Abstract: A rate of change of current sensor includes two measurement coils, arranged such that the turns of the first and second measurement coils are interleaved. The rate of change of current sensor further comprises two return coils, arranged to progress in an opposite direction to the measurement coils. The coils form loops and progress substantially around a target measurement conductor. This ensures that the two measurement coils both receive the same electrostatic coupling from external conductors which are not the target of the measurement operation. Further, the two measurement coils are arranged such that the first coil and the second coil are, on average, the same distance to the current-carrying conductor of interest.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: A rate of change of current sensor includes two measurement coils, arranged on a substrate or printed circuit board. The coils form loops and progress substantially around a target measurement conductor. This ensures that the two measurement coils both receive the same electrostatic coupling from external conductors which are not the target of the measurement operation. Further, the two measurement coils are arranged such that the first coil and the second coil are, on average, the same distance to the current-carrying conductor of interest.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media for training machine learning models. One method includes obtaining a machine learning model, wherein the machine learning model comprises one or more model parameters, and the machine learning model is trained using gradient descent techniques to optimize an objective function; determining an update rule for the model parameters using a recurrent neural network (RNN); and applying a determined update rule for a final time step in a sequence of multiple time steps to the model parameters.

Abstract: A rate of change of current sensor includes two measurement coils, arranged on a substrate or printed circuit board. The coils form loops and progress substantially around a target measurement conductor. This ensures that the two measurement coils both receive the same electrostatic coupling from external conductors which are not the target of the measurement operation. Further, the two measurement coils are arranged such that the first coil and the second coil are, on average, the same distance to the current-carrying conductor of interest.

Abstract: A rate of change of current sensor includes two measurement coils, arranged such that the turns of the first and second measurement coils are interleaved. The rate of change of current sensor further comprises two return coils, arranged to progress in an opposite direction to the measurement coils. The coils form loops and progress substantially around a target measurement conductor. This ensures that the two measurement coils both receive the same electrostatic coupling from external conductors which are not the target of the measurement operation. Further, the two measurement coils are arranged such that the first coil and the second coil are, on average, the same distance to the current-carrying conductor of interest.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media for training machine learning models. One method includes obtaining a machine learning model, wherein the machine learning model comprises one or more model parameters, and the machine learning model is trained using gradient descent techniques to optimize an objective function; determining an update rule for the model parameters using a recurrent neural network (RNN); and applying a determined update rule for a final time step in a sequence of multiple time steps to the model parameters.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media for training machine learning models. One method includes obtaining a machine learning model, wherein the machine learning model comprises one or more model parameters, and the machine learning model is trained using gradient descent techniques to optimize an objective function; determining an update rule for the model parameters using a recurrent neural network (RNN); and applying a determined update rule for a final time step in a sequence of multiple time steps to the model parameters.

Abstract: Compositions and devices for the production of treated wood products are disclosed. A moisture exchange container for treating a plurality of wood pieces with a chemical formulation of interest is described. Also, an adhesive suitable for gluing wood pieces that have been treated, or that are going to be treated, with a chemical formulation of interest is described. Also disclosed are methods and systems using the devices and compositions of the present invention, and treated wood products.

Abstract: An insulation pad includes a binderless pack of glass fibers and an envelope around the binderless pack of glass fibers. The glass fibers are mechanically entangled by needling such that the binderless pack has a density of from 4.5 to 5.5 pounds per cubic foot. The insulation pad is used to insulate pipes and vessels.

Abstract: Compositions and devices for the production of treated wood products are disclosed. A moisture exchange container for treating a plurality of wood pieces with a chemical formulation of interest is described. Also, an adhesive suitable for gluing wood pieces that have been treated, or that are going to be treated, with a chemical formulation of interest is described. Also disclosed are methods and systems using the devices and compositions of the present invention, and treated wood products.

Abstract: Compositions and devices for the production of treated wood products are disclosed. A moisture exchange container for treating a plurality of wood pieces with a chemical formulation of interest is described. Also, an adhesive suitable for gluing wood pieces that have been treated, or that are going to be treated, with a chemical formulation of interest is described. Also disclosed are methods and systems using the devices and compositions of the present invention, and treated wood products.

Abstract: Methods and systems for predicting a proposed electronic mail (email) for an email campaign to be sent to a list of email addresses received from a sender is spam. A hard bounce rate is predicted for the list of email addresses based on hard bounce rates calculated for respective email addresses on the list. If predicted hard bounce rate exceeds a threshold, the proposed email is deemed spam and sender may be informed. If predicted hard bounce rate falls within range of the threshold, proposed email may be sent to less than all the email addresses on list. If an acceptable number of the sent emails are received, the email may be sent to remaining email addresses on list. Prior to sending proposed email, the list of email addresses may be compared to purchased lists of email addresses. If there is a match, the proposed email may be deemed spam.

Abstract: A corona igniter 20 includes an electrode gap 28 between the central electrode 22 and the insulator 32 and a shell gap 30 between the insulator 32 and the shell 36. An electrically conductive coating 40 is disposed on the insulator 32 along the gaps 28, 30 to prevent corona discharge 24 in the gaps 28, 30 and to concentrate the energy at a firing tip 58 of the central electrode 22. The electrically conductive coating 40 is disposed on an insulator inner surface 64 and is spaced radially from the electrode 22. The electrically conductive coating 40 is also disposed on the insulator outer surface 72 and is spaced radially from the shell 36. During operation of the igniter 20, the electrically conductive coating 40 provides a reduced voltage drop across the gaps 28, 30 and a reduced electric field spike at the gaps 28, 30.

Abstract: A corona igniter 20 includes an electrode gap 28 between the central electrode 22 and the insulator 32 and a shell gap 30 between the insulator 32 and the shell 36. An electrically conductive coating 40 is disposed on the insulator 32 along the gaps 28, 30 to prevent corona discharge 24 in the gaps 28, 30 and to concentrate the energy at a firing tip 58 of the central electrode 22. The electrically conductive coating 40 is disposed on an insulator inner surface 64 and is spaced radially from the electrode 22. The electrically conductive coating 40 is also disposed on the insulator outer surface 72 and is spaced radially from the shell 36. During operation of the igniter 20, the electrically conductive coating 40 provides a reduced voltage drop across the gaps 28, 30 and a reduced electric field spike at the gaps 28, 30.