Scientific Setup

The main scientific inquiries concern:

* what are the existing and new biologial structures in the brain
* how do they translate into computational models
* how can those models be represented in computer form
* how can they be simulated, and how what new programming
models can be derived from it
* how can those models be applied
* can connections between nano/micro and macroscopic brain
activity be found, modeled and simulated, what is the
sensory behaviour
* what are the connections between nano-biological activities
involved in low level neuronal behaviour and quantum effects
* can quantum effects be observed also at higher levels of
abstraction for larger sets of combined neurons, and if
so what could be model-technical applications
* possibly: give some starting points for medical applications
of the acquired models
* can (nano-)biological structures be engineered?
* can (pieces of) nano-biological structures be used in other
fields, such as in more traditional computer technology?

Acquiring Biological and Neurological Knowledge

To get a good overview of current research programs and
personal interests of people in these fields, inventorization
is needed.
Contentwise, sufficient information can be found through library
references to acquire a basic working knowledge to be able to do
Various universities and medical centres could provide good
starting points.

Integrating Biological and Computational Models

The interplay between (nano-)biological knowledge and computer
science can be beneficial for both. The brain is incredibly
more complex and more capable of various types of computation
than any foreseeable generation of computers, and does these
computations in a higly efficient fashion.
On the other hand, digital computers offers well defined,
repeatable, transportable behaviour, and can be constructed and
programmed with good predictability as to their behaviour.
Computer oriented thinking, and computational models my provide
a good basis to form models for behaviour of micro, meso and
macroscopic behaviour of the brain (resp. sub-single neuron level,
neuron interaction for medium sized sets, and simulation of
larger parts of a brain), including sensitivies and radiation
at various levels.

Musical Instrument Simulation
A long standing interest of mine are electronic/digital
synthesizers, which lately are basically digital signal processing
units programmed to mimic various types of existing instruments
and create rich palettes of new ones. Basic synthesis for instance
in the form of sample processing is no longer a real scientific
challenge, simple and cheap hardware can do a good jos at that,
and can deliver musically satisfactory results, which are however
limited in their sonic appeal and range of instruments they can
faithfully reproduce.
Various forms of synthesis, that is signal processing algorithms
have evolved and are fairly well cristallized into a few handfull
of widely known synthesizer types.
None of these types until recently made an attempt to simulate
the actual behaviour of acoustical instruments, they merely
produce sound paletes that each have their own appeal and can
do a resonable to quite good simulation of various instruments.
There have been successfull attempts to use physical models
as the basis of synthesis, but they require a lot more computer
power than an ordinary digital synth currently has, and have been
limited mainly to off-line processing on large computers.
And even then, the models represent only fairly simple instruments,
such as a moving air column with simple exitation (as in wind
instruments such an organ pipe or a trumpet, or a relatively
simple string model).
Recent developments in DSP technology (see for instance the TI
tms 320c80 which delivers up to 200MOPS on one chip) have allowed
some synthesizer producing companies to construct commercial
synths that feature physical modeling as (one of) their
programming models. They can do very convincing simulations
of certain acoustic instrumentents, and are capable of responding
in similar fashion as their physical counterparts such as
the overblowing of a trumpet.

More complex instruments, with as an extreme example a piano,
can be modeled by specifically designing a model which captures
its characteristics, which has been done quite succesfully
in various electronic instruments, but at this time it is
not feasible to have a synthesizer/computer which allows you
to for instance give the main parameters of an instruments
physical properties (such as geometric features, material
properties, string tension) to generate the sound such an
instrument would make for something more complex than the
abovementioned examples.
In terms of imagination, a lot of complex instruments and
generator algorithms, especially involving large amounts
of non-linearly interacting elements, can be thought of
that could make interesting and appealing sounds, but
which cannot be built with reasonable amounts of currently
available technology.
Interestingly enough, these types of algorithms and models
almost coincide with certain types of behaviour of the brain,
which is illustrated by the fact that there are clearly
demonstrable relations between certain music and the brain
behaviour it induces, and the other way around.
In terms of computations on a computer system, there are also
clear correspondences between brain 'wave' patterns and
the way physical models are built (and the way I'd like to built
them): waveguides with non-linear connections built from large
amounts of basic elements.

Computing facilities
In line with the previous paragraphs, there is a need for huge
DSP power both for simulating brain-like behaviour (including
what I expect there is more or less going on at nano-scale)
and for complex (new) digital synthesizers.
Experience teaches that using existing technology in DSP is
not completely comparable with using existing computer
technology, where architectures and software buildup has
pretty much converged to certain base types, because the
variation in communication structure in coupled DSP technology
can be compared to computer networks in one respect: lots of
variation are available, but unfortunately the same doesn't
hold for the standardization of the communication protocols
and the access of the networks. One could see a complex
interconnected DSP system a local information highway with
lots of ferari's racing around without too many standard
traffic rules.

One of my interests is to have program models and implementation
means to facilitate the automation of organization of these
mini highways in an efficient, intuitively understandable
and portable fashion.
Graphical representations of programs have not been to much
replacing textual programs, even though lately even PC's have
sufficient graphical capabilities to make that possible.
On of the reasons is the our von Neumann based (serial) programming
model doesn't get much more clearer when picturally represented
instead of a text that more or less runs in the same way when
read by a computer or a human. It can be imagined though, that
when a musical instrument, or the characteristics of a
reverberating room are to be modelled it can more intuitively
be (partly) done by graphical representation. And in fact,
old (but recently revived) analog modular synths use a kind
of graphical representation where there are modules for various
functions available, each with their own control knobs, which
are interconnected by patch cords. This is effectively a 3D model
of the structure of a sound such a synth can produce, and
similar thinking has been since long used for computer
representations of synthesizer structures: block diagrams
consisting of graphical units connected by lines, albeit in
2 dimensional representation only.
In short: both new and existing programming models can be
fruitfully represented graphically, and the morecomplex the
program model it usually pays more to represent it graphically.
Intuitive and sufficiently efficent 3D interfaces can be
programmed even on current generation PC's, so they can be
considered as an interesting extension to the programmers
interface to a program, and might help conceiving of new
programming models suitable to deal with the kind of chalenges
mentioned above.

Programming of Neuronal Behaviour
Most of the above applies for programming neuronal behaviour.
the main representative of this type of behaviour currently is
the 'neural network', which is a type of programming that
incorporates learning and has some resemblances with biological
behaviour going on in the brain. There are some serious
limitations in this programming method however, related to
the (very) limited representation of what actually goes on in
the brain and to intrinsic difficulties in the model concerning
the limited predictablility of at least the accuracy of the
learning behaviour, beside some theory about theoretical learning
bounds and possible behavioural complexity.
In other words it wouldn't do justice to the behaviour of our
brain to call neural nets (partial) imitations of it. Without
going into all the details, the main reasons are absence of
an explicit time behaviour, non-deterministic programming
methods ('learning') without completely specifiable behaviour
bounds, limit complexitity of the node behaviour (e.g. the
transfer function is often just linear, or comprised of
o single non-linear map acting on a linear sun of input signals),
non-existent modeling of the 'fire' behaviour of a neuron
(related to the first point) and therefore absence of
possibilities for exhibiting the extraordinary behaviour range
of combined fire behaviour of large sets of neurons and
their interactions. Without even going into the fact that in
sheer numbers even our most powerfull computers can get nowhere
near simulating the behaviour of our 100 billion or so neurons,
the main problem is that at nano scale, that is at sub-neuron
scale there is so much going on that it is not unreasonable
to expect that simulating even a single(!) neuron resonably
accurate will take quite some years of research, programming
and will require already significant computer resources.

Organizational Forms

The biology and neurology I touched upon are most likely
found in academic/medical surroundings with specializations
in these areas.
Several fields of the research related to the modeling of
neuronal behaviour, the construction, simulation, theoretical
and practical properties of non-linearly connected waveguide
(and other) type of elements, main lines in user interface
and programming developments would probably fit naturally at
universities with an interest in these topics.

My previous activities and results concerning some solutions
to the programming methods mentioned above and their
implementation to solve DSP (and certain inter-computer)
communication problems, and to give a possibly fruitfull
starting point for a multi threading-like approach to
more parallel-oriented programming, could fit in the corporate
computer world, since it could produce applicabe results in
the foreseeable future, and since I've only seen simular
but not equal approaches around thusfar.

Recently I've experimented with and looked into an old 'hobby'
of mine starting ast the beginning of highschool: to build
fairly complex electronic circuitry with modest means, and
fond that it still holds that it pays to build circuitry
that is not mainstream technolgy complety out of low-level
discrete parts. In other words for instance computer main
boards are built beyond any competetion by hardware
manufacturers, but for technology used in smaller numbers
it holds that they can be build at fairly low cost. The main
factors besides a PCB board designer and etcher, preferably a
circuit simulator and a good set of databooks are craftmanship,
a working knowledge of available parts, and good ideas to
lead to prototypes or small series of specialistic machinery.
I learned that commonly available parts still allow circuitry
to be built that is compative with commercially available
products, offering perspectives to make small series of
circuits with specialistic functions, such as the non-linearly
linked waveguides mentioned above, for experimentation.

Time Path Proposals