AI for Science Interview 2: Professor Roarke Horstmeyer | Duke University

Introduction

In this episode of AI for Science, we had the pleasure of speaking with Professor Roarke Horstmeyer, an assistant professor in the Biomedical Engineering department at Duke University. With an impressive academic background that includes a Ph.D. from Caltech, an M.S. from MIT's Media Lab, and a B.S. in Physics from Duke, Professor Horstmeyer has made significant contributions to the intersection of machine learning (ML) and optics.

Collaboration and Vision

Professor Horstmeyer is currently collaborating with alumni from MIT's Julia Lab on an EdTech startup called Visua AI Labs. Their mission is to create educational content for scientific machine learning and to support research projects in the field. As part of this initiative, Professor Horstmeyer intends to interview prominent scientists, increasing awareness and understanding of the connections between physics and machine learning.

Career Journey and Research Interests

Roarke grew up in Palo Alto, California, and has always had a keen interest in science and math. While pursuing his undergraduate studies at Duke University, he engaged in research projects focusing primarily on scientific instrumentation, leading to a fascination with how devices function to measure complex phenomena, particularly in imaging technology.

In graduate school, his curiosity about machine learning blossomed, particularly with developments in neural networks, especially convolutional neural networks (CNNs). Although neural networks were less popular during his early research days, Roarke recognized their potential thanks to a course he took at Caltech and colleagues at the Media Lab. As interest in neural networks surged, he began applying them to his own research, specifically in microscopy, where he was inspired to merge physical imaging principles with computational approaches.

Physical Layers and Optical Neural Networks

An exciting aspect of Professor Horstmeyer's research is the implementation of physical layers in neural networks, which allows researchers to optimize imaging systems. This innovative approach not only enhances the classification of images in real-time (like identifying diseases from microscope images) but also adapts how these images are acquired. Through this integration, he is tailoring neural networks to improve optical systems rather than merely classifying standard datasets, such as images of cats and dogs.

Professor Horstmeyer's work also illustrates the challenges faced when trying to adopt new software tools. It requires constant recalibration to ensure students are comfortable and competent with coding, especially when tools evolve rapidly. Throughout his teaching career, he has advocated for project-based learning, where students can apply what they've learned, focusing on engineering principles rather than rote memorization of coding functions.

Advancements in Education

Developing his course "Machine Learning and Optics," Professor Horstmeyer aims to bridge the gap between optics and machine learning by encouraging students to engage with real-world problems. His emphasis on understanding both the mechanics of coding and the underlying theoretical concepts equips students to tackle complex issues in tech and science.

Throughout his academic journey, Roarke maintains that understanding the fundamentals is essential, and he encourages those interested in entering the data science and ML fields to adopt a practical understanding coupled with theoretical insights.

Future Collaborations

Professor Horstmeyer is actively collaborating on projects that harness the power of open-source software like JAX. He shared an exciting opportunity to work on an open-source package called Chromatics, which supports complex wave operations and modeling different elements of microscopes.

In conclusion, Professor Roarke Horstmeyer’s blend of optics expertise, passion for machine learning, and commitment to education stands as a testament to the significant role AI can play in scientific advancements.

Keywords

• Roarke Horstmeyer
• Duke University
• Machine Learning
• Optical Imaging
• Research Collaboration
• Education Technology
• Convolutional Neural Networks (CNN)
• Chromatics
• Physics
• Scientific Instrumentation

FAQ

Q1: Who is Roarke Horstmeyer?
A1: Roarke Horstmeyer is an assistant professor in the Biomedical Engineering department at Duke University, with expertise in optical imaging and machine learning.

Q2: What is the focus of Visua AI Labs?
A2: Visua AI Labs focuses on creating educational content for scientific machine learning and supporting research projects in the field.

Q3: How did Roarke get interested in machine learning?
A3: His interest in machine learning was sparked during graduate school, particularly through a course on neural networks at Caltech.

Q4: What is the significance of physical layers in Roarke's research?
A4: Physical layers allow for the joint optimization of imaging systems in his research, which enhances the efficiency of capturing and analyzing images, particularly in medical diagnostics.

Q5: What is the Chromatics project?
A5: Chromatics is an open-source project that supports complex wave operations and the modeling of different microscope elements, aimed at improving optical imaging techniques.