IEEE Computational Intelligence Magazine - August 2022 - 8

Future AI should be able to not only sift and extract
valuable information (Learn) from complex and changing
environments (Perceive) but also create new meanings
(Abstract) and have the ability to assist humans in
planning and deciding (Reasoning), while meeting human
needs (Integration) and concerns (Ethics and Security).
unexpected disturbances or even subsystem
failures suddenly occurring during
system operation, we proposed to
use the NN to reconstruct the upper
norm bound of the function, rather
than the function itself. Such treatment
not only simplifies the overall computations,
but also allows the discontinuity
possibly involved in the function to be
handled gracefully. With inspiration
from nervous systems that involve
abundant interconnected cells with a
variety of complex mechanisms dealing
with incoming signals in many different
ways, we proposed to use diversified
basis (activation) functions in constructing
NN-based approximator. This is
deemed useful and sensible because the
activation function with fixed single
structure, chosen from the commonly
used sigmoidal function, hyperbolic
tangent function, or raised cosine function,
might not be a wise choice as
each neural structure presents specific
characteristics. Thus, for some complex
systems undergoing a myriad of changes,
monotonous structure is not able to
meet the learning requirements. To
address
the tough question of " how
large is sufficiently large " for the number
of neurons in an NN unit in control
design, we proposed a new
structural NN unit with grouped neurons
and self-adjusting number of subneurons,
i.e., the unit starts with a
finite number of neurons in each
group and automatically adds or deletes
certain number of neurons if needed,
using a strategy developed by the
tracking error related criteria, as motivated
by the fact that the neurons in
the neurological systems are updating
themselves constantly.
My areas of research have been
closely related to neural networks, control
and the related learning systems.
Being appointed as the Editor-in-Chief
(EiC) for TNNLS is absolutely an honor
and a huge responsibility, requiring dedication,
commitment, enthusiasm and
devotion. This position also demands
strong leadership experience and management
skill. I am energetic and enthusiastic
in IEEE related services, and I
consider myself articulate in communicating
and interacting with the community,
strongly
self-motivated with
professional dedication.
IEEE TNNLS is an internationally
renowned journal, under the guidance
and leadership of the President and VP
for publication at CIS, I will work closely
and diligently with the editorial team
to maintain TNNLS at its top level and
advance the journal to an even higher
level of excellence.
3. Where will your research
take you from here?
In collaborating with Jennie Si from
Arizona State University, USA, and
Sonya Coleman and Dermot Kerr from
Ulster University, UK, we have recently
put together a special issue for TNNLS
on biologically inspired methodologies
for sensing, control and decision making,
which is actually the current focus of
research in my team. We are particularly
interested in innovations and technologies
that allow engineered systems to
achieve desired performance and be
resilient to externally or internally
caused errors and unpredictable failures.
In essence, this requires engineered
designs capable of learning and selfreconfiguring
and awareness of themselves
and the environment it operates.
Therefore, reliable sensing and decisionmaking
are needed in engineering systems.
Traditional engineering approaches
8 IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | AUGUST 2022
have taken into account these design
considerations, but usually, the solutions
are costly. Yet, biological organisms in
nature have successfully demonstrated
their superior capability of processing a
large amount of information, dealing
with uncertainties when perceiving and
processing data of their surroundings,
adapting to environmental changes, and
recovering from their internal errors and
failures. All of these essential attributes
are desired by engineered systems.
Therefore, it is expected that biologically
learned and inspired methods may
offer fundamentally new theoretical
frameworks and new design approaches
to address system robustness, reliability,
and effectiveness. Some progress has
been made in bio-inspired control
methodologies and our next step of
research is to further verify and demonstrate
the benefits and efficiency of the
developed algorithms and strategies.
4. AI has accomplished
remarkable achievements and
showed its benefits to human life;
on the other hand, people are
concerned about its potential
risks. Can you share your
thoughts about the development
and future of AI?
Since the victory of AlphaGo over Lee
Sedol and Ke Jie, it has been genuinely
recognized that AI can surpass humans
in specific areas. In fact, AI has made significant
progress in some data concentrated
fields (transportation, medical,
finance, etc.); however, the high dependence
of current AI algorithms on data
literally limits its cross-field applicability.
In the foreseeable future, AI will still
mainly play the role of assisting humans
but not replacing them, and the development
of AI will be a long and tortuous
process. I believe that the AI of the
future should be able to not only sift
and extract valuable information (Learn)
from complex and changing environments
(Perceive) but also create new
meanings (Abstract) and have the ability
to assist humans in planning and deciding
(Reasoning), while meeting human
needs (Integration) and concerns (Ethics
and Security). I have expectations that
IEEE Computational Intelligence Magazine - August 2022

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - August 2022
