Abstract: According to a first aspect a network system to generate output data values from input data values according to one or In more learned data distributions comprises an input to receive a set of observations, each comprising a respective first data value for a first variable and a respective second data value for a second variable dependent upon the first variable. The system may comprise an encoder neural network system configured to encode each observation of the set of observations to provide an encoded output for each observation. The system may further comprise an aggregator configured to aggregate the encoded outputs for the set of observations and provide an aggregated output. The system may further comprise a decoder neural network system configured to receive a combination of the aggregated output and a target input value and to provide a decoder output. The target input value may comprise a value for the first variable and the decoder output may predict a corresponding value for the second variable.

Abstract: According to a first aspect a network system to generate output data values from input data values according to one or In more learned data distributions comprises an input to receive a set of observations, each comprising a respective first data value for a first variable and a respective second data value for a second variable dependent upon the first variable. The system may comprise an encoder neural network system configured to encode each observation of the set of observations to provide an encoded output for each observation. The system may further comprise an aggregator configured to aggregate the encoded outputs for the set of observations and provide an aggregated output. The system may further comprise a decoder neural network system configured to receive a combination of the aggregated output and a target input value and to provide a decoder output. The target input value may comprise a value for the first variable and the decoder output may predict a corresponding value for the second variable.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for training a value neural network that is configured to receive an observation characterizing a state of an environment being interacted with by an agent and to process the observation in accordance with parameters of the value neural network to generate a value score. One of the systems performs operations that include training a supervised learning policy neural network; initializing initial values of parameters of a reinforcement learning policy neural network having a same architecture as the supervised learning policy network to the trained values of the parameters of the supervised learning policy neural network; training the reinforcement learning policy neural network on second training data; and training the value neural network to generate a value score for the state of the environment that represents a predicted long-term reward resulting from the environment being in the state.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for training a value neural network that is configured to receive an observation characterizing a state of an environment being interacted with by an agent and to process the observation in accordance with parameters of the value neural network to generate a value score. One of the systems performs operations that include training a supervised learning policy neural network; initializing initial values of parameters of a reinforcement learning policy neural network having a same architecture as the supervised learning policy network to the trained values of the parameters of the supervised learning policy neural network; training the reinforcement learning policy neural network on second training data; and training the value neural network to generate a value score for the state of the environment that represents a predicted long-term reward resulting from the environment being in the state.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for training a value neural network that is configured to receive an observation characterizing a state of an environment being interacted with by an agent and to process the observation in accordance with parameters of the value neural network to generate a value score. One of the systems performs operations that include training a supervised learning policy neural network; initializing initial values of parameters of a reinforcement learning policy neural network having a same architecture as the supervised learning policy network to the trained values of the parameters of the supervised learning policy neural network; training the reinforcement learning policy neural network on second training data; and training the value neural network to generate a value score for the state of the environment that represents a predicted long-term reward resulting from the environment being in the state.

Abstract: A separator is disclosed which provides an electrically insulating layer between groups of windings in an electrical apparatus, such as a rotating electrical machine. The separator comprises a matrix of interconnected rods. This can allow air gaps to be present between groups of windings, which may improve the cooling efficiency of the apparatus.

Abstract: A rotor disc for an electrical machine is disclosed. The rotor disc comprises a series of magnetisable members arranged in a circumferential series on a surface thereof and against peripheral retaining structure of the disc. The disc also includes a retainer fixable to the surface and arranged for assisting in retaining the members at least axially on the surface when so fixed. The retainer is arranged to abut at least one of a substantially circumferentially and substantially tangentially extending, radially inner part of each member and thereby to provide the axial retention.

Abstract: The present invention relates to perfumed compositions comprising polymer and nonionic surfactant and is explained with particular reference to insect-repellent compositions which are a preferred embodiment of the invention. The invention provides an aqueous cleaning composition comprising: a nonionic surfactant, a carboxylate polymer, and, a perfume. Preferably the compositions of the invention are aqueous cleaning composition of pH 3.5-5.5 comprising:a) 1-10% wt of an ethoxylated alcohol,b) 0.1-4% wt of a carboxylate polymer,c) 0.2-4% wt of a perfume component selected from the group comprising limonene, carvone, cineole, linalool, Gum Camphor, citronellal, alpha and beta terpenol, fencholic acid, borneol, iso borneol, bornyl acetate, iso bornyl acetate and mixtures thereof, and,d) 82-99% water,wherein the weight ratio of (a):(c) does not exceed 10:1.

Abstract: A rinse conditioner giving excellent deposition of a fabric treatment agent onto fabric. The rinse conditioner comprising a fabric softening compound, a water insoluble oil and from 0.1% to 10% by weight of the total composition of a Fabric Treatment Agent, wherein the Fabric Treatment Agent has a c.logP of 3.0 or more.