School of Engineering Department of Chemical and Biological Engineering 75 Generation of Adversarial Chemical Reactions Supervisor: GAO, Hanyu / CBE Student: OH, Jintaek / BIEN Course: UROP1100, Fall UROP2100, Spring This progress report outlines an ongoing study employing the T5 base model to predict reaction conditions, including agents and solvents, in chemical reactions. The model is initially trained to predict products of reactions given reactants, agents, and solvents, using the comprehensive Pistachio chemical reaction dataset. Despite hardware and time limitations, preliminary findings display promising progress, demonstrating a meaningful decrease in loss value. Future steps encompass pre-training the model, upgrading to a larger T5 model, expanding the dataset, optimizing hyperparameters, implementing evaluation metrics, and investigating real-world applications. This project strives to utilize machine learning to augment chemical reaction predictability, providing valuable tools for researchers and chemists. Skin-adherent Bioelectronic ECG Patch for Ambulatory Care Supervisor: HSING, I-ming / CBE Co-supervisor: NYEIN, Hnin Yin Yin / CBE Student: FONG, Sin Ning / CPEG HUI, Ka Ching / CENG Course: UROP1000, Summer UROP1000, Summer The need for continuous electrocardiography (ECG) monitoring to prevent sudden cardiac events and premature death has led to the development of wearable bioelectronics, providing non-invasive ambulatory monitoring for cardiovascular disease (CVD) detection and diagnosis. Here, a 12-lead ECG patch with improved compliance, conformity, and minimal skin damage over conventional ECG sensors, is presented. Clean, consistent ECG signals are recorded owing to the use of stretchable, skin-adherent, and biocompatible materials in the design. The patch’s unique 3-bridge patch structure, embedded liquid metal circuitry, and medical skin-safe hydrogels further improved its adhesiveness and electrophysiological sensing, so that more stable and continuous signals could be collected and further processed.
RkJQdWJsaXNoZXIy NDk5Njg=