Where can I find the problems to the lectures on Algebra by Benedict Gross?

12:15 it says Reimann but it should be Riemann. Other than that, thank you for the great video :D

How does a problem with the same initial conditions give different results?(Chaos Theory) Maybe they used slightly different initial conditions/approach to tackle this problem !!!

So, it depends on whether you count from zero or one when deciding "first" ;-) It also shows how lax we are with words and phrases and especially the failure to note those distinctions and their misunderstood conventions. E.g. Arbitrarily large and infinity are indistinguishable in the minds of the lay mathematicians. "To arbitrarily large and beyond" as the saying goes!

Hello author, I've recently been working on a mathematical physics chronicles content and I very much need to edit 5 seconds of footage using the footage from the video you created, I really, really need the 5 seconds of material. I'll be sure to credit the source of the footage and @ your channel when I'm done with my work. I very much recognize your work and thanks again. : ) (The above text was translated by me, using a translation program, so please forgive me if I use the wrong words) 😀

I always run aground on the same problem in Linear Algebra Done Right. I've even seen the solution to that problem. Nuthin'. Some polynomial thing, I think. I don't remember it exactly, but I know it when I see it. Thud.

Is this channel a gd reference???(joke)

Hey, amazing video. But what about statistics? What do you recommend for that? Like how should the base needs to be constructed and relationship between probability and statistics!

So if A^n+B^n=C^n has no integer solutions for n>2, does this imply a condition where the number of added terms must be < or = n for integer solutions to exist? So for n=3 would the sum need to be A^n + B^n + C^n = D^n to get integer solutions for A, B, C, and D? If there are integer solutions for A, B, C, and D would there be solutions for A, B, C, and D if N=4? Basically does there need to be n terms being added together for there to be integer solutions?

Ramanujan the joddest mathematician of all🛐🛐

This video is illuminating

Haha, that labeling machine was so unexpected

Great video! But I still don't understand how you connect the delta(q) (e.g. delta(q) = q[(1-q)(1-q^2)(1-q^3)...]^24), the resulting lattice to elliptic curve equation then the function f(q) = q - 2q^2 - q^3 +2q^4+ q^5+2q^6-2q^7-2q^9-2q^10+q^11+4q^13....). Where does the last f(q) polynomial expansion come from relative to the first? I get that the solutions (1+p-) (# of solutions mod p) reveal the same coefficients as this equation, but don't grok where it comes from as one steps through your video. Seems to pop out of nowhere? (e.g. the f(q))

Many thanks !

May I ask, which tab are you using to write on? Is it good for notetaking at uni?

I likd and learned machine learning; I didn't have a good teacher or experience -- I would like to start all over again

Thank you!

beautiful observation by sir srinivasa ramanujan 🙏

I don't understand. One can create a set X from a well defined subset of R, for example all real numbers with only one decimal. Then you would have |N| < |X| < |R| ? In my ignorance I would say that there is in fact an infinity of possible set X that satisfies |N| < |X| < |R|.

Honestly i just like math, im searching somewhere to study Calculus, limits, derivatives and integrals

PRML is the first book people should read, no question

7:56 as far as I understand it, there are no automorphisms not fixing the rationals, because the "prime field" (i.e. the minimal subfield in your field) always is fixed - which can easily be shown using less time than was taken here to repeat too often (imho) that only automorphisms fixing the rationals shall be taken into account...

very helpful thank you

Campusx KZbin Best

No idea why you selected cos t, right after saying "yeah periodic, so this is sin t". So the fact that squared they add to 1 wasn't surprising or meaningful to me, since it appeared your choice would lead to that.

well then, what about *geometric algebra*

I understand if one could not understand something we do not study or spend too much time on it. But is it bad?

I feel like if I had been presented algebraic geometry like this when it was my master's research area, I might have finished my PhD in mathematics.

This is one of my favorite results in mathematics. I used this categorical equivalence (along with the equivalence to algebraic function fields) in my master's thesis on prime Galois coverings of the Riemann sphere back in 2007.

Compared to the topics of your other videos, most machine learning is kindergarden level. Many people tend to say it is hard to boast about their achievement of learning it. The universities as well as ML-book writers happily play along as long as students keep paying. Frankly the math behind backpropagation is not beyond repeatedly applying the chain-rule for example. Most special tweaks have very little substantiation and are just applied because it worked for somebody else who found it by trial and error (choice of NN-architecture for example). People use PyTorch and Tensorflow not because they understand the limitations of SciKit-learn, but they want to put on their resume that they have experience with the 'professional' packages. In most cases learners of ML are looking for a well-paid job connecting black-boxes rather than a deep understanding. Having said that, this channel can make a difference by focusing on deeper understanding. For example of certain statistics. However the chances are low that anything of important practical use will be found given that nowadays millions of students are trying to improve many by trial and error (Yannic Kilcher YT presents a new paper every week mostly with sloppy math and many colorful pictures and score boards). Some years back the Caltech course 'Learning from data' (KZbin) was quite interesting with for example the Vapnik-Chervonenkis dimension explained. I also liked Welsh Labs (YT) 'learning to see'. Hope this helps if not just go with somebody else's opinion. ;-)

Bahen 🤩?

Hey man, if you haven't already, check out Michael Bronstein's Geometric Deep Learning, it's quite fascinating: youtu . be / w6Pw4MOzMuo

Thanks bro !

This concept reminds me of how glass condenses when its heated up; regardless of its initial shape it always wants to end up as a sphere with enough time and heat.

im glad i chose to be a neurosurgeon 😂😂😂

For reinforcement learning the video series by Mutual Information is an absolute must!

It's funny he specifically doesn't mention 3b1b's videos on ML, reckon they are bland

Liked this channel more when it was just about barebones exposition of math than what it's becoming now. Culling my old subsctibed channels, so I guess this is goodbye.

I think this video is the part one because Deep Learning is a branch of Machine learning. Also, learning these topics need the practical part: coding in notebooks, reading repositories, get the well known datasets and training some models.

When he talks about math, this channel was unique. Now he joined in red market and be one of them?

Thanks for this video! I’m actually really curious how machine learning might be used to search for more efficient algorithms

For a few weeks i have been wondering How could I learn machine learning once my studies were over (at the end of the year). I could not have dreamt about a better video. Thank you

Aleph...J.L.Borges?

Michael Nielson is one of the Greats!!!

I wish there was video like this for physics, a lot of channels recommend books but they lack lecture recordings

A video on exotic spheres would be nice

Why are the pure math elitists so pissed off in the comments 😂 There are still a lot of pure math topics in the backbone of ML and this is nothing close to the intense meat-riding of more mainstream AI/ML channels

Another resource that I think is great is MIT's Introduction to Deep Learning course on KZbin. Fully available online with lecture, slides, and labs. Look it up on youtube and theres a centralized link for all the lectures

The goat is back when we needed him the most

1:47 In reality, one will use scikit-learn most of time because: 1. You will not have enough data for training neural nets. 2. Even if you have enough data for training neural nets, you need a baseline model (let's say logistic regression or random forests) for performance comparison. 3. scikit-learn is more useful for rapid prototyping.