A neural network (NN) is a parameterised function that can be tuned via gradient descent to approximate a labelled collection of data with high precision. A Gaussian process (GP), on the other hand, is a probabilistic model that defines a distribution over possible functions, and is updated in light of data via the rules of probabilistic inference. GPs are probabilistic, data-efficient and flexible, however they are also computationally intensive and thus limited in their applicability. We introduce a class of neural latent variable models which we call Neural Processes (NPs), combining the best of both worlds. Like GPs, NPs define distributions over functions, are capable of rapid adaptation to new observations, and can estimate the uncertainty in their predictions. Like NNs, NPs are computationally efficient during training and evaluation but also learn to adapt their priors to data. We demonstrate the performance of NPs on a range of learning tasks, including regression and optimisation, and compare and contrast with related models in the literature.

No abstract

No abstract

In the last decade, remarkable developments have concerned the methods of space charge measurement in the field of insulation systems diagnostic. In particular, methods based on acoustic and thermal phenomena have been largely used. The present review provides a broad overview on the different techniques used describing, for each of them, the working principle, the main features and the most relevant applications. Further details are provided for the Pulsed Electro-Acoustic (PEA) method, as it seems to be the most used. This article provides more details on its historical evolution, showing evidence for its technological limits and taking into consideration the advantages and drawn from the different configurations of the measuring cell. A similar approach has been used for the group of thermal methods, whereas for the optical methods only the basic working principle and latest applications are reported

No abstract