Biomedical Engineering Reference
In-Depth Information
Theory of Brain
Information Processing
Computer Analogy
Bottom-Up Approach
FIgURE 1.3:
Bifurcation in the development of computational models of the brain.
it.” In either case, the computational approaches used in the development of BMIs contain four
pillars: hypothesis of how the brain functions, development of a model, analytically or numerically
evaluation of the models, and testing against experimental data. In this framework, modeling plays
a big part in the inductive or deductive reasoning used to establish functional associations among
neural assemblies. For this reason, the models can take several forms: computational, which focus on
which computation is to be performed, in terms of optimality, modularity and others, algorithmic,
which look at nature of computations performed, mechanistic, that are based on known anatomy
and physiology, descriptive, which summarizes large amounts of experimental data, or interpretive,
which explores behavioral and cognitive significance of nervous system function.
The integrative computational neuroscience perspective and the latest technology innovation
have been consistently applied to explain brain function, at least since the time of Rene Descartes.
Perhaps, the formal computer metaphor of Turing and Ince [
22
] and Von Neumann [
23
] is still
the one that has more followers today, but others are also relevant, such as the cybernetic model-
ing of Wiener [
24
], the universal computer analogy of McCulloch and Pitts [
25
], the holographic
approach of Pribram [
26
], the connectionist modeling of McClelland and Rumelhart [
27
], and
Rumelhart [
28
], Rumelhart et al. [
29
], and Hopfield [
30
], and the synergetic modeling of Haken
[
31
] and Freeman [
32
], to name some important examples. This description adapted from Freeman,
captures well the difficulty of the modeling task: “Unlike existing computational systems, brains and
the sensory cortices embedded within them are highly unstable. They jump from one quasi-stable
state to the next at frame rates of 5-10/sec under internal phase transitions. The code of each frame
is a 2D spatial pattern of amplitude or phase modulation of an aperiodic (chaotic) carrier wave.
The contents of each frame are selected by the input and shaped by synaptic connection weights
embodying past learning. These frames are found in the visual, auditory, somatic, and olfactory
cortices. Having the same code, they readily combine to form multisensory images (gestalts) that
provide a landscape of basins and attractors for learned classes of stimuli.”
