An introspection network is a machine-learned neural network that accelerates training of other neural networks. The introspection network receives a weight history for each of a plurality of weights from a current training step for a target neural network. A weight history includes at least four values for the weight that are obtained during training of the target neural network up to the current step. The introspection network then provides, for each of the plurality of weights, a respective predicted value, based on the weight history. The predicted value for a weight represents a value for the weight in a future training step for the target neural network. Thus, the predicted value represents a jump in the training steps of the target neural network, which reduces the training time of the target neural network. The introspection network then sets each of the plurality of weights to its respective predicted value.

Methods and apparatus for determining whether a media presentation device is in an on state or an off state are disclosed. A disclosed example method comprises determining contribution values from at least one of a signal measured from a sensing device or an output signal accessed from the presentation device, wherein the contribution values are indicative of a state of a presentation device. Summing, via a logic circuit, a first plurality of the contribution values corresponding to a first measurement cycle to generate a first intermediate fuzzy score for the first measurement cycle. Storing the first intermediate fuzzy score in a buffer including a plurality of intermediate fuzzy scores corresponding to respective measurement cycles. Combining, via the logic circuit, the intermediate fuzzy scores corresponding to a first time period to form a final fuzzy score. When the final fuzzy score satisfies a threshold, setting the state of the presentation device as on and enabling crediting of media presented by the presentation device.

A device may receive information associated with first and second bug reports to be classified as duplicate or non-duplicate bug reports. The device may identify first and second descriptions associated with the first and second bug reports, respectively. The first and second descriptions may be different descriptions having a shared description type. The device may identify a neural network for encoding the first and second descriptions, based on the shared description type. The device may encode the first description into a first vector using the neural network, and may encode the second description into a second vector using the neural network. The device may classify the first and second bug reports as duplicate or non-duplicate bug reports based on the first vector and the second vector. The device may perform an action based on classifying the first and second bug reports as duplicate or non-duplicate bug reports.

Provided is a method of processing data based on a neural network, the method including receiving input data; determining a hyper parameter of a first neural network that affects at least one of a speed of the first neural network and an accuracy of the first neural network by processing the input data based on a second neural network; and processing the input data based on the hyper parameter and the first neural network.

A system, apparatus and method for selecting a value for an independent variable that determines an operating state of a system described by a performance function. In one embodiment, the method includes establishing a range of values for the independent variable, selecting a number of values in the range of values to test the independent variable, and selecting random values within the range of values for the independent variable based on the number of values. The method also includes evaluating the performance function at the random values, and selecting the value of the independent variable from the random values that provides an extremum value for the performance function.

A system and method for constructing a mathematical model of a system. The method includes constructing an initial mathematical system representation with a combination of terms, the terms comprising mathematical functions including independent variables dependent on an input signal. A first set of known data is inputted to the initial mathematical representation to generate a corresponding set of output data. The corresponding set of output data of the initial mathematical representation and a second set of known data, correlated to the first set of known data, is fed to a comparator to generate error signals representing differences between output data and correlated members of the second set of known data. A parameter of the combination of terms is iteratively varied to produce a refined mathematical representation of the system until a measure of the error signals is reduced to a value wherein the set of corresponding output data of the refined mathematical representation over a desired range is approximately equivalent to the second set of known data.

A system and method for predicting a characteristic of an object in an artificial intelligence system. The method includes evaluating the object using a first model to produce a first prediction of a characteristic of the object. The object is evaluated using a second model to produce a second prediction of the characteristic of the object, the second model being dissimilar to the first model. A final prediction of the characteristic of the object is generated as a function of dynamic weightings of the first prediction and the second prediction.

A system and method for estimating a system state. The method includes constructing a first estimate of a system state at a first time including a first covariance matrix describing an accuracy of the first estimate. A second estimate of the state is constructed at a second time, after the first time, including a second covariance matrix. A value of a characteristic of the system state is measured at the second time and the second estimate of the system state and the second covariance matrix are adjusted based on the value of the characteristic. A third estimate of the system state is constructed at a third time, before the second time, including a third covariance matrix describing an accuracy of the third estimate. A fourth estimate of the system state is constructed at a fourth time being after the second time.

A simulation management service receives a request to perform reinforcement learning for a robotic device. The request can include computer-executable code defining a reinforcement function for training a reinforcement learning model for the robotic device. In response to the request, the simulation management service generates a simulation environment and injects the computer-executable code into a simulation application for the robotic device. Using the simulation application and the computer-executable code, the simulation management service performs the reinforcement learning within the simulation environment.

A fine-grained analog memory device includes: 1) a charge-trapping transistor including a gate and a high-k gate dielectric; and 2) a pulse generator connected to the gate and configured to apply a positive or negative pulse to the gate to change an amount of charges trapped in the high-k gate dielectric.