Abstract: Reinforcement learning enables a framework of information technology assets that include software elements, computational hardware assets, and/or, bundled software and computational hardware systems and products. The performance of successive sessions of an inner loop reinforcement learning is directed and monitored by an outer loop reinforcement learning wherein the outer loop reinforcement learning is designed to reduce financial costs and computational asset requirements and/or optimize learning time in successive instantiations of inner loop reinforcement learning training sessions. The framework enables consideration of the license costs of domain specific simulators, the usage cost of hardware platforms, and the progress of a particular reinforcement learning training. The framework further enables reductions of these costs to orchestrate and train a neural network under budget constraints with respect to the available hardware and software licenses available at runtime.

Abstract: Reinforcement learning enables a framework of information technology assets that include software elements, computational hardware assets, and/or, bundled software and computational hardware systems and products. The performance of successive sessions of an inner loop reinforcement learning is directed and monitored by an outer loop reinforcement learning wherein the outer loop reinforcement learning is designed to reduce financial costs and computational asset requirements and/or optimize learning time in successive instantiations of inner loop reinforcement learning training sessions. The framework enables consideration of the license costs of domain specific simulators, the usage cost of hardware platforms, and the progress of a particular reinforcement learning training. The framework further enables reductions of these costs to orchestrate and train a neural network under budget constraints with respect to the available hardware and software licenses available at runtime.

Abstract: Systems and methods for real time monitoring of acoustic environment using ultrasound are provided. An example method includes causing production of an acoustic sensing signal. The acoustic sensing signal can be a continuously repeating ultrasonic signal. The method further includes receiving a reflected acoustic signal, the reflected acoustic signal being a representation of an acoustic signal reflected within the acoustic environment. The method further includes analyzing the reflected acoustic signal based at least in part on the acoustic sensing signal to determine properties of the acoustic environment. The acoustic sensing signal can be produced based on a pseudorandom binary sequence. Analyzing the reflected acoustic signal includes correlating the reflected acoustic signal and the acoustic sensing signal to determine a real time estimate of impulse response of the acoustic environment. The method further allows determining that the properties of the acoustic environment have changed.

Abstract: A scalable neural network system may include a root processor and a plurality of neural network processors with a tree of synchronizing sub-systems connecting them together. Each synchronization sub-system may connect one parent to a plurality of children. Furthermore, each of the synchronizing sub-systems may simultaneously distribute weight updates from the root processor to the plurality of neural network processors, while statistically combining corresponding weight gradients from its children into single statistical weight gradients. A generalized network of sensor-controllers may have a similar structure.

Abstract: A multi-processor system for batched pattern recognition may utilize a plurality of different types of neural network processors and may perform batched sets of pattern recognition jobs on a two-dimensional array of inner product units (IPUs) by iteratively applying layers of image data to the IPUs in one dimension, while streaming neural weights from an external memory to the IPUs in the other dimension. The system may also include a load scheduler, which may schedule batched jobs from multiple job dispatchers, via initiators, to one or more batched neural network processors for executing the neural network computations.

Abstract: A multi-processor system for data processing may utilize a plurality of different types of neural network processors to perform, e.g., learning and pattern recognition. The system may also include a scheduler, which may select from the available units for executing the neural network computations, which units may include standard multi-processors, graphic processor units (GPUs), virtual machines, or neural network processing architectures with fixed or reconfigurable interconnects.

Abstract: Methods and systems for a subband acoustic echo canceller with integrated voice activity detection are disclosed and may include adjusting transmit and/or receive powers of wirelessly communicated audio signals based on voice activity detection via subband analysis of the wirelessly communicated audio signals. The receive power may be adjusted by utilizing a reduced duty cycle, or by conveying voice activity detection information via an asynchronous control channel in a Bluetooth application. A plurality of subbands may be generated utilizing a fast Fourier transform, and a first subset of the subbands corresponding to voice activity may be selected and a second subset of the subbands may be selected that corresponds to background noise. The processing of the subsets may be dynamically adjusted due to variations in the voice activity or background noise. Comfort noise may be generated and transmitted at a reduced bandwidth utilizing the second subset of the subbands.

Abstract: Provided are systems and methods for providing reduced audio latency in wireless communications. One electronic system providing reduced audio latency includes a host unit for converting audio data, a digital interface coupling the host unit and a wireless transceiver, where the wireless transceiver has a controller including a rate adapter, and where the controller is configured to monitor a rate mismatch between the host unit and the wireless transceiver and to compensate for the rate mismatch using the rate adapter, thereby reducing the audio latency. One controller includes an audio codec for encoding and decoding the audio data, where the controller is further configured to align a frame of encoded audio data and a transmission packet of the wireless transceiver, thereby further reducing the audio latency.

Abstract: Provided are systems and methods for providing reduced audio latency in wireless communications. One electronic system providing reduced audio latency includes a host unit for converting audio data, a digital interface coupling the host unit and a wireless transceiver, where the wireless transceiver has a controller including a rate adapter, and where the controller is configured to monitor a rate mismatch between the host unit and the wireless transceiver and to compensate for the rate mismatch using the rate adapter, thereby reducing the audio latency. One controller includes an audio codec for encoding and decoding the audio data, where the controller is further configured to align a frame of encoded audio data and a transmission packet of the wireless transceiver, thereby further reducing the audio latency.

Abstract: Allocating resources in a circuit switched data network, comprising receiving a request for a resource from a device coupled to the circuit switched data network and granting the resource to the requesting device if the resource is available. If the resource is not available, then examining the instantaneous quantity of data to be transmitted by the requesting device; the rate of change in the instantaneous quantity of data to be transmitted by the requesting device; and the time of utilization of the resource by the requesting device, and granting the resource to the requesting device based on the examination of the three factors.

Abstract: A charge stored in a battery is estimated by multiplying estimated charge currents supplied during charge phases and estimated discharge currents discharged during operational modes by the respective amounts of time of charging and discharge. The state of charge of the battery is calculated from the estimated stored charge. Furthermore, parameters of the battery may be measured to be used to adjust the estimated state of charge of the battery. In a first case, the estimated stored charge may be adjusted to a charge value corresponding to the detected completion of a constant current charge phase. In a second case, the estimated stored charge may be adjusted to a charge value corresponding to the detected completion of a constant voltage charge phase. In a third case, the estimated stored charge may be adjusted to a charge value corresponding to a detected predetermined discharge level of the battery.

Abstract: Methods and systems for a subband acoustic echo canceller with integrated voice activity detection are disclosed and may include adjusting transmit and/or receive powers of wirelessly communicated audio signals based on voice activity detection via subband analysis of the wirelessly communicated audio signals. The receive power may be adjusted by utilizing a reduced duty cycle, or by conveying voice activity detection information via an asynchronous control channel in a Bluetooth application. A plurality of subbands may be generated utilizing a fast Fourier transform, and a first subset of the subbands corresponding to voice activity may be selected and a second subset of the subbands may be selected that corresponds to background noise. The processing of the subsets may be dynamically adjusted due to variations in the voice activity or background noise. Comfort noise may be generated and transmitted at a reduced bandwidth utilizing the second subset of the subbands.

Abstract: Allocating resources in a circuit switched data network, comprising receiving a request for a resource from a device coupled to the circuit switched data network and granting the resource to the requesting device if the resource is available. If the resource is not available, then examining the instantaneous quantity of data to be transmitted by the requesting device; the rate of change in the instantaneous quantity of data to be transmitted by the requesting device; and the time of utilization of the resource by the requesting device, and granting the resource to the requesting device based on the examination of the three factors.

Abstract: Allocating resources in a circuit switched data network, comprising receiving a request for a resource from a device coupled to the circuit switched data network and granting the resource to the requesting device if the resource is available. If the resource is not available, then examining the instantaneous quantity of data to be transmitted by the requesting device; the rate of change in the instantaneous quantity of data to be transmitted by the requesting device; and the time of utilization of the resource by the requesting device, and granting the resource to the requesting device based on the examination of the three factors.

Abstract: Allocating resources in a circuit switched data network, comprising receiving a request for a resource from a device coupled to the circuit switched data network and granting the resource to the requesting device if the resource is available. If the resource is not available, then examining the instantaneous quantity of data to be transmitted by the requesting device; the rate of change in the instantaneous quantity of data to be transmitted by the requesting device; and the time of utilization of the resource by the requesting device, and granting the resource to the requesting device based on the examination of the three factors.