[Main contributions] – The primary contributions of this work are (a) extending the SMC-EM technique to include the DH-PSF for three dimensional motion, (b) demonstrating the ability to do accurate localization and parameter estimation of confined motion at low signal and low SBR levels, and (c) providing a quantitative comparison to the standard method of GF-MSD and to other approaches based on optimal estimation across a range of (very low) SBR levels and confinement lengths [More …].
My main work during the internship: explored and implemented machine learning techniques to analyze financial information and predict economic conditions. Linked the Risk, Finance, and Treasury models together and visualized the dependency relationship among models. Conducted quantitative data analysis and model simulation work based on model execution framework (MEF) to do balance sheet forecasting and stress testing across Barclays Group.
I'm grateful for the continuous help and support provided by Barclays colleagues who motivate me to keep learning and making progress!
It's a great honor to be a speaker at the Grace Hopper Celebration this year [More ...]. Meanwhile, I'm excited to be awarded the Grace Hopper Celebration Student Scholarship by Anitab.org. Dare to dream, and dare to make a difference!
[Main contributions] In this work, we describe three modifications to SMC-EM aimed at improving its computationally efficiency and demonstrate it through analysis of simulated SPT data of a particle in a three dimensional confined environment. The first two modifications use approximation methods to reduce the complexity of the original motion and measurement models without significant loss of accuracy. The third modification replaces the previous SMC methods with a Gaussian particle filter combined with a backward simulation particle smoother, trading off some level of generality for improved computational performance. In addition, we take advantage of the improved efficiency to investigate the effect of data length on performance in localization and parameter estimation [More … ]
The above figures show the mathematical simplification on confined motion model and measurement model (If you use the figures, please cite the above paper).
[Main contributions] - There are two primary contributions of this work. The first is the extension of our existing algorithms to data captured using an sCMOS camera. Due to their relatively low cost, high speed, and performance, sCMOS cameras are becoming popular tools for SPT data acquisition and including them in our EM-based approach extends the impact our algorithms can have. The second is the detailed, quantitative comparison of our EM based methods to a standard in the field, namely GF-MSD, and to an existing alternative that is also based on optimal estimation theory and which has previously been shown to outperform the standard approach in the analysis of diffusion, namely GF-MLE. This comparison is done across a wide range of SBRs and across a wide range of diffusion coefficients, validating the performance of our methods and guiding users in algorithm selection based on their particular experimental conditions [More ...].
The following video (DOI: 10.5061/dryad.9w0vt4bf5) shows the relationship among trajectory driven by Brownian motion, observation, and properties of readout noise brought by sCMOS.
Simultaneous Localization and Parameter Estimation for Single Particle Tracking in Two Dimensional Confined Environments [poster]
I'm so excited to attend the Grace Hopper Celebration for the first time in my life! This event provides a great platform for women in tech to get a lot of innovation and inspiration. Also, it's a great honor to present one of my research work in the GHC poster session [Poster]. Again, thanks for SE/CISE awarding me academic access to vGHC2020. Cheers!
A Time-Varying Approach to Single Particle Tracking with a Nonlinear Observation Model
[Abstract] - Single Particle Tracking (SPT) is a powerful class of tools for analyzing the dynamics of individual biological macromolecules moving inside living cells. The acquired data is typically in the form of a sequence of camera images that are then post-processed to reveal details about the motion. In this work, we develop a local time-varying estimation algorithm for estimating motion model parameters from the data considering nonlinear observations. Our approach uses several well-known existing tools, namely the Expectation Maximization (EM) algorithm combined with an Unscented Kalman filter (UKF) and an Unscented Rauch-Tung-Striebel smoother (URTSS), and applies them to the time-varying case through a sliding window methodology. Due to the shot noise characteristics of the photon generation process, this model uses a Poisson distribution to capture the measurement noise inherent in imaging. In order to apply our time-varying approach to the UKF, we first need to transform the measurements into a model with additive Gaussian noise. This is carried out using a variance stabilizing transform. Results from simulations show that our approach is successful in tracing time-varying diffusion constants at a range of physically relevant signal levels. We also discuss the initialization for the EM algorithm based on the available data.
Biophysical Society (BPS) 64th Annual Meeting (February 15 - February 19)
[Reference]
@article{lin2020quantitative,
title={Quantitative Comparison of Single Particle Tracking Algorithms Across Different Signal and Noise Levels},
author={Lin, Ye and Andersson, Sean B},
journal={Biophysical Journal},
volume={118},
number={3},
pages={616a},
year={2020},
publisher={Elsevier}
}
[Brief illustration] - Single Particle Tracking (SPT) is a powerful class of tools for acquiring information about the dynamics of biological macromolecules moving inside living cells. In this work, we are interested in comparing algorithms with respect to the performance of localization and motion model estimation at very low signal levels. A standard approach for estimation is to use Gaussian Fitting (GF) to determine the position of an emitter in an image, followed by linking to produce a trajectory. These trajectories are then analyzed by using Mean Square Displacement (MSD) or Maximum Likelihood (ML) to estimate motion parameters, such as diffusion coefficients. However, both MSD and the standard formulation of ML assume a linear measurement model with simple additive Gaussian noise, which begins to fail as image intensity decreases. We have introduced two nonlinear methods that more accurately model camera-based measurements: Sequential Monte Carlo combined with Expectation Maximization (SMC-EM), and Unscented Kalman Filter combined with Expectation Maximization (U-EM).
Minor in English Literature, Beihang University.
[Abstract] - Single Particle Tracking (SPT) is a powerful class of tools for analyzing the dynamics of individual biological macromolecules moving inside living cells. The acquired data is typically in the form of a sequence of camera images that are then post-processed to reveal details about the motion. In this work, we develop an algorithm for jointly estimating both particle trajectory and motion model parameters from the data. Our approach uses Expectation-Maximization (EM) combined with an Unscented Kalman filter (UKF) and an Unscented Rauch-Tung-Striebel smoother (URTSS), allowing us to use an accurate, nonlinear model of the observations acquired by the camera. Due to the shot noise characteristics of the photon generation process, this model uses a Poisson distribution to capture the measurement noise inherent in imaging. In order to apply a UKF, we first must transform the measurements into a model with additive Gaussian noise. We consider two approaches, one based on variance stabilizing transformations (where we compare the Anscombe and Freeman-Tukey transforms) and one on a Gaussian approximation to the Poisson distribution. Through simulations, we demonstrate the efficacy of the approach and explore the differences among these measurement transformations.
[Example of measurement] -
[Reference] -
@inproceedings{lin2019simultaneous,
title={Simultaneous Localization and Parameter Estimation for Single Particle Tracking via Sigma Points based EM},
author={Lin, Ye and Andersson, Sean B},
booktitle={2019 IEEE 58th Conference on Decision and Control (CDC)},
pages={6467--6472},
year={2019},
organization={IEEE}
}
In the Fall semester of 2019 & 2020, I was the teaching assistant of the class ME/SE/EC 501 Dynamic Systems Theory taught by Professor John Baillieul and Professor Theodore Djaferis. Course Website Available Here.