BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Neuroevolutionary approaches to generating increasingly intelligen
 t behaviours in virtual creatures / (simulated) robots - Alastair Channon\
 , University of Keele
DTSTART:20191206T140000Z
DTEND:20191206T150000Z
UID:TALK135568@talks.cam.ac.uk
CONTACT:Amanda Prorok
DESCRIPTION:Dr Channon and his Evolutionary Systems group carry out resear
 ch into the use of evolution to generate increasingly intelligent agents. 
 Their work has included evolving deep neurocontrollers (deep neuroevolutio
 n) since 1996\, using neural development (augmenting topologies) and gener
 ative encodings within that and other work also since 1996\, and incorpora
 ting (initially static\, hand-designed and later evolvable) convolutional 
 neural networks with rectified linear units (ReLUs) since 2005. This talk 
 will give an overview of some incremental approaches to the evolution of n
 eural networks for the control of intelligent agents / virtual creatures /
  simulated robots. The focus here will be on 'long-term' evolution and app
 roaches that generate increasingly intelligent behaviours. This part of th
 e talk will conclude with a very brief overview of related work on open-en
 ded evolution\, in which the aim is to achieve ongoing adaptive novelty an
 d ongoing growth of complexity.\n\nThe second part of the talk will focus 
 on two specific examples of the incremental neuroevolution of virtual crea
 tures / simulated robots.  In the first example\, a simple simulated quadr
 uped is evolved to walk over obstacles (walls) of different heights. A ran
 ge of different evolutionary complexification strategies are compared and 
 'heterogeneous strategies' are shown to overcome previous approaches' shor
 tcomings in relation to loss-of-gradient and over-fitting (analogous to ca
 tastrophic forgetting in neural systems).  The second example demonstrates
  the incremental neuroevolution of reactive and deliberative 3D Agents.  T
 he 3D physically-based setting requires that a successful agent continuall
 y and deliberately adjust its gait\, turning and other motor control over 
 the many stages and sub-stages of these tasks\, within its individual eval
 uation.  Achieving such complex interplay between motor control and delibe
 rative control\, within a neuroevolutionary framework\, is the focus of th
 is work. The results demonstrate a variety of intricate lifelike behaviour
 s being used\, separately and in combination.
LOCATION:James Dyson Building\, Teaching Room\, main Engineering site
END:VEVENT
END:VCALENDAR