Kernel Filter-Based Image Construction from Scanning Electrochemical Microscopy Approach Curves

Abstract

A scanning electrochemical microscope (SECM) with artificial intelligence could generate the sample's activity image from approach curves measured at several points, minimizing the time of measurement and calculating sample activity in points of interest. The time used to shape the separation between the feature space and regression line in the convolutional process for CNNs and DNNs constitutes a significant contribution to the accuracy of the AI model performance. Kernel functions and pre- processing of synthetic data can achieve higher efficiency and localization precision by applying them to the initial layers of an MLP network; assuming an infinite-width network, we can use the theory from NTK to consider the kernel shape and function. In this paper, we compare model accuracy and performance with and without rescaled data for under-sampled data sets and determine the effectiveness of CNN Fourier kernel interpolation filtering to detect high-frequency features for non-image data in multi-layer perceptron (MLP) networks. The results show how engineered feature mapping and data shaping can affect the convergence of several types of Gaussian and Fourier mapping methods for training and validation convergence with custom activation and modified filter functions, as well as model learning of highly sampled features in proprietary data bands.