| Time：1130410 (Wed.) 14:20~16:20 Speaker：Prof.王道維(Daw-Wei Wang) 國立清華大學物理學系/Department of physics, National Tsing Hua University Title：Applications of Self-Supervised Machine Learning in Condensed Matter Physics Abstract：We propose a self-supervised machine learning approach to accurately identify and classify different types of phase transitions in various condensed matter and solid-state systems solely based on experimental measurements or simulated data, without relying on a priori knowledge of the theoretical models. Our methodology employs self-supervised ensemble learning (SSEL) to analyze the fluctuation properties of machine learning outputs, enabling the detection of higher-order correlations between physical quantities. We demonstrate the robustness and versatility of our approach by applying it to several benchmark models, including the 2D Potts model, 2D Clock model, 1D Ising model, and 1D XXZ model. Our SSEL method can reliably discern first-order, second-order, and Berezinskii-Kosterlitz-Thouless transitions using various input features, such as time-of-flight images, spatial correlation functions, density-density correlation functions, and in situ spin configurations. Notably, our self-supervised learning approach identifies phase transition points by detecting the largest deviation of the predicted results from the known system parameters and the highest confidence through a systematic shift of the training regions, overcoming the sensitivity to training region and data labeling issues faced by conventional supervised learning methods. We will also provide some more examples for the investigation of other physical properties in condensed matter systems. Place：B101, Gongguan Campus, NTNU |
