Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: A machine learning system that includes one or more machine learning models implemented in one or more hardware processors, a first-level feature creation module, and a combination module provides an output based on one or more channel inputs. Each of the one or more machine learning models receives the channel inputs and additional feature inputs based on the channel inputs to produce the output. The first-level feature creation module receives the channel inputs, performs a feature creation operation, creates the additional feature inputs, and provides the additional feature inputs to at least one of the machine learning models. The first-level feature creation operation performs a calculation on one or more aspects of the channel inputs, and the combination module receives the one or more machine learning model outputs and produce a machine learning channel output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: A machine learning system that includes three machine learning models implemented in a hardware processor, a first-level feature creation module, and a combination module provides an output based on one or more channel inputs. Each of the three machine learning models receives the channel inputs and additional feature inputs based on the channel inputs to produce the output. The first-level feature creation module is implemented in hardware and receives the channel inputs, performs a feature creation operation, creates the additional feature inputs, and provides the additional feature inputs to at least one of the machine learning models. The first-level feature creation operation performs a calculation on one or more aspects of the channel inputs, and the combination module receives the one or more machine learning model outputs and produce a machine learning channel output.

Abstract: A machine learning system that includes one or more machine learning models implemented in one or more hardware processors, a first-level feature creation module, and a combination module provides an output based on one or more channel inputs. Each of the one or more machine learning models receives the channel inputs and additional feature inputs based on the channel inputs to produce the output. The first-level feature creation module receives the channel inputs, performs a feature creation operation, creates the additional feature inputs, and provides the additional feature inputs to at least one of the machine learning models. The first-level feature creation operation performs a calculation on one or more aspects of the channel inputs, and the combination module receives the one or more machine learning model outputs and produce a machine learning channel output.

Abstract: A machine learning system that includes three machine learning models implemented in a hardware processor, a first-level feature creation module, and a combination module provides an output based on one or more channel inputs. Each of the three machine learning models receives the channel inputs and additional feature inputs based on the channel inputs to produce the output. The first-level feature creation module is implemented in hardware and receives the channel inputs, performs a feature creation operation, creates the additional feature inputs, and provides the additional feature inputs to at least one of the machine learning models. The first-level feature creation operation performs a calculation on one or more aspects of the channel inputs, and the combination module receives the one or more machine learning model outputs and produce a machine learning channel output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network used to select actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a plurality of actor computing units and a plurality of learner computing units. The actor computing units generate experience tuple trajectories that are used by the learner computing units to update learner action selection neural network parameters using a reinforcement learning technique. The reinforcement learning technique may be an off-policy actor critic reinforcement learning technique.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: A phased-array antenna filter and diplexer for a super economical broadcast system are provided. The filter and diplexer includes a signal divider tee diplexer, a receive filter and a transmit filter. The diplexer includes a tee branch point, an antenna port, a transmit port and a receive port. The receive filter includes a flat, multi-pole bandpass filter, an input port and an output port, where the input port is coupled to the diplexer receive port to define a receive signal path, from the tee branch point to the receive input port, that has a length of approximately one quarter receive wavelength. The transmit filter includes a folded, multi-pole bandpass filter, an input port and an output port, where the output port is coupled to the diplexer transmit port to define a transmit signal path, from the tee branch point to the transmit output port, that has a length of approximately one quarter transmit wavelength.

Abstract: A filter includes a cross-coupling link which includes a crossbar, a first vertical support attached to one end of the crossbar, a second vertical support attached to another end of the crossbar, a first coupling arm attached to the first vertical support, a second coupling arm attached to the second vertical support, a first adjustable support attached to the first coupling arm at one end and grounded at another end, and a second adjustable support attached to the second coupling arm at one end and grounded at another end.

Abstract: A filter includes a cross-coupling link which includes a crossbar, a first vertical support attached to one end of the crossbar, a second vertical support attached to another end of the crossbar, a first coupling arm attached to the first vertical support, a second coupling arm attached to the second vertical support, a first adjustable support attached to the first coupling arm at one end and grounded at another end, and a second adjustable support attached to the second coupling arm at one end and grounded at another end.

Abstract: A phased-array antenna filter and diplexer for a super economical broadcast system are provided. The filter and diplexer includes a signal divider tee diplexer, a receive filter and a transmit filter. The diplexer includes a tee branch point, an antenna port, a transmit port and a receive port. The receive filter includes a flat, multi-pole bandpass filter, an input port and an output port, where the input port is coupled to the diplexer receive port to define a receive signal path, from the tee branch point to the receive input port, that has a length of approximately one quarter receive wavelength. The transmit filter includes a folded, multi-pole bandpass filter, an input port and an output port, where the output port is coupled to the diplexer transmit port to define a transmit signal path, from the tee branch point to the transmit output port, that has a length of approximately one quarter transmit wavelength.

Abstract: A multi-spectra photon therapy device and method of use for providing treatment to a subject in need thereof is described. A multi-spectra photon therapy device comprising an array of LEDs having at least two sets of LEDs configured according to a predetermined pattern, with a first set of LEDs emitting light in the about red wavelength and at least a second set of LEDs emitting light in the about infrared wavelength; a frequency generator for modulating a signal to the LED array; and a controller for selecting the frequency applied to the signal and the duration during which the signal is modulated at the selected frequency; a protocol database including a plurality of predetermined sequences of frequencies and durations of time specific to treat or ameliorate predetermined diseases or conditions.