Biomedical Engineering Reference
In-Depth Information
Note that this model for an individual neuron is a greatly simplified representation of the function of
an actual neuron in the human nervous system. For example, neurons in the brain are regularly
bathed in substances—from naturally produced endorphins to drugs such as seratonin release
inhibitors—that dynamically alter the strength of connections, represented by the
fixed
weights (
w
) in
the neuron model. The advantage of ignoring the intricacies of the actual nervous system is
computational efficiency and lower overhead associated with developing a model. A simpler model is
also easier to develop and maintain compared to developing and maintaining a more complex model.
The challenge is defining a model that is simple enough computationally and yet is rich enough to
accurately define the behavior of the system.
Although spreadsheets are still used for modeling and simulation applications in business, science,
and engineering, all but the simplest modeling is performed with software optimized for particular
domains. For example, nuclear physicists use custom modeling and simulation programs running on
supercomputers to simulate the power of nuclear explosions. Similarly, life scientists use a variety of
microcomputer-based simulations to explore everything from population dynamics to the docking of
proteins.
The downside of using a general-purpose spreadsheet as a platform for modeling and simulation is
related to performance, flexibility, visualization capabilities, standards, and startup time. A general-
purpose spreadsheet, like a general-purpose language such as eXtensible Markup Language (XML) or
C++, is designed to solve a variety of problems. As such, it represents a compromise between
flexibility and performance. Although a spreadsheet can be used to prototype virtually any type of
simulation, the simulation will likely run several orders of magnitudes slower than a simulation
developed in an environment designed for modeling and simulation.
Similarly, coding a simulation in C++ may result in a system with a higher performance than can be
obtained with a dedicated simulation system. However, the startup time associated with a domain-
specific simulation will likely be several orders of magnitude lower that that associated with the
general-purpose language. For example, classification systems based on a neural network simulation
are typically outperformed by classification systems developed in C++ or some other compiled
language. However, creating a classification system with a neural network system may take only
