科技高速发展的时代，我们通常走得很快，却少了很多思考的时间。这个周末，我们想邀请人们暂时停下脚步，参与 FAIR 2020: Grand Challenges and Opportunities， 聆听人工智能与机器人领域的大师们分享他们的思想，并近距离与他们交流前沿技术，以激发我们每个人对这个领域的独立思考，抓住时代赋予我们的机遇，一起应对人类共同的挑战。
来吧，加入由深圳市人工智能与机器人研究院（AIRS）和香港中文大学（深圳）联合举办的全球人工智能与机器人前沿研讨会2020: Grand Challenges and Opportunities (Frontiers in AI & Robotics - FAIR 2020: Grand Challenges and Opportunities)。
Optimization and Operations Research in Mitigation of a Pandemic
We present several Optimization, Statistics and Operations Research models and methods in mitigation the ongoing Covid-19 pandemic. In particular, we describe in details of following topics:
● Inventory and Risk Pooling of Medical Equipment/Resources in a Pandemic
● New Norm: Operation/Optimization helps to maintain Social Distancing
● Indoor GPS and Tracking by Sensor Network Localization for Contact-Tracing
● Dynamic and Equitable Region Partitioning for Hospital/Health-Care Services
● Efficient Public Good Allocating under Tight Capacity Restriction via Market Equilibrium Mechanisms/Platforms
The Era of Human-Robot Collaboration
Robotics is undergoing a major transformation in scope and dimension with accelerating impact on the economy, production, and culture of our global society. The generations of robots now being developed will increasingly touch people and their lives. They will explore, work, and interact with humans in their homes, workplaces, in new production systems, and in challenging field domains. The emerging robots will provide increased support in mining, underwater, hostile environments, as well as in domestic, health, industry, and service applications. Combining the experience and cognitive abilities of the human with the strength, dependability, reach, and endurance of robots will fuel a wide range of new robotic applications. The discussion focuses on design concepts, control architectures, task primitives and strategies that bring human modeling and skill understanding to the development of this new generation of collaborative robots.
Benjamin Van Roy
Hypermodels for Exploration
We study the use of hypermodels to represent epistemic uncertainty and guide exploration. This generalizes and extends the use of ensembles to approximate Thompson sampling. The computational cost of training an ensemble grows with its size, and as such, prior work has typically been limited to ensembles with tens of elements. We show that alternative hypermodels can enjoy dramatic efficiency gains, enabling behavior that would otherwise require hundreds or thousands of elements, and even succeed in situations where ensemble methods fail to learn regardless of size. This allows more accurate approximation of Thompson sampling as well as use of more sophisticated exploration schemes. In particular, we consider an approximate form of information-directed sampling and demonstrate performance gains relative to Thompson sampling. As alternatives to ensembles, we consider linear and neural network hypermodels, also known as hypernetworks. We prove that, with neural network base models, a linear hypermodel can represent essentially any distribution over functions, and as such, hypernetworks are no more expressive.
Deep Learning for Algorithm Design
Algorithms are step-by-step instructions designed by human experts to solve a problem. Effective algorithms play central roles in modern computing, and have impacted many industrial applications, such as recommendation and advertisement in internet, resource allocation in cloud computing, robot and route planning, disease understanding and drug design.
However, designing effective algorithms is a time-consuming and difficult task. It often requires lots of intuition and expertise to tailor algorithmic choices in particular applications. Furthermore, when complex application data are involved, it becomes even more challenging for human experts to reason about algorithm behavior.
Can we use deep learning and AI to help algorithm design? There have been a number of recent advancements that have allowed algorithms to designed from specific algorithmic families automatically using data, often leading to either state-of-the-art empirical performance or provable performance guarantees on observed instance distributions. In this talk, I will provide an introduction to this area, and explain a few pieces of work along this direction.
From Deep Learning to Deep Understanding
Communication with Speech and Language – A Hallmark of Artificial Intelligence
The ability to communicate in speech and language has long been regarded as a hallmark of human intelligence. Recent technological advancements have made great strides in enabling machines to simulate the human ability to communicate verbally and create a hallmark of Artificial Intelligence (AI). This talk presents an overview of ongoing research at CUHK that enables AI to not only speak and listen, but also to enhance learning of a new language, to serve users with communicative impairments, as well as to combat dementia.
People have envisioned tiny machines and robots that can explore the human body, find and treat diseases since Richard Feynman’s famous speech, “There's plenty of room at the bottom,” in which the idea of a “swallowable surgeon” was proposed in the 1950s. Even though we are at a state of infancy to achieve this vision, recent intense progress on nanotechnology, MEMS/NEMS technology and micro-/nanorobotics has accelerated the pace toward the goal. A number of research efforts have been recently published regarding the development of tiny swimming machines/robots from the basic principles and fabrication methods to practical applications.
I will present the past and recent research progress on medical micro-/nanorobots. The challenges and opportunities of using these tiny agents for biomedical applications will be discussed.
Aggregating Intelligence with the Internet of Intelligent Things (IoIT)
Incentive Mechanism Design for Crowd Systems
Crowd systems can help solve complicated problems through the collective efforts of many non-expert agents. A key to success is to incentivize enough agents to participate and exert efforts. We will introduce the challenges and opportunities of incentive mechanism designs in diverse types of crowd systems.
我将介绍我们最近在Edge AI方面的研究。通过智能地分布和调度从云到物联网端的计算，存储，控制和网络资源，边缘智能计算技术可以应对下一代物联网的挑战。首先，我将介绍我们的基于实时边缘中间件(real-time Edge middleware)的智能路边设施RSI (smart roadside infrastructure)系统，通过对边缘系统进行编程并在网络层之间划分计算任务，我们的实时边缘中间件可以在满足应用程序时延要求的同时最大程度地降低系统功耗。在此框架上我们进行了智能多传感器融合和实时多深度学习任务调度等工作。最后我将简要介绍我们在移动健康、联邦学习、火山地震监测、NB-IoT等方向的工作。我们研发的系统已经进行了大规模的现场部署，包括在厄瓜多尔和智利的两个活火山上安装的地震传感器网络。
香港理工大学电子计算学系讲座教授、IEEE Fellow、ACM Distinguished Member
Distributed Intelligence at the Edge
The emerging IoT applications in connected healthcare, industrial internet, multi-robot systems, and other areas demand higher intelligence of the connected devices, larger scale of the systems, and better decision making leveraged by analyzing the data being continuously generated. In this context, centralized cloud computing would face high data transmission cost, high response time, and data privacy issues. The edge cloud paradigm seeks to alleviate these inefficiencies by moving the computation and analytics tasks closer to the end devices. It facilitates the evolution of IoT from instrumentation and interconnection to distributed intelligence. This talk focuses on collaborative edge computing where edge nodes share data and computation resources and perform tasks by leveraging distributed intelligence. It covers the major problems in distributed collaboration we are currently studying, namely collaborative task execution, distributed machine learning, and distributed cooperation in autonomous multi-robot systems. Solutions need to address the challenging issues such as distributed data sources, conflicting network flows, heterogeneous devices, consistency, and mutual influence during the training.
Medical Biometrics- A Computerized TCM Data Analysis Approach
Traditional Chinese Medicine (TCM) diagnosis methods are mainly relied on Doctor's experience and not quantified. In this presentation, we will try to develop a novel approach by using Medical Biometrics technology to solve these problems. By some TCM-orient diagnosis acquisition devices, we could collect many kinds of date like tongue/pulse/odor with a priori knowledge from healthy/sub-healthy in Body Checking Station or from different diseases in Hospitals. Then, we use a statistical pattern recognition method to extract all possible features from these images/waveforms, including color, texture, shape, and so on. After matching between our training data and testing data, some decision rules will be made. Finally, we apply our results to the practical diseases diagnosis to illustrate the effectiveness of our approach.
Embracing Mechanical Intelligence for Agile Locomotion
Understanding the locomotion principle behind animals is crucial in developing next generation of agile robotic platform. Over the past decades, a wide range of bio-inspired legged robots have been developed that can run, jump, and climb over a variety of challenging surfaces. However, in terms of maneuverability they still lag far behind animals. Animals have instinct to use their mechanical body and external appendages (such as tails) effectively to achieve spectacular maneuverability, energy efficient locomotion, and robust stabilization to large perturbations which may not be easily attained in the existing legged robots.
In this talk, we will present our efforts on the development of innovative legged robots with greater mobility/efficiency/robustness, comparable to its biological counterpart. We will discuss the fundamental challenges for legged robots and show our initial results to demonstrate the feasibility of developing such systems through the use of external appendages and advanced intelligent algorithms. We believe our solutions could potentially lead to more efficient legged robot design and give the legged robot greater mobility and robustness for moving through complex real-world environments, comparable to its biological counterpart.
|2020.07.19 09:40-10:10||Optimization and Operations Research in Mitigation of a Pandemic||叶荫宇教授|
|2020.07.19 10:10-10:40||The Era of Human-Robot Collaboration||Oussama Khatib教授|
|2020.07.19 10:40-11:10||Hypermodels for Exploration||Benjamin Ven Roy教授|
|2020.07.19 09:30-09:35||Deep Learning for Algorithm Design||宋乐教授|
|2020.07.19 14:10-14:40||From Deep Learning to Deep Understanding||沈向洋博士|
|2020.07.19 14:40-15:10||Communication with Speech and Language – A Hallmark of Artificial Intelligence||蒙美玲教授|
|2020.07.19 16:40-17:10||Aggregating Intelligence with the Internet of Intelligent Things (IoIT)||李世鹏博士|
|2020.07.19 17:10-17:40||Incentive Mechanism Design for Crowd Systems||黄建伟教授|
|2020.07.20 11:10-11:40||Distributed Intelligence at the Edge||曹建农教授|
|2020.07.20 11:40-12:10||Medical Biometrics- A Computerized TCM Data Analysis Approach||张大鹏教授|
|2020.07.20 12:10-12:40||Embracing Mechanical Intelligence for Agile Locomotion||欧国威教授|