Abstract: A method is proposed for training a multitask computer system, such as a multitask neural network system. The system comprises a set of trainable workers and a shared module. The trainable workers and shared module are trained on a plurality of different tasks, such that each worker learns to perform a corresponding one of the tasks according to a respective task policy, and said shared policy network learns a multitask policy which represents common behavior for the tasks. The coordinated training is performed by optimizing an objective function comprising, for each task: a reward term indicative of an expected reward earned by a worker in performing the corresponding task according to the task policy; and at least one entropy term which regularizes the distribution of the task policy towards the distribution of the multitask policy.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network used to select actions to be performed by an agent interacting with an environment. One of the methods includes: obtaining data identifying a set of trajectories, each trajectory comprising a set of observations characterizing a set of states of the environment and corresponding actions performed by another agent in response to the states; obtaining data identifying an encoder that maps the observations onto embeddings for use in determining a set of imitation trajectories; determining, for each trajectory, a corresponding embedding by applying the encoder to the trajectory; determining a set of imitation trajectories by applying a policy defined by the neural network to the embedding for each trajectory; and adjusting parameters of the neural network based on the set of trajectories, the set of imitation trajectories and the embeddings.

Abstract: A neural network system is proposed. The neural network can be trained by model-based reinforcement learning to select actions to be performed by an agent interacting with an environment, to perform a task in an attempt to achieve a specified result. The system may comprise at least one imagination core which receives a current observation characterizing a current state of the environment, and optionally historical observations, and which includes a model of the environment. The imagination core may be configured to output trajectory data in response to the current observation, and/or historical observations. The trajectory data comprising a sequence of future features of the environment imagined by the imagination core. The system may also include a rollout encoder to encode the features, and an output stage to receive data derived from the rollout embedding and to output action policy data for identifying an action based on the current observation.

Abstract: A neural network system is proposed to select actions to be performed by an agent interacting with an environment to perform a task in an attempt to achieve a specified result. The system may include a controller to receive state data and context data, and to output action data. The system may also include an imagination module to receive the state and action data, and to output consequent state data. The system may also include a manager to receive the state data and the context data, and to output route data which defines whether the system is to execute an action or to imagine. The system may also include a memory to store the context data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for data-efficient reinforcement learning. One of the systems is a system for training an actor neural network used to select actions to be performed by an agent that interacts with an environment by receiving observations characterizing states of the environment and, in response to each observation, performing an action selected from a continuous space of possible actions, wherein the actor neural network maps observations to next actions in accordance with values of parameters of the actor neural network, and wherein the system comprises: a plurality of workers, wherein each worker is configured to operate independently of each other worker, wherein each worker is associated with a respective agent replica that interacts with a respective replica of the environment during the training of the actor neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network. One of the methods includes maintaining a replay memory that stores trajectories generated as a result of interaction of an agent with an environment; and training an action selection neural network having policy parameters on the trajectories in the replay memory, wherein training the action selection neural network comprises: sampling a trajectory from the replay memory; and adjusting current values of the policy parameters by training the action selection neural network on the trajectory using an off-policy actor critic reinforcement learning technique.

Abstract: A system includes a neural network system implemented by one or more computers. The neural network system is configured to receive an observation characterizing a current state of a real-world environment being interacted with by a robotic agent to perform a robotic task and to process the observation to generate a policy output that defines an action to be performed by the robotic agent in response to the observation. The neural network system includes: (i) a sequence of deep neural networks (DNNs), in which the sequence of DNNs includes a simulation-trained DNN that has been trained on interactions of a simulated version of the robotic agent with a simulated version of the real-world environment to perform a simulated version of the robotic task, and (ii) a first robot-trained DNN that is configured to receive the observation and to process the observation to generate the policy output.

Abstract: A neural network control system for controlling an agent to perform a task in a real-world environment, operates based on both image data and proprioceptive data describing the configuration of the agent. The training of the control system includes both imitation learning, using datasets generated from previous performances of the task, and reinforcement learning, based on rewards calculated from control data output by the control system.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training an actor neural network used to select actions to be performed by an agent interacting with an environment. One of the methods includes obtaining a minibatch of experience tuples; and updating current values of the parameters of the actor neural network, comprising: for each experience tuple in the minibatch: processing the training observation and the training action in the experience tuple using a critic neural network to determine a neural network output for the experience tuple, and determining a target neural network output for the experience tuple; updating current values of the parameters of the critic neural network using errors between the target neural network outputs and the neural network outputs; and updating the current values of the parameters of the actor neural network using the critic neural network.

Abstract: Image processing using masked restricted Boltzmann machines is described. In an embodiment restricted Boltzmann machines based on beta distributions are described which are implemented in an image processing system. In an embodiment a plurality of fields of masked RBMs are connected in series. An image is input into a masked appearance RBM and decomposed into superpixel elements. The superpixel elements output from one appearance RBM are used as input to a further appearance RBM. The outputs from each of the series of fields of RBMs are used in an intelligent image processing system. Embodiments describe training a plurality of RBMs. Embodiments describe using the image processing system for applications such as object recognition and image editing.

Abstract: Image processing using masked restricted Boltzmann machines is described. In an embodiment restricted Boltzmann machines based on beta distributions are described which are implemented in an image processing system. In an embodiment a plurality of fields of masked RBMs are connected in series. An image is input into a masked appearance RBM and decomposed into superpixel elements. The superpixel elements output from one appearance RBM are used as input to a further appearance RBM. The outputs from each of the series of fields of RBMs are used in an intelligent image processing system. Embodiments describe training a plurality of RBMs. Embodiments describe using the image processing system for applications such as object recognition and image editing.

Abstract: A hand-guided power jigsaw (10) with a jigsaw blade (33) which can be detachably clamped between the free ends of two essentially parallel arms (122, 222) preferably comprised of a U-shaped frame (22), in particular comprised of a tube, and which can be driven, in particular in a reciprocating manner, by a motor (11) integrated into the jigsaw (10), wherein a laterally protruding handle (13) with a switch button (15) of an on/off switch is disposed at the free end of one of the arms (221, 222), can be manufactured as lightweight, compact, and inexpensive because the lower arm (122) carries the motor (11) and transmission mechanism (114, 115, 116, 119, 67) for moving the saw blade (33) back and forth, particularly in a base housing (12), wherein a motor shaft (1103) with a crankshaft (116) is coupled parallel to the arm (221) and is coupled by means of a connecting rod (67) to an elastic support (51, 57) in order to clamp the jigsaw blade (33) in place.

Abstract: Method for determining the position of a terminal (14) in a cellular mobile radio network (10). with network measurements being executed in the terminal (14) for selection of a service cell (CEL), with the following steps: transfer (S2) of a measurement data request message from a position determination server (16) to the terminal (14), reading out (S3) of measurement data determined by the network measurements from a measurement data memory (40) in the terminal (14) in response to the request message, transfer (S4) of a response message with the read out measurement data from the terminal (14) to the position determination server (16), calculation (S5) of the position of the terminal (14), based on the measurement data transferred, in the position determination server (16).

Abstract: An electric hand tool guided with both hands has a tool housing, a stationary handle, and an additional movable handle, the additional handle being turnable about a turning axis and being also displaceable longitudinally substantially along the turning axis.

Abstract: A hand machine tool operated with two hands has a machine housing composed of insulating synthetic plastic shells, a fixed, web-shaped main handle formed as a part of the machine housing, an auxiliary handle, a drive including a motor and a transmission, at least one inner housing accommodating the motor and the transmission and surrounded by the synthetic plastic shells, and a tool driven by the drive. The handles, the machine housing, the inner housing and the tool are arranged so that forces applied to at least one of the handles are withdrawn from the machine housing by transferring the forces by the main handle directly to the inner housing without deforming the machine housing and then transferring the forces from the inner housing to the tool.

Abstract: A power-operated sword saw comprises a plate-shaped sword having a lower region with a guiding groove, two saw blades reciprocatingly movable in opposite directions parallel to one another and each having a front saw blade tip and an opposite clamping end. The saw blades have a toothed side provided with a plurality of teeth and an opposite saw blade back guided in the guiding groove of the sword. A drive is coupled with the clamping ends of the saw blades through a clamping unit. The saw blades are jointly coupled with the clamping unit and released from the clamping unit by a blocking element which blocks the clamping ends in a coupling position and release the clamping ends in a contact-free manner.

Abstract: An electric hand saw has a stationary sword having a lower side, a reciprocatable saw blade received in the lower side of the sword and having a rear end portion, a machine housing having a transmission outlet in which the rear end portion of the saw blade is received. The machine housing has a longitudinal slot located under the rear end portion of the saw blade. A suction passage is mounted on the lower side of the machine housing and has a sucking-in opening extending from the front end side of the suction passage to the longitudinal slot of the machine housing located under the rear end portion of the saw blade.

Abstract: A hand held power tool has a working disc and a working sheet held on the working disc by a burdock connection. The working disc has a burdock surface. An intermediate pad is held on the burdock surface of the grinding disc. The intermediate pad has a first side provided with a velour coating and releasably connectable with the burdock surface of the grinding working disc and a second side releasably connectable with the working sheet.