Engineering topics

Zhao, Tuo

Publications and source records attributed to Zhao, Tuo.

5 records

DOE OSTI2024

Deep Nonparametric Estimation of Operators between Infinite Dimensional Spaces

Learning operators between infinitely dimensional spaces is an important learning task arising in machine learning, imaging science, mathematical modeling and simulations, etc. This paper studies the nonparametric estimation of Lipschitz operators using deep neural networks. Non-asymptotic upper bounds are derived for the generalization error of the empirical risk minimizer over a properly chosen network class. Under the assumption that the target operator exhibits a low dimensional structure, our error bounds decay as the training sample size increases, with an attractive fast rate depending on the intrinsic dimension in our estimation. Our assumptions cover most scenarios in real applications and our results give rise to fast rates by exploiting low dimensional structures of data in operator estimation. We also investigate the influence of network structures (e.g., network width, depth, and sparsity) on the generalization error of the neural network estimator and propose a general suggestion on the choice of network structures to maximize the learning efficiency quantitatively.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

High dimensional binary classification under label shift: phase transition and regularization

Label Shift has been widely believed to be harmful to the generalization performance of machine learning models. Researchers have proposed many approaches to mitigate the impact of the label shift, e.g., balancing the training data. However, these methods often consider the underparametrized regime, where the sample size is much larger than the data dimension. The research under the overparametrized regime is very limited. Here, to bridge this gap, we propose a new asymptotic analysis of the Fisher Linear Discriminant classifier for binary classification with label shift. Specifically, we prove that there exists a phase transition phenomenon: Under certain overparametrized regime, the classifier trained using imbalanced data outperforms the counterpart with reduced balanced data. Moreover, we investigate the impact of regularization to the label shift: The aforementioned phase transition vanishes as the regularization becomes strong.

binary classification↗

DOE OSTI2021

A universal method to fabricating porous carbon for Li-O 2 battery

The main bottleneck in the application of biomass-based carbon material in energy field is the development of an economical, nanoscale controllable and universal fabrication method. Herein, we demonstrate that a biomass-derived carbon consisting of loose carbon nanosphere clusters could serve as an ideal cathode material for high performance Li-O 2 batteries. The carbon nanomaterial can be prepared from both biomass derivatives and crude biomass by a low cost, facile and nanoscale controllable method. The open slit-shaped hierarchical pore structure endows the electrode sufficient active surface sites and mass transport channels that are not easily asphyxiated by the discharge products. The concentration of oxygenic groups and structure defects on the surface of obtained carbon materials is very low, which minimizes the side reactions associated with carbon decomposition. Finally, the unique pore structure and surface properties of the carbon electrode give rise to a superhigh specific capacity of 20,300 mA h g -1 and an extremely long cycle life which a cyclic capacity retention of 100% is maintained for 543 cycles with current density of 0.2 mA cm -2 at controlled discharge-charge depths of 500 mA h g -1 .

25 ENERGY STORAGE↗

DOE OSTI2020

Optimization of Oxygen Electrode Combined with Soluble Catalyst to Enhance the Performance of Lithium-Oxygen Battery

Lithium-air batteries possess a high theoretical energy density among the current battery technologies, and are expected to become the driving power for electric vehicles. However, the practical application of lithium air batteries is handicapped by their poor cycle life, high over-charge potential and low energy efficiency. Numerous studies have performed to improve the kinetics of the electrochemical reaction. Making a good oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) catalyst system is under intense scrutiny. In this work, we report a rational design of air electrode, which derives from biomass and presents a three-dimensional porous carbon architecture with high specific surface and large volume expansion coefficient. Combining with a soluble catalyst, this oxygen electrode presents an ultrahigh areal specific capacity of 7.95 mAh/cm 2 and a long cycling life up to 1000 cycles with suppressed polarization (about 0.46 V). Finally, the greatly improved overall performance is due to the synergetic effect of the novel designed oxygen electrode capable of mediating the redox mediators to trigger the ORR and OER process.

25 ENERGY STORAGE↗

DOE OSTI2013

Natural CD8{sup +}25{sup +} regulatory T cell-secreted exosomes capable of suppressing cytotoxic T lymphocyte-mediated immunity against B16 melanoma

Highlights: •CD8{sup +}25{sup +} regulatory T cells secrete tolerogenic exosomes. •CD8{sup +}25{sup +} regulatory T cell-derived exosomes exhibit immunosuppressive effect. •CD8{sup +}25{sup +} regulatory T cell-derived exosomes inhibit antitumor immunity. -- Abstract: Natural CD4{sup +}25{sup +} and CD8{sup +}25{sup +} regulatory T (Tr) cells have been shown to inhibit autoimmune diseases. Immune cells secrete exosomes (EXOs), which are crucial for immune regulation. However, immunomodulatory effect of natural Tr cell-secreted EXOs is unknown. In this study, we purified natural CD8{sup +}25{sup +} Tr cells from C57BL/6 mouse naive CD8{sup +} T cells, and in vitro amplified them with CD3/CD28 beads. EXOs (EXO{sub Tr}) were purified from Tr cell’s culture supernatants by differential ultracentrifugation and analyzed by electron microscopy, Western blot and flow cytometry. Our data showed that EXO{sub Tr} had a “saucer” or round shape with 50–100 nm in diameter, contained EXO-associated markers LAMP-1 and CD9, and expressed natural Tr cell markers CD25 and GITR. To assess immunomodulatory effect, we i.v. immunized C57BL/6 mice with ovalbumin (OVA)-pulsed DCs (DC{sub OVA}) plus Tr cells or EXO{sub Tr}, and then assessed OVA-specific CD8{sup +} T cell responses using PE-H-2K{sup b}/OVA tetramer and FITC-anti-CD8 antibody staining by flow cytometry and antitumor immunity in immunized mice with challenge of OVA-expressing BL6–10{sub OVA} melanoma cells. We demonstrated that DC{sub OVA}-stimulated CD8{sup +} T cell responses and protective antitumor immunity significantly dropped from 2.52% to 1.08% and 1.81% (p < 0.05), and from 8/8 to 2/8 and 5/8 mice DC{sub OVA} (p < 0.05) in immunized mice with co-injection of Tr cells and EXO{sub Tr}, respectively. Our results indicate that natural CD8{sup +}25{sup +} Tr cell-released EXOs, alike CD8{sup +}25{sup +} Tr cells, can inhibit CD8{sup +} T cell responses and antitumor immunity. Therefore, EXOs derived from natural CD4{sup +}25{sup +} and CD8{sup +}25{sup +} Tr cells may become an alternative for immunotherapy of autoimmune diseases.

60 APPLIED LIFE SCIENCES↗