Final answer:
Gaussian Processes, a machine learning concept, allow for the computation of confidence intervals with predictions, offer support for custom kernels, and their efficiency is not inherently tied to the number of features or sparsity.
Step-by-step explanation:
The question is regarding Gaussian Processes, which is a concept from machine learning, a subfield of Computers and Technology. Among the statements listed in the question, the one that is true about Gaussian Processes is that predictions enable users to compute confidence intervals. This means that when making predictions using Gaussian Processes, not only is a prediction made about the expected value of the target, but a confidence interval can also be provided, giving a measure of certainty around the prediction. Gaussian Processes are not necessarily sparse, and efficiency does not depend on the number of features. Custom kernels are indeed supported by Gaussian Processes which allows them to be very flexible and adapted to the particular characteristics of the data.