|
BioComp's core
technologies fall into four major categories: Data
Analytics, Modeling,
Understanding and Optimization. Each
category has numerous technologies within it that are used for varying purposes
under differing conditions. Please note that most core technologies are
vital to our business and are protected, in part, by trade secret.
Accordingly, we do not divulge their inner workings so that we do not lose
competitive advantage. Instead, we offer external views that provide you
functional utility and understanding.
The application of the
appropriate technology for a particular use or "problem" requires two things:
1) The necessary algorithms existence, invention or adaptation and 2) The
expertise to select the appropriate technology and apply it. For over 20
years, the key BioComp technology staff have researched, applied, and learned
the utility of different algorithms. We look for and invent "Robust"
algorithms, ones that are flexible and powerful enough to be applied in many
ways with little or no modification. Through this process we have modified
select algorithms to improve their performance and results. This is true
of "Neural Networks" which we have replaced with our invention, "Meshes", for
modeling. It is also true with our refinement of common generation-based
"Genetic Algorithms" (GA) with our GA-CPG algorithms, which are simpler yet more
effective, two important advantages. We are not the repository nor the
appliers of all possible algorithms, but we masters of a subset that are
effective for many of the key challenges our customers face.
We are proud of our technological
history, which includes being the first company to...
-
Offer a commercial Genetic
Algorithms library for Microsoft Windows (GAWindows) in the 1980's.
-
Market general purpose
"genetically optimized neural networks" (circa 1994). This model
development strategy was the first ever to employ genetic input selection while
simultaneously optimizing model type and structure. This capability was
delivered through our "NeuroGenetic Optimizer" (NGO) product, which is still
widely used today because of its base of 10's of thousands of licenses
distributed. This product has been retired and replaced with our new
desktop tools using "Mesh" modeling algorithms.
-
Market a high volume neural
network server ("Enterprise Modeling Server") that built and managed the quality
of 10's, 100's or 1000's of NGO-constructed models.
All of the above products are now
retired, replaced with better methods through our evolution in technology
leadership. Most of all, however, we are PROUD of the benefits our
customers have achieved using our technologies, helping save lives, defeat
diseases, properly value and protect assets and optimize their operations to
achieve savings or increased revenues of 100's of millions of dollars per year.
Data Analytics
(back to top)
Purpose
The purpose of data analytics is to perform analysis on data (raw,
cleaned,
converted and/or otherwise
pre-processed) in order to see cause and effect relationships before modeling
and optimization. This analysis gives insights into which variables impact
our variable(s) of interest, to what degree and when (time lags).
Advantages
With this information, one can reduce or eliminate redundant variables, find
key drivers of performance and enable improved modeling and optimization.
Technologies
Used in Products
Modeling
(back to top)
Purpose
To capture the functional cause and effect relationships between predicted variables using a set of one or more explaining (input or
causal) variables. Generally, these models are flexible mathematical
systems that are most often "regressed" using historical data, or smartly
summarize historical data for useful recall later. Clustering technologies
determine similarities and differences between entities and system states and
enable recognition of states seen before or are of similar nature.
Advantages
The models can predict future events, conditions and values and can be
inverted to create predictive control mechanisms that take corrective action
before adverse conditions are created.
Technologies
BioComp "Meshes" that perform linear, logistic and non-linear multivariate
regression, look-up tables and clustering
Used in Products
Understanding
(back to top)
Purpose
To explore the functional cause and effect relationships between one or more
causal (input) variables and one or more variable(s) of interest.
This helps the user, or the system itself, better understand what is driving performance,
the ability (or inability) to prediction it and to optimize it.
Advantages
Enhances the human's or system's understanding of cause and effect and the
nature of the relationships between causes and the results obtained.
Technologies
Used in Products
Optimization
(back to top)
Purpose
To determine actions to take, or settings to make, that improve the performance of a system of
interest.
Technical Uses
- Feed-forward predictive non-linear
multivariate constrained cascade control to achieve one or more objectives
[That is a mouthful, but very useful ! ]
- Non-linear direct action adaptive process
control
Example Applications
- Controlling multiple setpoints to increase
product conformance to specifications
- Maximizing production
- Mixing and matching sub-assemblies to maximize
resulting assemblies conformance to specification
Advantages
Significant gains are achieved through taking appropriate actions at
the right time.
Technologies
BioComp GA-CPG based predictive model inversion with constraints, limits and
desirabilities.
Gradient Ascent/Decent direct action (change-sense-change optimization)
BioComp GA-CPG based direct action (change-sense-change optimization)
Used in Products
|
