About
Microtrace’s Alerion AI Classification technology goes far beyond limitations of conventional support vector machine (SVM) and neural network (NN) modeling. Alerion Classification is trained using artificial intelligence to recognize specific targets and has an additional, expansion class for previously unknown samples. Samples falling outside the trained classes are uniquely classified into the additional class until they are trained otherwise—which is vastly different than conventional modeling where unknown samples are forced into a trained class. Since model creation and continuous development and optimization can be done using data collected from one or more sensors, a limitless number of internet of things (IOT) devices can be used for field or market surveillance to create a true smart network. None of this is possible with conventional SVM or NN modeling.
Media
Products media viewer
-
Alerion? AI Classification Technology
Patent-pending AI classification trained to automatically recognize targets from both known and unknown classes for adaptive learning - add new classes, enhanced targeting and warning thresholds.
-
Alerion? AI Classification Technology
Pushing past the limits of conventional modeling
-
Adaptive Learning: Addition of new classes
New classes can be created from previously unclassified samples that are captured in the expansion category. Regardless of how many sensors you used, your modeling will always provide the most accurate and up-to-date insights for all samples in the market. Uniquely, Alerion provides accurate classification of any sample even if only some are known at the time of training. Knowledge of what might be encountered in the future is not needed to achieve accurate classification of all sample types.
-
Adaptive Learning: Enhanced Targeting
Individual classes can be optimized to accurately account for variations in samples captured.
-
Adaptive Larning: Warning Thresholds
Warning thresholds allow you to take appropriate action and can provide you with unique insights. Thresholds (one or more) can be placed around class boundaries to alert you if: - A shift is detected in your known classes, which may indicate manufacturing issues, degradation or variants - Unknown samples approach your known classes, which could mean competitors have a close match to your genuine sample, or discoveries of similar previously unknown samples