro. I always "struggled" with matrices and linear algebra. Now you gave powerful tools for interpretation and intuition behind every single concept behind them. Great content. Impressive animations and sound effects. A piece of art.
for those confused why is sqrt(2)/2 a unit vector, cuz at least I thought it should be equal to 1, here's the proper explanation: A unit vector is a vector with a magnitude of 1. we find magnitude using |v| =√(x^2 + y^2) or also known as the distance formula between points OR ALSO KNOWN AS THE PYTHAGORUS THEOREM-ish ("-ish" cuz Pythagoras theorem is for right-angle triangles but it's fine it works here as well cuz when finding distance from two points, we can just imagine the distance to be a hypotenuse of an imaginary right-angle triangle and then apply the formula using √(x^2 + y^2) where y is the 'rise' and x is the 'run') . normally, |v| =√ (1^2 + 0^2) = 1 OR |v| =√ (0^2 + 1^2) = 1, hence they're a unit vector, but it's not always limited to that. The expression √2/2 can represent the x or y component of a unit vector in two dimensions. For instance, normally if you have a unit vector pointing in the positive x-direction, its x-component would be 1 and its y-component would be 0. However, if you have a unit vector pointing at a 45-degree angle from the positive x-axis aka √2/2, both its x and y components would be √2/2, and the magnitude of the vector formed by these components would indeed be 1. This is because when you square and add the components, you get 1, then √1 is still 1, hence satisfying the condition of being a unit vector.
It's a great piece to summarize a hard topic. He has also made it very simple with the computer graphics. I've learnt linear algebra from Prof G. Strang and this just gives it life to explain those hard topics.
I asked ChatGpt and it could not compete with this guy's videos. This is Gold. Whoever did these videos should share their name. A great honour beholds them. 🎉❤
Projection matrices are idempotent matrices. What you considered are orthogonal projection matrices i.e., symmetric idempotent matrices. Orthogonality requires an inner product but general projectors exist in every vector space not just inner product spaces.
great video but here is one error/correction: at 3:16 it is stated that orthogonal matrices produce pure rotations with "no reflection" - this is in fact false - orthogonal matrices can indeed produce reflections. The defining property of an orthogonal matrix is U^T U = UU^T = I; check this property for any reflection matrix and you will find that it is satisfied. Orthogonal matrices can produce both rotations and reflections. Note that in 2D, a product of two reflection matrices is also a rotation matrix.
in case anyone is wondering, the subset of orthogonal matrices that produce pure rotations (not reflections) are those that have determinant(U) = +1 (and with -1 a reflection is additionally involved in general). for orthogonal matrices the determinant is always +/- 1 so this covers all cases.
Seriously, you're going from basic visualisation to spectral decomposition I'm the second vid?! Brave man. I don't mind you skipping Gaussian elimination but i do think you should have shown what a linear combination is. Also independent and dependent bases. Was a lovely video though and, like you, I prefer the geometric interpretation.
It's not true that all orthogonal matrices represent rotations. Some orthogonal matrices are reflections. Specifically, orthogonal matrices with determinant = 1 are rotations, and orthogonal matrices with determinant = -1 are reflections.
@@mb59621 I wouldn't say so. A reflection inverts the orientation of the space, hence the negative determinant, whereas a rotation does not. Reflections are also their own inverse. Applying a reflection twice lands you back where you started, whereas it doesn't with most rotations. Since orthogonal matrices are defined as those with an inverse equal to its transpose, this means that the transpose of a reflection matrix is equal to itself (since it is its own inverse). Therefore, reflection matrices are symmetric across their main diagonal (taking the transpose doesn't change the matrix), whereas most rotation matrices are not.
@@APaleDot good points to ponder while still learning the subject. At first I thought reflection is just a rotation about the normal , but ty for your reply , i found an inaccuracy in my intuition that a single rotation matrix built on this principle cannot reflect anything more than 1 point ! Which is what makes reflection matrices a bit unique.
at 2:07 , orthogonal matrices need not have a unit column vectors, only orthonormal matrix, anyway which is a subset of orthogonal matrix, has unit column vectors. Am i right? If not, please enlighten me. Thank you for the awesome video.
These videos, and other visualization videos of a similar nature, are very helpful in understanding much of underlying concepts. One thing, however... is it possible to perform some spell checking within the visualizations themselves? Sometimes the misspellings diminish the impact of the video. Is spell checking within the visualizations difficult, or is it difficult to correct the visualizations after a misspelling is detected later in the production process? Anyway, good work.
Ugh, you made me go and check. It's correct. The references are weighted. A to itself is 0.2 and this is the top left. All the diagonals are correct. And those between the nodes are correct.
@@davidmurphy563 WTH you mean by weighted? Matrix consist nothing but simple transition probabilities, but he collects the values in wrong places. Only first row is correct, i.e. transition probabilities from A to A, B, and C.
@@woowooNeedsFaith The matrix entries can be equally referred to as transition probabilities or transition weights. Just google it, it's in use. I prefer the latter as you're often not looking a probabilities so the former can be a bit of a misnomer in my humble. I'll concede the 2nd point though. I checked the top row and the right hand column - figured it was a good sample - tutted at you and stopped there. Bloody unlucky to check 5/9 and see no issues when 4/9 were wrong... I was questioning my life choices at that point. He got col 1 & 2 switched on rows 2 & 3. Should have been: {{0.2, 0.4, 0.4}}, {0.6, 0.3, 0.1}, {0.4, 0.1, 0.5}} Meh, a silly slip up / typo. No biggie.
@noitnettaattention @rhaegartargaryen0947 basically what he is saying is that for a vector ( column of the matrix) to be a unit vector then it needs to have a module of 1. The module of a vector is just like doing a Pythagoras theorem with x and y to find the hypothenuse. this said, generally people don't go that far and just use the following formula: module = sqrt(x^2 + y^2). the module of the vector is therefore sqrt(1/2+1/2) = sqrt(1) = 1 thus making it a unit vector. this vector is very commonly seen on many applications since it has a simple 45º angle with the axis but any vector you can imagine that goes from the origin to a radius 1 circumference really qualifies as a unit vector (in R^2, 2 dimensions). What he wants to explain with this is that there can be matrixes that have two unit vectors, but to have them orthogonal is yet another important and distinct characteristic that he shows some seconds after. The good thing about the orthogonal unit vectors is that they are called "eigenvectors" and together with "eigenvalues" they will set a new set of axis for this transformation, as if the whole original axis is being rotated, as to have two perpendicular axis in a 2D space you can basically say that they are just some rotation or inversion of the original two axis. Hope I helped! Cheers
Your videos are brilliant, they should be 1K times more popular! But why do you need jazz in the background, it just distracts? ). Thanks, helped me a lot!
For orthogonal matrices, I don’t believe that implies they have unit vectors for columns. Isn’t that reserved for orthonormal matrices? Just want to correct my understanding it it is wrong, great video!
A unit vector is any vector with a length of 1. We can find the length of a vector using the Pythagorean theorem: (length)² = x² + y² = (sqrt(2) / 2)² + (-sqrt(2) / 2)² = (2 / 4) + (2 / 4) = 1 The square root of 1 is just 1, so we conclude that the length of the vector is 1, a unit vector.
This channel is criminally underrated, watching the video i was baffled by how well explained and how WELL EDITED THIS IS, I cannot fathom the amount of work that went into this especially the visualisations ! I thought this channel had millions of subscribers and i was wondering how i missed it, only 9K ?!!! I'm sorry this is outrageous.
PS: I think 3Blue1Brown would definitely be proud.
This is magic. You are an incredibly talented math communicator. A billion thanks for your content.
The world needs you bro
You just brought the whole linear algebra of quintals of books of one full academic year less than an hour of enjoyable stuff to the senses of even common people. Kudos. Thank you very much.
Dude unfortunately we are only told to calculate them without any meaning. Now that I know the meaning it all makes sense. Thank you pouring so much information into one single video
In these 2 parts I could say I've learnt more about matrices than (sadly) during the high school and even university courses on linear algebra. It gave me so much needed intuition behind all those theoretical concepts. I could only wish our educational system was so engaging and demonstrative like these videos! Thank you very much!
This channel deserves much more attention! ❤
you made 4 videos on topics i didnt fully understand and just dipped lmao. legend
Awesome, simply awesome. Only 51 comments for such a fabulous job is just not fair.
I deeply appreciate your videos on matrix visualization, bro. I always "struggled" with matrices and linear algebra. Now you gave powerful tools for interpretation and intuition behind every single concept behind them.
Great content. Impressive animations and sound effects. A piece of art.
Please do not stop doing videos, are just invaluable.
thanks for making me fall in love with math all over again
for those confused why is sqrt(2)/2 a unit vector, cuz at least I thought it should be equal to 1, here's the proper explanation:
A unit vector is a vector with a magnitude of 1. we find magnitude using |v| =√(x^2 + y^2) or also known as the distance formula between points OR ALSO KNOWN AS THE PYTHAGORUS THEOREM-ish ("-ish" cuz Pythagoras theorem is for right-angle triangles but it's fine it works here as well cuz when finding distance from two points, we can just imagine the distance to be a hypotenuse of an imaginary right-angle triangle and then apply the formula using √(x^2 + y^2) where y is the 'rise' and x is the 'run') .
normally, |v| =√ (1^2 + 0^2) = 1 OR |v| =√ (0^2 + 1^2) = 1, hence they're a unit vector, but it's not always limited to that.
The expression √2/2 can represent the x or y component of a unit vector in two dimensions. For instance, normally if you have a unit vector pointing in the positive x-direction, its x-component would be 1 and its y-component would be 0. However, if you have a unit vector pointing at a 45-degree angle from the positive x-axis aka √2/2, both its x and y components would be √2/2, and the magnitude of the vector formed by these components would indeed be 1. This is because when you square and add the components, you get 1, then √1 is still 1, hence satisfying the condition of being a unit vector.
Looking forward to Chapter 2 !!
It's a great piece to summarize a hard topic. He has also made it very simple with the computer graphics. I've learnt linear algebra from Prof G. Strang and this just gives it life to explain those hard topics.
I like the character development of the potato!
This is really great. You will get a better understanding at 3:39 with a unit circle.
Man, these videos are gold!
amazing finish of part 2, thank you
I rarely comment or like anything.
Made an exception for your work
The music is unexpectedly cool for a Math video
This video was beautiful and emotional. Thank you
Keep doing these videos man, We just love them!
Shared it with all my friends as a token of gratitude.
A lovely channel. Personally I wish there were no music, but the clear examples are golden.
Please upload more videos these are extremely helpful!
magical video
insane - many thanks for this!!!
7:01 Death's end reference?
Thank you so much love from india. You sorted out Lot of things
I asked ChatGpt and it could not compete with this guy's videos. This is Gold. Whoever did these videos should share their name. A great honour beholds them. 🎉❤
These videos are invaluable! Thanks a lot. Please create more of such videos.
Nujabes music in the back and Watanabe characters to describe. I like this channel!
Phenomenal work! I'm very thankful to you for such a great content
awesome videos and I love the graphics, nice to see some of my favorite anime characters while studying linear algebra ;)
0:42 you can always work out the equations and it will be quite apparent what is going on!
excellent, why is this channel not producing more videos like these.
Love the detail of the least squares normal equation when talking about the data matrix (10:46)😂😹
Projection matrices are idempotent matrices. What you considered are orthogonal projection matrices i.e., symmetric idempotent matrices. Orthogonality requires an inner product but general projectors exist in every vector space not just inner product spaces.
great video but here is one error/correction:
at 3:16 it is stated that orthogonal matrices produce pure rotations with "no reflection" - this is in fact false - orthogonal matrices can indeed produce reflections. The defining property of an orthogonal matrix is U^T U = UU^T = I; check this property for any reflection matrix and you will find that it is satisfied. Orthogonal matrices can produce both rotations and reflections. Note that in 2D, a product of two reflection matrices is also a rotation matrix.
in case anyone is wondering, the subset of orthogonal matrices that produce pure rotations (not reflections) are those that have determinant(U) = +1 (and with -1 a reflection is additionally involved in general). for orthogonal matrices the determinant is always +/- 1 so this covers all cases.
Thank you for this fantastic video!
Visualizing matrix helped me to lock in e knowledge and make sense of it. Thank you 🫡
Seriously, you're going from basic visualisation to spectral decomposition I'm the second vid?! Brave man.
I don't mind you skipping Gaussian elimination but i do think you should have shown what a linear combination is. Also independent and dependent bases.
Was a lovely video though and, like you, I prefer the geometric interpretation.
Incredible work!!
perfect.
TREMENDOUS
Is that a reference to 二向箔 :)
nujabes in the background with anime characters to explain. 10/10 banger
It's not true that all orthogonal matrices represent rotations. Some orthogonal matrices are reflections. Specifically, orthogonal matrices with determinant = 1 are rotations, and orthogonal matrices with determinant = -1 are reflections.
A reflection is a rotation too , right ?
@@mb59621
I wouldn't say so. A reflection inverts the orientation of the space, hence the negative determinant, whereas a rotation does not.
Reflections are also their own inverse. Applying a reflection twice lands you back where you started, whereas it doesn't with most rotations.
Since orthogonal matrices are defined as those with an inverse equal to its transpose, this means that the transpose of a reflection matrix is equal to itself (since it is its own inverse). Therefore, reflection matrices are symmetric across their main diagonal (taking the transpose doesn't change the matrix), whereas most rotation matrices are not.
@@APaleDot good points to ponder while still learning the subject. At first I thought reflection is just a rotation about the normal , but ty for your reply , i found an inaccuracy in my intuition that a single rotation matrix built on this principle cannot reflect anything more than 1 point ! Which is what makes reflection matrices a bit unique.
@@mb59621 a reflection is a discrete transformation whereas rotation is the sum of infinitesimal transformations.
I wish you could also do the same for Statistics and calculus topics. What can ML/AI/DS students ask for?
Amaizing content!!!!!!!!!!
Thank you for sharing your video. Wish you health and wealth.
very clear. like it very much. Thanks.
This is so good.
at 2:07 , orthogonal matrices need not have a unit column vectors, only orthonormal matrix, anyway which is a subset of orthogonal matrix, has unit column vectors. Am i right? If not, please enlighten me. Thank you for the awesome video.
Agreed
These videos, and other visualization videos of a similar nature, are very helpful in understanding much of underlying concepts. One thing, however... is it possible to perform some spell checking within the visualizations themselves? Sometimes the misspellings diminish the impact of the video. Is spell checking within the visualizations difficult, or is it difficult to correct the visualizations after a misspelling is detected later in the production process? Anyway, good work.
Thank you for your information. Thanks for telling us about Manim. I want to make videos to teach 7 years old kids math with animation.
Nujabes = Chefs kiss.
these are wonderful videos!
"Unfortunately no one can be told what the Matrix is" - Morpheus
“- you have to see it for yourself.”
What do you use to visualise the transformations?
Using cowboy bebop to teach linear algebra is something i didn’t know i needed
sublime
Where is the captions 🥺
10:36 - From this on matrix is wrong. You seem to place numbers into the matrix quite randomly.
Ugh, you made me go and check. It's correct. The references are weighted. A to itself is 0.2 and this is the top left. All the diagonals are correct. And those between the nodes are correct.
@@davidmurphy563 WTH you mean by weighted? Matrix consist nothing but simple transition probabilities, but he collects the values in wrong places. Only first row is correct, i.e. transition probabilities from A to A, B, and C.
@@woowooNeedsFaith The matrix entries can be equally referred to as transition probabilities or transition weights. Just google it, it's in use. I prefer the latter as you're often not looking a probabilities so the former can be a bit of a misnomer in my humble.
I'll concede the 2nd point though. I checked the top row and the right hand column - figured it was a good sample - tutted at you and stopped there. Bloody unlucky to check 5/9 and see no issues when 4/9 were wrong... I was questioning my life choices at that point.
He got col 1 & 2 switched on rows 2 & 3. Should have been:
{{0.2, 0.4, 0.4}},
{0.6, 0.3, 0.1},
{0.4, 0.1, 0.5}}
Meh, a silly slip up / typo. No biggie.
wait can someone tell what's special about projection matrices?
Long live the king
brilliant videos, thanks a lot
Sir, What is Geometric interpretation of Trace of a Matrix. Kindly make a video.
how you made a [ -sqrt(2)/2, sqrt(2)/2 ] into unit vector remains complete mystery to me....
i was stuck there too. can somebody explain this to me
@noitnettaattention @rhaegartargaryen0947 basically what he is saying is that for a vector ( column of the matrix) to be a unit vector then it needs to have a module of 1. The module of a vector is just like doing a Pythagoras theorem with x and y to find the hypothenuse. this said, generally people don't go that far and just use the following formula: module = sqrt(x^2 + y^2). the module of the vector is therefore sqrt(1/2+1/2) = sqrt(1) = 1 thus making it a unit vector.
this vector is very commonly seen on many applications since it has a simple 45º angle with the axis but any vector you can imagine that goes from the origin to a radius 1 circumference really qualifies as a unit vector (in R^2, 2 dimensions).
What he wants to explain with this is that there can be matrixes that have two unit vectors, but to have them orthogonal is yet another important and distinct characteristic that he shows some seconds after.
The good thing about the orthogonal unit vectors is that they are called "eigenvectors" and together with "eigenvalues" they will set a new set of axis for this transformation, as if the whole original axis is being rotated, as to have two perpendicular axis in a 2D space you can basically say that they are just some rotation or inversion of the original two axis.
Hope I helped!
Cheers
Great great great
vsauce music was a nice touch!
7:12 is the big reveal of dark forest
three body reference??/
Your videos are brilliant, they should be 1K times more popular!
But why do you need jazz in the background, it just distracts? ).
Thanks, helped me a lot!
Nujabes ❤
Bravooooooooo
For orthogonal matrices, I don’t believe that implies they have unit vectors for columns. Isn’t that reserved for orthonormal matrices?
Just want to correct my understanding it it is wrong, great video!
yes I had the same thought. Too lazy to check now...
I know this transformation, 3 bodies, reduce dimension 😁
How is root 2/2 , - root2/2 a unit vector
A unit vector is any vector with a length of 1. We can find the length of a vector using the Pythagorean theorem:
(length)² = x² + y²
= (sqrt(2) / 2)² + (-sqrt(2) / 2)²
= (2 / 4) + (2 / 4)
= 1
The square root of 1 is just 1, so we conclude that the length of the vector is 1, a unit vector.
background music too loud
3:13
Orthonormal
your content is amazing! but i feel you can probably rename your videos so they aren't discriminated by the algorithms!!!!
Where are you!!!
brilliant!!
Woow...
bro is the asian 3b1b
nujabes + maths
二向箔哈哈哈