/explore
|
You are here 
|
fa.bianp.net | ||
| | | | |
djalil.chafai.net
|
|
| | | | | This post is mainly devoted to a probabilistic proof of a famous theorem due to Schoenberg on radial positive definite functions. Let us begin with a general notion: we say that \( {K:\mathbb{R}^d\times\mathbb{R}^d\rightarrow\mathbb{R}} \) is a positive definite kernel when \[ \forall n\geq1, \forall x_1,\ldots,x_n\in\mathbb{R}^d, \forall c\in\mathbb{C}^n, \quad\sum_{i=1}^n\sum_{j=1}^nc_iK(x_i,x_j)\bar{c}_j\geq0. \] When \( {K} \) is symmetric, i.e. \( {K(x,y)=K(y,x)} \) for... | |
| | | | |
xcorr.net
|
|
| | | | | Earlier, I discussed how I had no luck using second-order optimization methods on a convolutional neural net fitting problem, and some of the reasons why stochastic gradient descent works well on this class of problems. Stochastic gradient descent is not a plug-and-play optimization algorithm; it requires messing around with the step size hyperparameter, forcing you... | |
| | | | |
windowsontheory.org
|
|
| | | | | Previous post: ML theory with bad drawings Next post: What do neural networks learn and when do they learn it, see also all seminar posts and course webpage. Lecture video (starts in slide 2 since I hit record button 30 seconds too late - sorry!) - slides (pdf) - slides (Powerpoint with ink and animation)... | |
| | | | |
igorstechnoclub.com
|
|
| | | This week I learned something that finally made "transfer learning" click. I had always heard that you can hit strong accuracy fast by reusing a pretrain... | ||
