PhD Candidate
Los Angeles, USA
lzhuang0324@ucla.edu
+(1)424-402-2604
I am a Machine Learning Researcher at TikTok. I received my Ph.D. in Oct 2024 at UCLA, Master of Science in June 2021 at UCLA, and Bachelor of Engineering (BEng) degree in June 2019 at Zhejiang University, China.
I am currently working on content understanding and recommendation systems. My research interests also include computational imaging and microscopy. My goal is to apply advanced machine and deep learning technologies to microscopy imaging, and also enhance the robustness, generalizability and interpretability of models.
My oral presentation "Self-supervised, physics-informed learning for hologram reconstruction" has been selected as the AI/ML Best Paper in SPIE Photonics West 2024.
Our roadmap paper on label-free super-resolution imaging is published on Laser & Photonics Reviews.
Our research paper on self-supervised learning (GedankenNet) is published on Nature Machine Intelligence.
It is my great honor to receive the Dissertation Year Fellowship from UCLA! Thanks for the support from Prof. Aydogan Ozcan and Prof. Liang Gao.
Existing applications of deep learning in computational imaging and microscopy mostly depend on supervised learning, requiring large-scale, diverse and labelled training data. Here we report a self-supervised learning model, termed GedankenNet, that eliminates the need for labelled or experimental training data, and demonstrate its effectiveness and superior generalization on hologram reconstruction tasks. Without prior knowledge about the sample types, the self-supervised learning model was trained using a physics-consistency loss and artificial random images, and successfully generalized to experimental holograms of unseen biological samples after its self-supervised training. This self-supervised learning of image reconstruction creates new opportunities for solving inverse problems in holography, microscopy and computational imaging.
Deep learning-based image reconstruction methods have achieved remarkable success in phase recovery and holographic imaging, but their generalization to new types of samples remains a challenge. Here we introduce a deep learning framework, termed Fourier Imager Network (FIN) based on spatial Fourier transform modules that process the spatial frequencies of its inputs using learnable filters and a global receptive field. Compared with existing convolutional deep neural networks used for hologram reconstruction, FIN exhibits superior generalization to new types of samples, while also being much faster in its image inference speed. Beyond holographic microscopy and quantitative phase imaging, FIN and the underlying neural network architecture might open up various new opportunities to design broadly generalizable deep learning models in computational imaging and machine vision fields.
Volumetric imaging of samples using fluorescence microscopy plays an important role in various fields including physical, medical and life sciences. In this work we report a deep learning-based volumetric image inference framework that uses 2D images that are sparsely captured by a standard wide-field fluorescence microscope at arbitrary axial positions within the sample volume. Our work demonstrates the first application of recurrent neural networks in microscopic image reconstruction and provides a flexible and rapid volumetric imaging framework.
In OCT angiography (OCTA), decorrelation computation has been widely used as a local motion index to identify dynamic flow from static tissues, but its dependence on SNR severely degrades the vascular visibility, particularly in low-SNR regions. We developed a multi-variate time series (MVTS) model, and based on this model, derived a universal asymptotic linear relation of decorrelation to inverse SNR (iSNR) and a linear classifier, termed as ID-OCTA.