Abstract: Computer-implemented techniques can include obtaining, by a client computing device, a digital media item and a request for a processing task on the digital item and determining a set of operating parameters based on (i) available computing resources at the client computing device and (ii) a condition of a network. Based on the set of operating parameters, the client computing device or a server computing device can select one of a plurality of artificial neural networks (ANNs), each ANN defining which portions of the processing task are to be performed by the client and server computing devices. The client and server computing devices can coordinate processing of the processing task according to the selected ANN. The client computing device can also obtain final processing results corresponding to a final evaluation of the processing task and generate an output based on the final processing results.

Abstract: A learning method for transforming a virtual video on a virtual world to a more real-looking video is provided. And the method includes steps of: (a) a learning device instructing a generating CNN to apply a convolutional operation to an N-th virtual training image, N-th meta data and (N-K)-th reference information to generate an N-th feature map; (b) the learning device instructing the generating CNN to apply a deconvolutional operation to the N-th feature map to generate an N-th transformed image; (c) the learning device instructing a discriminating CNN to apply a discriminating CNN operation to the N-th transformed image to generate a category score vector; (d) the learning device instructing the generating CNN to generate a generating CNN loss by referring to the category score vector and its corresponding GT, and to perform backpropagation by referring to the generating CNN loss to learn parameters of the generating CNN.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network. One of the methods includes receiving training data; training a neural network on the training data, wherein the neural network is configured to: receive a network input, convert the network input into a latent representation of the network input, and process the latent representation to generate a network output from the network input, and wherein training the neural network on the training data comprises training the neural network on a variational information bottleneck objective that encourages, for each training input, the latent representation generated for the training input to have low mutual information with the training input while the network output generated for the training input has high mutual information with the target output for the training input.

Abstract: Aspects of this disclosure include technologies to detect unpackaged, unlabeled, or mislabeled products based on product images. Leveraging from improved machine learning techniques, the disclosed technical solution can reduce a full product space for product search to a partial product space. Accordingly, a limited number of product candidates in the partial product space may be visually presented for user evaluation. Sometimes, a product candidate is to be comparatively presented with a live image of the product via a uniquely designed graphic user interface, which further improves the confidence of the user and the accuracy of the underlying transaction.

Abstract: Disclosed are systems, methods, and non-transitory computer-readable storage media for predicting a future context of a computing device. In some implementations, a context daemon can use historical context information to predict future events and/or context changes. For example, the context daemon can analyze historical context information to predict user sleep patterns, user exercise patterns, and/or other user activity. In some implementations, a context client can register a callback for a predicted future context. For example, the context client can request to be notified ten minutes in advance of a predicted event and/or context change. The context daemon can use the prediction to notify a context client in advance of the predicted event.

Abstract: A method for transferring acoustic knowledge of a trained acoustic model (AM) to a neural network (NN) includes reading, into memory, the NN and the AM, the AM being trained with target domain data, and a set of training data including a set of phoneme data, the set of training data being data obtained from a domain different from a target domain for the target domain data, inputting training data from the set of training data into the AM, calculating one or more posterior probabilities of context-dependent states corresponding to phonemes in a phoneme class of a phoneme to which each frame in the training data belongs, and generating a posterior probability vector from the one or more posterior probabilities, as a soft label for the NN, and inputting the training data into the NN and updating the NN, using the soft label.

Abstract: Provided are a computerized image interpretation method and a device for analyzing a medical image. The image interpretation method may include receiving, at a processor, a medical image, and receiving report information including a healthcare worker's judgement result of the medical image. The method may also include generating, at the processor, result information representing correspondence between first lesion information, which is related to a lesion in the medical image acquired on the basis of the medical image, and second lesion information, which is related to a lesion in the medical image acquired on the basis of the report information, by applying the first lesion information and the second lesion information to a third analysis model. The method may further include outputting, at the processor, the result information.

Abstract: Methods, systems, and devices are disclosed for determining causality among data. In one aspect, a method includes receiving a first data and a second data associated with a first event and a second event, respectively; and determining a data set including quantitative causality dependence values between the first and the second data.

Abstract: Various techniques employed by an application performance management service to generate an application behavior learning based capacity forecast model are disclosed. In some embodiments, such a capacity forecast model is at least in part generated by clustering collected transaction data into one or more usage patterns, analyzing collected usage pattern data, and solving a mathematical model generated from the usage pattern data to determine a sensitivity of a resource to each type of transaction associated with an application.

Abstract: A computer-implemented method for deriving biopsy results in a non-invasive manner includes acquiring a plurality of training data items. Each training data item comprises non-invasive patient data and one or more biopsy derived scores associated with an individual. The method further includes extracting a plurality of features from the non-invasive patient data based on the one or more biopsy derived scores and training a predictive model to generate a predicted biopsy score based on the plurality of features and the one or more biopsy derived scores.

Abstract: Embedding representation for a document is generated based on clustering words in the document. Representative clusters are selected and a weighted sum of the embeddings of the words in the selected clusters is determined as a document embedding. Documents are labeled based on document embeddings. A machine learning algorithm is trained using the documents. The machine learning algorithm predicts a label of a given document based on the given document's document embedding.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for environment simulation. In one aspect, a system comprises a recurrent neural network configured to, at each of a plurality of time steps, receive a preceding action for a preceding time step, update a preceding initial hidden state of the recurrent neural network from the preceding time step using the preceding action, update a preceding cell state of the recurrent neural network from the preceding time step using at least the initial hidden state for the time step, and determine a final hidden state for the time step using the cell state for the time step. The system further comprises a decoder neural network configured to receive the final hidden state for the time step and process the final hidden state to generate a predicted observation characterizing a predicted state of the environment at the time step.

Abstract: In an example, one or more member profiles and corresponding Boolean attributes indicating, for each of the one or more member profiles, whether the corresponding member of a social networking service interacted with a request for confidential data, are obtained. A first set of one or more features are extracted from the one or more member profiles. The first set of one or more features and corresponding Boolean attributes are fed into a machine learning algorithm to train a confidential data response propensity prediction model to output a predicted propensity to interact with a request for confidential data for a candidate member profile. A second set of one or more features are extracted from the candidate member profile. The extracted second set of one or more features are fed to the confidential data response propensity prediction model, outputting the predicted propensity to interact with a request for confidential data.

Abstract: Techniques for inferring the identity (e.g., member profile attributes) of members of an online social network service are described. According to various embodiments, a member profile attribute missing from a member profile page associated with a particular member of an online social network service is identified. Member profile data and behavioral log data associated with a plurality of members of the online social network service is then accessed. Thereafter, a prediction modeling process is performed, based on a prediction model and feature data including the member profile data and the behavioral log data, to generate a confidence score associated with the particular member and the missing member profile attribute, the confidence score indicating a likelihood that the missing member profile attribute corresponds to a candidate value.

Abstract: An approach is provided in which an information handling system trains multiple classifiers using a set of training samples. The information handling system selects a leader classifier from the multiple classifiers that generates the most amount of correct decisions corresponding to the set of training samples. Next, the information handling system identifies an endorser classifier from the multiple classifiers that generates the highest proportion of correct decisions among the endorser classifier's decisions matching the leader classifier's decisions, and combines the leader classifier and the endorser classifier into a combined classifier stage. In turn, the information handling system utilizes the combined classifier stage to process inquiries and generate results.

Abstract: In an example embodiment, a machine learning algorithm is used to train an engagement score model to calculate an engagement score for a particular member indicating a probability that the particular member would increase engagement with the social networking service if provided with statistical information about confidential data submitted by other members. Member usage information is obtained corresponding to a first member of a social networking service. Then a plurality of features are extracted from the member usage information corresponding to the first member. This plurality of features is inputted into the engagement model to obtain an engagement score for the first member. It is then determined whether or not to provide statistical information to the first member about confidential data submitted by other members based on the engagement score for the first member.

Abstract: A computer-implemented method for selecting features for classification may include (1) generating a matrix X, a column vector Y, and a matrix Z from a training dataset that includes a plurality of samples with a plurality of features, (2) generating an augmented matrix from the matrix X, the column vector Y, and the matrix Z, (3) identifying one or more most-relevant features from the plurality of features by iteratively applying a sweep operation to the augmented matrix, and (4) training a classification model using the most-relevant features from the plurality of features rather than all of the plurality of features. Various other methods, systems, and computer-readable media may have similar features.