SARS was caused by a D3 deficiency due to the use of sunblock. The new flu drugs at that time were D3 derivatives and massive doses of D3. The rest is a mafia con game for taxation without representation.
Sure. At any time instance t, the number of customers is equal to a(t)-b(t), i.e., the gap between the two piecewise linear functions (broken-line plots) at time t. So the average number of customers is equal to the average gap between the two broken-line plots. That is equal to the area in between divided by the "width", which is τ
This video was so disappointing in that they should of showed this pool thumping and they didn’t. All we got was blah, blah, blah, blah, blah, blah and more blah blah blah!!! 😡😡😡
To me word2vec and attention models looks similar, both comes up with a dimension of a word based on surrounding word. Could you explain how is different these are fundamentally? Thanks
Word2vec just creates a representation for words (or tokens). But attention is the way of how to process the representation vectors. On a high level, attention simply allocates more weights to representations that are more relevant.
Great video. Really appreciate it. But I do have a question. How is alpha^k chosen at each iteration? Or are there various ways to determine the step size?
do you mean the beginning? this is a trick to absorb the constraint. When we rewrite the original constrained optimization problem as a min-max problem, the inner optimization (maximizing y) will ensure that the value is infinite (and hence can never be chosen by the outer minimization) when the constraint is violated.
alpha does not "measure" thing per se. It is a parameter to control D_alpha. D_alpha is a generalization of the KL-divergence. You vary alpha and you can get different kinds of divergences. D_alpha and KL-divergence are used to measure the difference between two functions. Often time, two probability density functions.
I just put it at nbviewer.org/gist/phsamuel/6e90037a68782b3c1d1517e9ac4f0ca1 but I don't have a realsense camera anymore and cannot test the code again.
Very good explanation of GaBP, but the quality of video and noise in the background made it very difficult to understand properly. Kindy, if possible can you record another video with better audio and video? Thanks
