Signal Processing

◊	Statistical Analysis of HRR Signals

◊	Higher Order Statistics


◊	Tomographic Imaging

◊	Multiaspect Partial Reconstruction

Machine Learning

◊	Learning under Imperfections

◊	Multitask Learning

◊	Plan-in-Advance Active Learning

Planning and Scheduling

◊	Information-Theoretic Myopic Planning


◊	Partially Observable Markov Decision Process

Bioinformatics

Computer Vision

Signals are carriers of information and signal processing (SP) is a technique of reconstructing and extracting information by manipulation of signals. The ability to cognise and acquire information, in a large sense, constitutes intelligence, and it is not exaggerating to say that human knowledge accumulate through cognition, discovery, and exchange of information. While the nature of intelligence is still an unsolved mystery, artificial intelligence (AI), a science of studying and imitating intelligence by means of computation, has been fruitful through several decades' pursuit.

A good example illustrating how SP and AI integrate is a robot. The robot uses its sensors to receive signals from the environment and its "brain" to cognise the information contained in the signals. An inseparable part of each sensor is its signal processor, which transforms the raw signals into internal representations of information understandable by the "brain". While signal processing (SP) is a study concerned with designing and implementing the transformations of signals, the core mission of artificial intelligence (AI) is to build computing machines that function as the "brain".

A smart signal processor produces a neat internal representation that requires simple brain to understand. On the contrary, a naive signal processor produces a scruffy internal representation that requires a sophiscated brain to understand. What is the right balance between SP and AI? The fact is, SP techniques are usually built upon the underlying physics of signals, while AI is traditionally a black or gray box that learns the input-to-output mappings but cares little about the underlying processes. Therefore, if the underlying process governing the phenomenon is known, this information should be exploited and maybe this exploitation should be put in the SP category.

The boundary is becoming fuzzy, however. With the advance in biotechniques, we are now looking into the biological brain and approaching a better understanding of the mystery of intelligence. This has brought SP and AI even closer.