I like this. I'd watch more of this guy explaining graph kernels and beyond!

stop interrupting the guy over and over. let him explain!

Nice presentation. More content on graphs optimizations

so text language models use a mapping coordinate system to find what words are most likely near it.. it would work with the graph triangle demos easily

Instead of three red and three blue nodes, what if we had three blue and one brown? 😁

i think you should apologize to dogs and cats

After office hours teaching me further on Kernel Hilbert space my ML professor said with disappointment (paraphrased) "But, all of this doesn't matter practically because deep learning, while it's inner workings are not understood rigorously, performance are so much better..." Do you mathematicians have any unput? It makes the work I study and teach in Discrete math seem fascinating but in the end, not practically useful...

I'd think the similarity between two objects is the % of overlapping arrangements. Identical objects would have 1 similarity, while objects that share a large pattern of elements would have about 0.5 similarity. The problem is that going through all those arrangements would get astronomically many right away.

the molecule is methanol...

How can the inner vector show similarity, if you do not normalize the vectors? If you multiply a vector by a factor k, the inner product will also be multiplied with k.

In case anyone is a bit lost as to what's going on with the dot product. Basically, it's a way of comparing two vectors for similarity. So if you've got two identical vectors of length 1 - so both pointing the same way - the dot is 1. Meaning identical.

Really interesting. For all interested in those directions I recommend researching in the direction of graph neural networks and specifically for molecules topological neural networks!

Nice to see David on here!

The kernel trick is simply a matrix multiplication. You have a embeddding set A of shape (m,n) and you get your kernel(gram matrix) by computing A@AT (A matmul A Transpose) which will be of shape (m,m) which tells you how similar each vector is to every other vector in the batch. if you flip it AT@A you can use that matrix of shape (n,n) and this is used to compute the covariance matrix and this tells you how how two features vary with respect to each other.

You can use graph convolutional networks to classify certain kinds of abuse by the users of your product. It's really flexible when you have sparse bits of information about the relationships between people and products. Graphs are absolutely everywhere.

Sean's really into it this time

So, it seems Dr David is my favorite now :)

i love this channel ngl

I think we need more videos on Graph theory.

The "plots", distinguished from graphs, which Sean raises at