Embodiments relate to a system, program product, and method for processing an untrusted data set to automatically determine which data points there are poisonous. A neural network is trained network using potentially poisoned training data. Each of the training data points is classified using the network to retain the activations of at least one hidden layer, and segment those activations by the label of corresponding training data. Clustering is applied to the retained activations of each segment, and a clustering assessment is conducted to remove an identified cluster from the data set, form a new training set, and train a second neural model with the new training set. The removed cluster and corresponding data are applied to the trained second neural model to analyze and classify data in the removed cluster as either legitimate or poisonous.

According to an embodiment, a processing apparatus includes a hardware processor. The hardware processor is configured to: cut out, from an input signal, a plurality of partial signals that are predetermined parts in the input signal; execute processing on the plurality of partial signals using neural networks having the same layer structure with each other to generate a plurality of intermediate signals including a plurality of signals corresponding to a plurality of channels; execute predetermined statistical processing on signals for each of the plurality of channels for each of the plurality of intermediate signals corresponding to the plurality of partial signals, to calculate statistics for each channel and generate a concatenated signal by concatenating the statistics of the plurality of respective intermediate signals for each channel; generate a synthetic signal by performing predetermined processing on the concatenated signal; and output an output signal in accordance with the synthetic signal.

A system for validating installation correctness of a component in a test assembly includes a housing having a platform including a tiered surface. The tiered surface forms an abutment surface configured as a stop against which a test assembly is abutted. A plurality of cameras is positioned to capture different views of the test assembly. A processing device is configured to execute instructions to capture an image from each of the plurality of cameras of the test assembly that includes a plurality of components. Each of the plurality of components is analyzed within each image of the plurality of images. A matching score is determined and an indication of whether the plurality of components was correctly installed in the test assembly is generated.

A computer assesses device log entries. The computer receives a training log entry and an input log entry from a log entry corpus. The computer determines for the training log entry, status indicators respective to the group of log entries, The indicators are based on processing the training log entry with a group of unsupervised Machine Learning models calibrated to identify outliers. The computer assigns an outlier status based on the processing to the training log entry. The computer trains a supervised ML learning model with a data pair of the training log entry and an associated data label representing the assigned outlier status value. The computer processes the input log entry with the supervised ML model to predict an input log classification, and the log classification indicates whether the input log is anomaly. The computer generates an input log entry assessment report including the input log entry classification.

The disclosure herein relates to AI-based methods and systems of using machine-learning to identify deceptive merchants in payment transactions such as recurring payment transactions. For example, the AI-based systems and methods may train and use an aggregate merchant matcher based on entity matching to identify merchant identifiers and/or acquirers that may be used by a merchant, train and use transaction classifiers to classify transactions as deceptive, recognize merchants based on an N-density aware transaction embedding learned from transaction data, and train and use a merchant classifier to classify merchants as deceptive.

In accordance with various aspects of the present disclosure, methods and systems for fire detection are disclosed. In some embodiments, a method for fire detection includes: acquiring data related to a monitored area, wherein the data comprises image data related to the monitored area; determining, based on the acquired data, whether a first mode or a second mode is to be executed for fire detection, wherein the first mode comprises a first smoke detection, and the second mode comprises a first flame detection; and executing, by a hardware processor, at least one of the first mode or the second mode based on a result of the determination.

Provided is a method for detecting a driver's abnormalities based on a CAN (Controller Area Network) bus network communicating with an ECU (Electronic Control Unit) of a vehicle. The method may include: acquiring a CAN bus signal related to an operation of the vehicle from the CAN bus network; extracting a detection vector from the CAN bus signal using an auto encoder; and detecting a driver's abnormality based on the detection vector.

Described are methods and systems for training a machine learning (ML) model to detect anomalies in images of documents. A first image of a first set of images of documents is obtained. Each first image relates to a region of the document and the first set of images comprises an image of a document containing an anomaly and an image of a document not containing an anomaly. Signal processing algorithms are applied to the first images to generate a signal for each first image and each algorithm, and a discriminative power of each algorithm is evaluated. Based on the discriminative power, a signal processing algorithm is selected and ML model input data is generated using signals generated by applying the algorithm to second digital images. The ML model is trained using the input data to produce output indicating whether an image of a document contains an anomaly.

A method of object authentication based on digital blueprints and physical fingerprints comprising the steps of acquiring a set of training blueprints and fingerprints, training, object enrollment and object authentication. The method uses a pair of a mapper realized as an encoder and a decoder and a set of multi-metric scores originating from the decomposition of mutual information and applied to both the output of the encoder and decoder and producing a feature vector for a one-class classifier. The method is trained only on the original physical objects and does not use any fakes for reliable authentication.

A method of handling outliers is provided. The method includes determining a set of residuals, wherein each residual represents a difference between a measurement included in a set of measurements and a predetermined estimate; clustering the residuals into a plurality of clusters; calculating a consistency value for each of the plurality of clusters based on a number of measurements included in the set of measurements and a standard deviation of the measurements; identifying a cluster having a maximum consistency value among the plurality of clusters as inliers by applying the consistency function to the plurality of clusters; and handling the outliers based on an approximation of one or more parameters as a function of a statistical relationship of the inliers included in the cluster having the maximum consistency value among the plurality of clusters and an initial estimation of the one or more parameters.