Please take a look at this proposal: Machine Learning > On Topic
I would like your opinion if these on-topic questions would be welcomed on this site:
- I am using a neural network with the backpropagation algorithm. How can I calculate significance of each input variable on the output?
- What is the best software for performing Gaussian Process regression?
- What is the best R package to use for support vector machines (SVMs)?
- Given a dataset, how do I determine the best ratio of training to test data for a supervised learning algorithm?
- Generally, when are Hessian-based optimization methods preferable to first order methods?
- If I have an undirected, tree-structured model, what's the most efficient way to pull an exact sample from its distribution?
- Why does the kernelized support vector machine algorithm care that the kernel function be positive semi-definite? What happens if it isn't?
If they are consistently on-topic, I will be closing that proposal as a duplicate and inviting those users to Stats.stackexchange.