Now Available to Software Developers:
The Pattern Memory Engine (PME)
The Building Block for Intelligent Learning Machines
PME Learns to Recognize or Cluster Complex Patterns that are
Subject to Noise
Patterns in Large Data Bases or from Multiple Real-Time
Sensors .....Fast and
Simple.....
Application Need: Technology is
creating an environment of exponentially increasing
volume of information stored in databases, on the
internet, and from different real time sensors such
as video. Application fields include data mining,
signal processing, robotic guidance, biometric
security, and video surveillance. Extracting useful
patterns of information from large databases or
multiple real time sensors requires speed.
PME provides the need for speed.
Technical Problem Solved: The technical
problem is to cluster or recognize similar high
dimensional (complex) patterns that are distorted by
noise. Alternative algorithms recognize or cluster
patterns by comparing it to known patterns in a
database. With large databases the search for a
match is too slow. PME eliminates this search and
does not need to store the originally trained
(known) patterns. PME
is typically more than 10,000 times faster than
alternatives.
The Product: PME is
available as a software developer kit (SDK).
Thousands of installations are using earlier
versions of PME to recognize hand printed
characters. The technology originated from neural
models of human pattern recognition. Patents have
been granted and others are pending. The SDK enables
corporate product developers to enhance their
existing and future products.
PME's simplicity allows it to
be easily embedded in software or hardware
applications.
Application Examples: * Recognizing
one face, finger print, or iris in a data base of
100,000 for biometric
security
(recognition in microseconds)
* Real time object tracking and recognition in video
for smart surveillance
cameras and vehicle/robot guidance
* Learning normal patterns from multiple sensors for
real time outlier detection in process monitoring
* Recognizing malware signatures that are similar to
known threats for
internet security
* Recognizing / clustering similar images for
content based image retrieval in search engines
* Finding all similar clusters of customer
preferences in a large database for data mining
* Implementing PME in hardware (semiconductor) as a
general purpose tool in high speed computing hardware
* Fusion of patterns from different types of sensors
for signal processing
PME is a tool
and not specific to any one application. In all
cases patterns are represented by a list of numbers
that describe the pattern. Basically, PME clusters
or recognizes similar lists of numbers - the common
application challenge.