top of page

Topic: Using statistical learning and the Physics of image formation to evolve exploratory microscopy

Speaker: Dr. Duane Loh from National University of Singapore



We’re super delighted to host Dr. Duane Loh from National University of Singapore to share with us:


Using statistical learning and the Physics of image formation to evolve exploratory microscopy


Duane Loh has been a thought leader in the AI in science field in Singapore! his talk is long-awaited! He’s also an awesome speaker!


Abstract

The evolution of human vision is a fascinating interaction between our eyes and brains. At the very beginning, so the story goes, variations of primitive eyes receive input that is interpreted by simple yet plastic neural networks that in turn trigger responses. When these responses confer the organism with evolutionary fitness, the causal “wetware” (biological equivalent of hardware) and “software” are selectively propagated to the progeny. Iterating such selective propagations over many generations, purportedly, created the powerful visual perception that you are using to read this abstract.

A different type of interaction between hardware and software also occurs in the iterative evolution of microscopy. Humankind has a storied history of inventing many novel forms of microscopy using visible light, x-rays, and electrons. Each invention, however wild or incremental, produced images that fed the development of Physical theories of image formation, which in turn empowered the design of the next generations of microscopy.


Generations of such iterative design later, it is often impossible for humans to interpret the raw micrographs without using the knowledge from the theory of image formation. This is especially true when the imaging setup only permits us a very noisy and incomplete view of the sample. In these scenarios, the combination of statistical machine learning with the theory of image formation is critical.


In this talk, I will describe how my group has been combining the Physics of image formation with statistical learning to create statistical lenses for high-resolution microscopy. These combinations, we hope, will create vistas to explore novel, complex phenomena in the fragile, fleeting, and chaotic nanometer world.


Biography

Duane Loh is a faculty member in the departments of Physics and Biological Sciences at NUS. His is interested in combining machine learning with scientific and instrument priors to create computational lenses that help make sense of the chaotic and nearly invisible dynamics that occur at the nanometer-scale. Duane did his Bachelor of Science at Harvey Mudd College, specializing in Theoretical and Mathematical Physics. He would go on to earn his PhD in Physics from Cornell University for his feasibility studies of single-particle diffractive imaging with extremely noisy, and incomplete data. Thereafter, he moved to SLAC National Accelerator Laboratory at Stanford University, the world’s first hard X-ray Free-electron laser, to help realize these single-particle diffractive imaging experiments. At SLAC, Duane pioneered unsupervised machine learning to discover transient intermediate states in highly heterogeneous and dynamic phenomena. In 2013, he joined NUS as a Lee Kuan Yew post-doctorate fellow, where he extended these methods to electron microscopy to study dynamic order-disorder phenomena at the nanometer-scale. He started his research group NUS in 2016, and continues to develop core ideas and technologies towards far-reaching imaging modalities that are too challenging for hardware-based microscopy alone.

Very much looking forward to the seminar – hope to see you all on Wednesday: (SGT) 9am, Wed, 28 Apr! https://mit.zoom.us/j/96231985116

Warmest regards, @Siyu Tian (Isaac) @kedar @Tonio

Topic: An overview on material informatics including global trends, outlook, key players etc.

Speakers: Ms. Cheng Meiling (Customer Success Manager, Lux Research) and Mohammed Ali (Managing Director, Lux Research APAC)



We have a special talk lined up this week! The team at Lux Research, led by Customer Success Manager, Ms. Cheng Meiling and Mohammed Ali (Managing Director, Lux Research APAC) will share an overview on material informatics including global trends, outlook, key players etc.

Very much looking forward to the seminar – hope to see you all on Wednesday: (SGT) 9am, Wed, 14 Apr! https://luxresearchinc.zoom.us/j/92698965876?pwd=TGE1aktJQkp2VUcwQ2dHNVlGd0VZQT09

Warmest regards, @Siyu Tian (Isaac) @kedar @Tonio

Topic: Data-driven Approaches to Accelerated Discovery of Complex and Metastable Materials

Speaker: Dr. Apurva Mehta from SLAC National Accelerator Laboratory



With great joy, we host Dr. Apurva Mehta from SLAC National Accelerator Laboratory this week to outline a combination of machine-learning and AI-directed (high throughput) experiments to accelerate discoveries of structural and functional multi-principal-element alloys.


Abstract

The complex material systems are beginning to play a growing role in addressing some of our current technological challenges. They include multi-principal element alloys and oxides (often also called high entropy alloys/oxides), metastable materials like metallic glasses and carbon nanotubes, solid-state energy storage devices, heterogeneous catalysts, and increasingly complex devices for power electronics and quantum computing. However, there are easily tens of million and possibly even a billon compositions out of which a very small fraction are potential candidates. Many of the desired materials are metastable and therefore processing paths play a crucial role in synthesizing them. The composition-processing combinatorial search space is too vast for a dependence on serendipitous discoveries, to provide urgently need novel materials and devices; and a brute-force survey of the combinatorial space is too slow and expensive.


We need guidance to navigate the vast composition-processing space. Our physiochemical understanding of complex material systems is still too nascent to provide it. Recent advances in machine-learning (ML) and other AI-tools, however, suggests that a data-driven approach that builds up on insights from physiochemical theories, and bridges the gaps in the insights by experimental observations can provide the needed navigation. Application of ML to large experimental observations can highlight hidden and complex trends that theories may miss. These predictions also come with uncertainty quantification. The combination of predictions and uncertainties allow us to device different experimental strategies, ranging from exploitive to exploratory to help direct us to the next high impact experiment to perform, leading to accelerated discoveries and new physiochemical insights. Here, I will demonstrate these approaches using examples from amorphous and crystalline Multi-Principal-Element Alloys with desired structural and thermoelectric properties.


Biography

As the lead scientist at SLAC, he has been leading the effort of high-throughput structural characterization of new inorganic materials at the beamline, with machine learning, computations and optimization in the loop.

Very much looking forward to the seminar – hope to see you all on Wednesday: (SGT) 9am, Wed, 31 Apr! https://mit.zoom.us/j/96231985116

Warmest regards, @Siyu Tian (Isaac) @kedar @Tonio

© 2024 by Kedar Hippalgaonkar. Created with Wix.com

Materials Science and Engineering, Nanyang Technological University

Institute of Materials Research and Engineering, Singapore

bottom of page