Skip to main content

Carlos Santos's Library tagged optimization   View Popular

08 Dec 09

Nuit Blanche: CS: YALL1, fitness function Eureqa, Data Driven CS, Matlab, Python, NIPS 2009 papers, Bayesian statistics in AI and Social Sciences

"they just came out with a GUI that allows anyone to try that experiment for themselves. It's called Eureqa. [...] I am surprised to see that there is no response to the criticism made on the fitness function of the original article and now this program.

nuit-blanche.blogspot.com/...ness-function-eureqa-data.html - Preview

GeneticAlgorithms Optimization evolutionarycomputation for:mirwox

20 May 09

[0905.2924] Colorization of Natural Images via L1 Optimization

They claim better results than the already awesome algorithm of Levin et al. Convexifying the problem through l1 optimization

arxiv.org/0905.2924 - Preview

sparsity colorization l1 optimization regularization via:chl imageprocessing

09 Apr 09

Convergence is Relative: SGD vs. Pegasos, LibLinear, SVM^light, and SVM^perf « LingPipe Blog

  • a nice paper that introduces a blocked gradient optimization approach for support vector machines (SVMs). Setting the blocks to a single item yields stochastic gradient, and setting them to the whole corpus gives you a traditional gradient approach. The innovation is a step that handles L2 (aka Gaussian prior) regularization by projecting the coefficients down onto an L2-ball (this technique’s since also been used for L1 norms). The benefit over standard SGD is that it automatically sets the pesky learning rate parameter.
  • a nice paper that introduces a blocked gradient optimization approach for support vector machines (SVMs). Setting the blocks to a single item yields stochastic gradient, and setting them to the whole corpus gives you a traditional gradient approach. The innovation is a step that handles L2 (aka Gaussian prior) regularization by projecting the coefficients down onto an L2-ball (this technique’s since also been used for L1 norms). The benefit over standard SGD is that it automatically sets the pesky learning rate parameter.
1 - 20 of 75 Next › Last »
Showing 20 items per page

Highlighter, Sticky notes, Tagging, Groups and Network: integrated suite dramatically boosting research productivity. Learn more »

Join Diigo