An image classifier comprises a first classifier and a second classifier. The first classifier comprises L individual classifiers, which are trained at different, respective image resolutions from a first full-resolution level to a lowest-resolution level. Outputs of the first set of classifiers are used to train the second classifier at the full-resolution level. Accordingly, the second classifier exploits contextual information at multiple different image resolutions. The classifiers may be trained to optimize a joint posterior probability at multiple resolutions.

Systems and methods are described for user classification with semi-supervised machine learning. The systems and methods may include receiving user information for a first set of users, receiving survey data for a second set of users wherein the second set of users is a proper subset of the first set of users, training a first neural network and a second neural network based on the second set of users, mapping the user information for the first set of users to the embedding space using the first neural network, predicting category membership propensities for the first set of users using a low-density separation algorithm on the user information for the first set of users mapped to the embedding space, updating the first neural network and the second neural network based on the prediction, and reclassifying the first set of users based on the updated first neural network and the updated second neural network.

Computational techniques are applied to video images of polyps to extract features and patterns from different perspectives of a polyp. The extracted features and patterns are synthesized using registration techniques to remove artifacts and noise, thereby generating improved images for the polyp. The generated images of each polyp can be used for training and testing purposes, where a machine learning system separates two types of polyps.

A valuation platform may receive, from a user device, video data associated with a vehicle, and may receive a vehicle history report of the vehicle based on a vehicle identification number of the vehicle. The valuation platform may receive, from the user device, feature data associated with the vehicle, and may process the video data, the vehicle history report, and the feature data, with a machine learning model, to determine one or more values for the vehicle. The valuation platform may determine a valuation for the vehicle based on the determined one or more values for the vehicle. The valuation platform may create a vehicle profile for the vehicle based on the video data, the vehicle history report, the feature data, the determined one or more values for the vehicle, and the valuation for the vehicle, and may perform one or more actions based on the vehicle profile.

Neural network classification may be performed by inputting a training data set into each of a plurality of first neural networks, the training data set including a plurality of samples, obtaining a plurality of output value sets from the plurality of first neural networks, each output value set including a plurality of output values corresponding to one of the plurality of samples, each output value being output from a corresponding first neural network in response to the inputting of one of the samples of the training data set, inputting the plurality of output value sets into a second neural network, and training the second neural network to output an expected result corresponding to each sample in response to the inputting of a corresponding output value set.

A product defect detection method, device and system are disclosed. The product defect detection method comprises: constructing a defect detection framework including a classification network, a locating detection network and a judgment network; training the classification network by using a sample image of a product containing different defect types to obtain a classification network capable of classifying the defect types existing in the sample image; training the locating detection network by using a sample image of a product containing different defect types to obtain a locating detection network capable of locating a position of each type of defect in the sample image; inputting an acquired product image into the defect detection framework, inputting a classification result and a detection result obtained into the judgment network to judge whether the product has a defect, and detecting a defect type and a defect position when the product has a defect.

The technology disclosed relates to a system. The system comprises a trained multi-label support vector machine running a one-vs-the-rest classifier. The trained multi-label support vector machine running a one-vs-the-rest classifier is configured with trained parameters. The trained parameters are learned from training the trained multi-label support vector machine running the one-vs-the-rest classifier on document features of documents belonging to a plurality of label classes, and hyperplane determinations on label classes in the plurality of label classes. The trained parameters include distributions of distances between the label classes and the hyperplanes.

An automatic target recognizer system including: a database that stores target recognition data including multiple reference features associated with each of multiple reference targets; a pre-selector that selects a portion of the target recognition data based on a reference gating feature of the multiple reference features; a preprocessor that processes an image received from an image acquisition system which is associated with an acquired target and determines an acquired gating feature of the acquired target; a feature extractor and processor that discriminates the acquired gating feature with the reference gating feature and, if there is a match, extracts multiple segments of the image and detects the presence, absence, probability or likelihood of one of multiple features of each of the multiple reference targets; a classifier that generates a classification decision report based on a determined classification of the acquired target; and a user interface that displays the classification decision report.

An element construction unit compares output values of one or more elements included in an intermediate layer calculated by an output value calculating unit with a threshold value, and the number of elements included in the intermediate layer is maintained when any of the output values out of the output values of the one or more elements included in the intermediate layer is greater than the threshold value, and the number of elements included in the intermediate layer is increased when all of the output values of the one or more elements included in the intermediate layer are equal to or less than the threshold value.

A method for training thresholds controlling data flow in a plurality of cascaded classifiers for classifying malicious software, comprising: in each of a plurality of iterations: computing a set of scores, each for one of a set of threshold sequences, each threshold sequence is a sequence of sets of classifier output thresholds, each set of classifier output thresholds used to control a flow of data from a first cascaded classifier of the plurality of cascaded classifiers to a second cascaded classifier of the plurality of cascaded classifiers, each score computed when classifying, using the respective threshold sequence, each of a plurality of software objects as one of a set of maliciousness classes; computing a set of new threshold sequences by applying a genetic algorithm to the set of threshold sequences and the set of scores; and using the set of new threshold sequences in a consecutive iteration.