Next: Analysis: Real- and Complex-Valued
Up: No Title
Previous: No Title
The discipline of computer science seems to draw mostly upon constructs
and operations from discrete mathematics, such as the propositional calculus
(logic, syllogisms, truth tables),
set membership & relations, and combinatorics. The fundamental notion of an
algorithm is a discrete sequence of discrete operations. The
elementary hardware devices that implement algorithms are discrete gates,
governed by Boolean algebra,
and the elementary entities that they manipulate are bits, which are
discrete states.
Information is generated, transmitted, and stored in discrete form, and
everything that happens in computing
happens at discrete sequences of points in time - the edges of clock ticks.
So why study continuous mathematics?
- Answer 1:
- Because the natural world is continuous. It is the
discreteness of digital computing that is unnatural! If we want to model
a world that is governed by the laws of physics, we must come to
computational terms with continuous processes.
- Answer 2:
- Because the distinction between discrete and
continuous processes is illusory. Just as every continuous process can be
approximated by discrete ones, every discrete process can be modeled
as a continuous one.
- Answer 3:
- Because the two domains are inextricably
intertwined, mathematically or physically.
Semiconductor devices such as TTL logical gates really operate through
continuous quantities (voltage, current, conductance); continuous theoretical
constructs such as differentials and
derivatives are only defined in terms of limits of discrete quantities
(finite differences); etc.
- Answer 4:
- Because some of the most interesting and powerful
computers that we know about are continuous. Non-linear dynamical systems in
continuous time can be viewed as automata having great computational power;
and the most powerful known ``computer," the human brain, has the following
properties that distinguish it from a digital computer: it lacks numerical
calculations; its communications media are stochastic; its components are
unreliable and widely distributed; it has no precise connectivity blueprints;
and its clocking is asynchronous and extremely slow (milliseconds). Yet its
performance in real-time tasks involving perception, learning, and motor
control, is unrivaled. As computer scientists we need to be able to study
neural processes, and at many levels this requires continuous mathematics.
This short course is intended to be a refresher on some of the major ideas
and tools used in continuous mathematics. Its practical purpose
within the CST curriculum is to serve
as groundwork for the following
Pt. II and Diploma courses:
Information Theory and Coding; Neural
Computing; and Computer Vision.
Next: Analysis: Real- and Complex-Valued
Up: No Title
Previous: No Title
Neil Dodgson
2000-10-23