Classifier accuracy is increased in machine learning applications by training a machine learning (ML) model including a classifier across classes by determining weighted input points for a contributing set to store the values for true positive and true negative predications. In a following step, input data is provided to the classifier of the machine learning model at runtime, and a classification output is determined from the classifier. For the classification output, values for input during runtime are compared with a sample of inputs stored for training the machine learning model to determine the distance in spread for the classification output. A class is determined from the classification output having a smallest distance and spread. The method can further determine if the class with the smallest distance and spread is a true positive or true negative by comparing the class with the smallest distance and spread with the classification output.

An apparatus, method, and a computer readable medium for attenuating visual artifacts in processed images. An annotated dataset of images to be processed by an image processing system is created. An adversarial control network is trained to operate as an image quality expert in classifying images. After the adversarial control network has been trained, the adversarial control network is used to supervise the image processing system on-the-fly.

A method for detecting fraudulent interactions may include receiving interaction data, including a first plurality of interactions with (first) fraud labels and a second plurality of interactions (without fraud labels). Second fraud label data for each of the second plurality of interactions may be generated with a first neural network (e.g., classifying whether each interaction is fraudulent or not). Generated interaction data and generated fraud label data may be generated with a second neural network. Discrimination data for each of the second plurality of interactions and generated interactions may be generated with a third neural network (e.g., classifying whether the respective interaction is real or not). Error data may be determined based on the discrimination data (e.g., whether the respective interaction is correctly classified). At least one of the neural networks may be trained based on the error data. A system and computer program product are also disclosed.

Techniques for machine learning optimization are provided. A video comprising a plurality of segments is received, and a first segment of the plurality of segments is processed with a machine learning (ML) model to generate a plurality of tags, where each of the plurality of tags indicates presence of an element in the first segment. A respective accuracy value is determined for each respective tag of the plurality of tags, where the respective accuracy value is based at least in part on a maturity score for the ML model. The first segment is classified as accurate, based on determining that an aggregate accuracy of tags corresponding to the first segment exceeds a predefined threshold. Upon classifying the first segment as accurate, the first segment is bypassed during a review process.

A method, system and computer-readable storage medium for performing a counterfactual generation operation. The counterfactual generation operation includes: receiving a subject data point; classifying the data point via a trained classifier, the classifying providing a classified data point; identifying a counterfactual using the classified data point, the counterfactual comprising another datapoint, the another data point being close to the subject data point, the another data point resulting in production of a different outcome when provided to a model when compared to an outcome resulting from the subject data point being provided to the model; and, providing the counterfactual to a destination.

The exemplary embodiments disclose a system and method, a computer program product, and a computer system for diagnosing technical issues. The exemplary embodiments may include collecting data relating to one or more technical issues, extracting one or more features from the collected data, determining one or more diagnoses based on the extracted one or more features and one or more models, and suggesting to a support agent one or more actions based on the one or more determined diagnoses.

Provided is a process including: writing classes using object-oriented modelling of modeling topics; scanning the classes to determine class definition information; receiving from a subscribing modeling object a request for a subscription to a given modeling topic in a given modeling topic class, the subscription request including a modeling topic filter to select the given modeling topic from a plurality of modeling topics described by the given modeling topic class; registering a modeling topic accessor associated with the subscribing modeling object and a modeling topic mutator associated with the subscribing modeling object; processing, through the modeling topic filter a modeling topic that is accessed through an accessor and is described by the modeling topic class, the modeling topic being received from a modeling publisher object; notifying the subscribing object of the received modeling topic through a registered modeling topic listener; and mutating the received modeling topic.

Techniques for synthetic data generation in computer-based reasoning systems are discussed and include receiving a request for generation of synthetic data based on a set of training data cases. One or more focal training data cases are determined. For undetermined features (either all of them or those that are not subject to conditions), a value for the feature is determined based on the focal cases. In some embodiments, the generated synthetic data may be checked for similarity against the training data, and if similarity conditions are met, it may be modified (e.g., resampled), removed, and/or replaced.

In a variety of embodiments, machine classifiers may model multi-turn dialogue as a one-to-many prediction task. The machine classifier may be trained using adversarial bootstrapping between a generator and a discriminator with multi-turn capabilities. The machine classifiers may be trained in both auto-regressive and traditional teacher-forcing modes, with the generator including a hierarchical recurrent encoder-decoder network and the discriminator including a bi-directional recurrent neural network. The discriminator input may include a mixture of ground truth labels, the teacher-forcing outputs of the generator, and/or noise data. This mixture of input data may allow for richer feedback on the autoregressive outputs of the generator. The outputs can be ranked based on the discriminator feedback and a response selected from the ranked outputs.

A method for identifying errors associated with named entity recognition includes recognizing a candidate named entity within a text and extracting a chunk from the text containing the candidate named entity. The method further includes creating a feature vector associated with the chunk and analyzing the feature vector for an indication of an error associated with the candidate named entity. The method also includes correcting the error associated with the candidate named entity.