Congratulations to Song Leilei on successfully graduating! As the saying goes, "Heaven rewards diligent efforts." In this world, no one becomes a genius without hard work. May you continue to work diligently day and night, and may you achieve success at an early stage!
Work during graduate studies:
To analyze and solve typical small-sample problems in materials, the following approaches can be used to handle and analyze three different types of small-sample data. Targeted solutions for small-sample problems can be developed using relevant theoretical knowledge of machine learning, deep learning, and complex networks.
1. A machine learning-based method for crystal structure recognition has been proposed to address data related to crystal structures. This method involves augmenting the sample data and defining feature representations and feature selection based on the distribution characteristics of atoms in space. Finally, machine learning techniques are utilized to achieve crystal structure recognition.
2. A transfer learning-based recognition method is proposed for Atomic Force Microscopy (AFM) image data. This method involves preprocessing the images to remove noise and improving the watershed segmentation technique for segmenting overlapping objects. Finally, transfer learning is utilized to recognize the target objects.
3. A complex network theory-based image segmentation method is proposed for microscopic structure images of ceramics. This method generates a set of nodes in network space based on the pixel value distribution of the image. It then defines the similarity between nodes and generates a network topology structure. Finally, it optimizes the network topology structure and performs image segmentation.
