Title: Recent Advances in Dynamic Processes over Networks
Abstract: Many interesting phenomena happening around us can be described as dynamic processes over networks. A dynamic process on a network refers to any process whose state, represented by variables defined on network nodes, changes over time due to interactions among the nodes according to certain rules. One prominent example is spread of infectious diseases though networks of human contact, whose study dates back to almost one century ago, but draws close attention nonetheless these days in the face of COVID-19. News or gossips are similarly spread through social interactions, increasingly over the social platforms. Such information dynamics can dramatically influence individuals’ perception and shape their behavior, the understanding of which has become increasingly important for modern politics and commercial marketing. On the technological side, how to disseminate information quickly and efficiently over large-scale complex networks, and how to prevent the propagation of cascading failures over interdependent networks, are two important questions that network scientists and engineers endeavor to address. This talk will give an overview of these interesting problems, illustrate the common models and analytical tools used for their studies, and discuss recent advances in this area.
Title: Coming soon...
Abstract: Coming soon...
Title: From linear to deep collaborative learning with applications to multi-modal fMRI and genomics data integration
Abstract: Canonical correlation analysis (CCA) has been used to find correlations between two or multiple data modalities. However, it is unrelated to phenotypes or disease status. On the other hand, regression models can find the association between a phenotype and multi-modal imaging and genomics data but overlook the cross-modal data correlation. To this end, we first propose a collaborative regression to combine both regression and CCA models. Then, we extend it to the deep learning framework by introducing deep collaborative learning (DCL), which includes deep CCA as a special example. As a result, DCL can better combine complex correlations between multiple data sets in addition to their fitting to phenotypes. Finally, we demonstrate its application to brain development study using integrative fMRI and genomics analysis. We show that DCL outperforms several existing models in predicting populations with different ages and intelligence quotients (IQ).
Title: Advanced Control of Surgical Robotics Systems
Abstract: To improve the quality of life, utilization of commercial industrial robots have been successfully adopted and further developed in precise automation processes for a variety of applications, especially robotic assisted systems in surgery operation. In general, a robotic surgery implementation is at the need of a well-developed control technology and precision of the surgical tools. Hence, from both theoretical and practical aspects the problem of surgical robotics control is a challenging topic.
The objective of this talk is to present some challenges and recent results on intelligent control of redundant surgical robot manipulators. Specifically, learning algorithms and optimization techniques under remote center of motion constraints are presented and then followed by model development for the surgical operation. The talk will be concluded with some concluding remarks on both technical and practical aspects of adaptive control schemes for surgical robotic systems.
Title: Effective Decoupling Techniques in Improving MIMO Antenna Isolation Characteristic
Abstract: Multiple-input multiple-output (MIMO) antennas have been a mainstream technology in fifth generation (5G) communications. However, strong mutual coupling between both adjacent and nonadjacent elements is inevitable, especially among highly integrated devices. In this presentation, three effective decoupling techniques are used and combined to improve MIMO antenna isolation. A slit embedded mushroom electromagnetic bandgap structure (EBG) is primarily proposed to suppress the propagation of surface wave between antenna elements. Subsequently, complementary split-ring resonators (CSRRs) are applied to steer the surface wave. By effectively utilizing the unusual electromagnetic property of EBG and CSRR to manipulate the propagation of surface wave, the mutual coupling between antenna elements has immensely alleviated. Finally, an H-shape defected ground structure (DGS) is introduced to reinforce decoupling effect. In order to validate the feasibility of the design principle, a prototype of the proposed antenna has been fabricated and measured. Measured results demonstrate that the decoupling concept presented is reasonable and approximately 12 dB reduction of mutual coupling is realized.
Title: Experimental study on fiber coupling control of adaptive optics wavefront correction
Abstract: In order to improve the coupling efficiency of free space coherent optical communication, the influence of the discrete distorted wavefront on the coupling efficiency is derived, and the influence of the distorted wavefront on the coupling efficiency and the power in the bucket is analyzed, and the coupling efficiency of the measured wavefront phase at different distances is analyzed through experiments. The theoretical analysis results show that: the coupling efficiency decreases with the increase of wavefront distortion, and the power in the barrel decreases with the increase of turbulence intensity; the experimental results show that after eliminating the non-common optical path aberration, the spot at the focal plane will diverge with the increase of turbulence intensity. After the optical wavefront is transmitted through indoor, 600m, 1km, 5km, 10km, 100km, the coupling efficiency is respectively From 10%, 10%, 10%, 3%, 5%, 5% to 90%, 80%, 90%, 60%, 40% and 10%. The effect of short-range correction is better than that of long-distance correction, and the fluctuation of coupling efficiency after short-range correction is smaller than that of long-distance correction.