Smart city sensing represents a pivotal paradigm shift in urban development, characterized by the infusion of cutting-edge technology to improve the quality of life in densely populated areas. At its essence, smart city sensing revolves around the deployment of sensors and data collection infrastructure throughout urban landscapes. The use of the collected data spans various domains, including environmental monitoring, transportation, energy management, and public safety. By continuously gathering and analyzing real-time data, smart cities gain unprecedented insights into their operations and the needs of their citizens. Unlocking the potential of smart cities, this talk presents effective solutions to leverage collected data to optimize the quality of the offered services. Specifically, this talk illuminates the innovative use of cellular data to analyze footfall and transportation modes, enabling precise localization and urban insights. By tracking citizens’ mobility profiles inferred from channel state information (CSI) measurements using efficient artificial intelligence (AI) tools, it reveals intricate pedestrian and transportation dynamics, informing optimal urban planning, crowd management, and resource allocation. Furthermore, the talk discusses effective solutions to enable the perpetual operation of IoT infrastructure and environmental monitoring sensors. Following an interdisciplinary methodology that combines experimental implementation and theoretical analysis, the use of 6G enablers such as unmanned aerial vehicles (UAVs) is investigated to optimize the data collection from the IoT sensors and minimize energy waste. 

Title: SWAP: SPARSE ENTROPIC WASSERSTEIN REGRESSION FOR ROBUST NETWORK PRUNING

Abstract: This talk addresses the challenge of encountering inaccurate gradients during neural network pruning. We introduce our recently proposed method, SWAP, a formulation of Entropic Wasserstein regression (EWR) for pruning, capitalizing on the geometric properties of the optimal transport problem. The “swap” of the commonly used sparse linear regression with the EWR in optimization is analytically demonstrated to offer noise mitigation effects by incorporating neighborhood interpolation across data points with only marginal additional computational cost. The unique strength of SWAP is its intrinsic ability to balance noise reduction and covariance information preservation effectively. Extensive experiments performed on various networks and datasets show comparable performance of SWAP with state-of-the-art (SoTA) network pruning algorithms. Our proposed method outperforms the SoTA when the network size or the target sparsity is large, the gain is even larger with the existence of noisy gradients, possibly from noisy data, analog memory, or adversarial attacks. 

Photonic startup business success relies heavily on the uniqueness and quality of its technical know-how. Success also depends heavily on a range of skills and activities beyond technical prowess. Business aspects such as marketing, sales, team leadership, commercial opportunity development, proposal preparation, program management, quality control, budgeting, funding and financial planning among others, are crucially important. This presentation will focus on dissecting all the skills and activities required for photonic startup business success. The presenter will attempt to do this through examples drawn from his own experience as a technical and business practitioner in large and small photonic companies.