About the Center

About the Center

UC San Diego researchers at the Center for Machine-Intelligence, Computing and Security are integrating hardware, software and massive data sets in new ways in order to invent the future of machine learning, real-time data analytics, deep learning, security and privacy. Advances in the integration of hardware, software, algorithms and data are necessary for developing new generations of systems that make decisions and take actions based on data that are collected and analyzed in real time. Our team is continually on the cutting edge of innovation. The team, for example, was the first to report real-time analysis of streaming data using machine learning algorithms running on mobile platforms and other resource-constrained platforms such as drones. 

Real-Time Data Analytics

Hardware, software and algorithm co-design for real-time data analytics. Our customized performance optimization engine is automated and works across platforms, from low-power sensors to data centers and the cloud. Our solutions integrate adaptive data collection processes with training, learning, and inference in real-time and streaming applications.

Paradigm Shift in Deep Learning

Automated acceleration and adaptive retraining of deep learning. Our framework allows for training of deep learning networks that are platform independent, and scale from sensors to mobile to data centers. We introduced a paradigm shift when we built and demonstrated the first training of deep learning on Edge devices.

Security & Privacy for Cyber-Physical Systems 

To secure cyber-physical systems, we fully consider hardware, software, algorithms and data - and their isolation and interactions. We offer new approaches to security and privacy. Safe machine learning / defense against adversarial attacks, secure embedded medical devices, and privacy-preserving computing (DNA, learning, biometrics) are examples.

Our work is crucial for developing scalable and secure machine intelligence for cloud computing, data centers and many other autonomous and semi-autonomous applications including surgical robots, imaging systems and low-power sensor networks.

One key technological hurdle that must be cleared to develop these kinds of systems is the ability to analyze – on the edges of networks – massive data sets coming in from multiple sources in real time. This will require advanced machine learning algorithms to be training in real time on mobile platforms and embedded systems that are constrained by power, computational resources and bandwidth.

These machine learning algorithms will also need to process incoming data in real time, and adapt their behavior accordingly. In this way, real-time data analytics on mobile and embedded computing platforms will guide real-time decision making, and real-world actions taken by autonomous systems.

Our researchers are integrating hardware, software and massive data sets in new ways in order to invent this future.