Page 12 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

xxiv Introduction

insights into key principles of intelligence in biological brains and helps in building
more powerful artificially intelligent devices.
In Chapter 11, Morabito et al. introduce a comprehensive investigation of DL ap-
plications in brain engineering and biomedical signal processing, with a particular
focus on the processing of multivariate time-series coming from electrophysiology.
Electroencephalography (EEG) and high density EEG and magnetoencephalography
technologies are reviewed, as they constitute the measurement systems yielding
multivariate electrophysiological time-series. The use of DL methods for multivariate
EEG time-series processing are then detailed to permit the reader to easily enter this
fascinating application framework. Future direction of research in DL, encompassing
interpretability, architectures and learning procedures, and robustness aspects are then
discussed, so as to provide the reader with some relevant open research topics.
In Chapter 12, Alippi et al. present a timely and thorough review on the use of
computational intelligence and machine learning methods in cyber-physical systems
(CPS) and IoT. The review goes over four major research topics in this domain: (1)
system architecture; (2) energy harvesting, conservation, and management; (3) fault
detection and mitigation; and (4) cyberattack detection and countermeasures.
Importantly, the authors challenge assumptions that are taken for granted but do
not apply anymore to the increasingly more complex CPS and IoT systems. These
assumptions include high energy availability, stationarity, correct data availability,
and security guarantees. This chapter provides an excellent review of the status of
CPS and IoT and how to overcome the issues in these emerging fields.
In Chapter 13, Tagliaferri et al. present an innovative approach of multiview
learning as a branch of machine learning for the analysis of multimodal data in
biomedical applications, among which, in particular, the authors focus on bioinfor-
matics (i.e., gene expression, microRNA expression, proteineprotein interactions,
genome-wide association). The approach proposed allows capturing information
regarding different aspects of the underlying principles governing complex biolog-
ical systems. The authors also propose an example of how both clustering and
classification can be combined in a multiview setting for the automated diagnosis
of neurodegenerative disorders. They also show using some application examples
how recent DL techniques can be applied to multimodal data to learn complex
representations.
In Chapter 14, Choe provides educated discussions and analyses about some key
concepts central to brain science and AI, namely those associated with the dichot-
omies: meaning vs. information, prediction vs. memory, and question vs. answer.
The author shows how a slightly different view on these concepts can help us
move forward, beyond current limits of our understanding in these fields. In detail,
the chapter elaborates over the intriguing definition of information as seen from
different perspectives and its relationship with the concept of meaning. Then, it inves-
tigates the role of plasticity, that of memory, and how they relate to prediction.
Finally, the focus moves on to the question vs. answer issue in AI algorithms and
how it impacts on their ability to solve problems.