Fast and accurate discrimination high-speed Raman imaging combined with multi-armed bandit algorithm in reinforcement learning

We propose a method that combines high-speed Raman imaging with a machine learning technique, multi-armed bandit, to achieve rapid and accurate identification of samples under observation. First, our method dvides the field of view of the sample into small sections, and it returns either 'positive' or 'negative' based on whether the sections with high anomaly indices exceed a certain proportion. Moreover, the points to be measured are determined dynamically and automatically generating a series of optimal illumination patterns.