Databases Reference
In-Depth Information
brain. Harnessing the potential of this interconnectivity of high-performance
machines over large-scale networks may provide recognition schemes for large-
scale and complex data. With the advent of high-resolution digital instruments
and sensors in areas such as biomedical and satellite imaging, such large-scale
and complex data are becoming increasingly available.
Machine intelligence is important in large-scale data applications. In
biomedical imaging, intelligence schemes are commonly used to analyze and
extract important and critical features from high-dimensional images obtained
through sophisticated imaging techniques, such as Magnetic Resonance Imag-
ing (MRI). In addition, computational intelligence schemes can be used by
medical experts to assist in their diagnosis. With the advancements in high-
speed networking technology, medical experts can conduct a collaborative di-
agnosis by collecting data from instruments over large networks and storing
or updating these data in distributed repositories. With this capability at
hand, the amount of data generated as part of the distributed system is at
the Internet-scale.
Depending on the in-depth resolution of satellite imaging, the size of the
generated data can be huge. Satellite imaging is important in a number of
applications, including the geographical information system (GIS) and the
global positioning system (GPS). To produce useful geographical images, the
raw images taken from the satellite camera must be processed. A number of
processes are required, including image extraction and manipulation. These
processes ensure that the data satisfy the resolution and size requirements
for specific applications. Machine intelligence schemes can be very useful in
performing these operations effectively.
A rapid growth in large-scale scientific analysis activities has inspired the
development of sophisticated and state-of-the-art facilities. One example is a
synchrotron, a scientific facility that performs cyclic particle acceleration. The
data generated by such facilities are images of the interaction of the particle
beam with targets at a sub-atomic scale. An average beam line can produce
hundreds of megabytes (Mb) of images continuously throughout the year. Syn-
chrotron facilities are being used for a number of applications including large
molecule crystallography and other chemical analyses. In addition, sophisti-
cated data-capture instruments and sensors developed for high energy physics
facilities, such as the Large Hadron Collider (see Figure 1.1) and Interfero-
metric Synthetic Aperture Radar (InSAR), consistently generate extremely
large volumes of highly complex and often invaluable data.
These state-of-the-art data capture and storage technologies are the key
factors that have led to the generation of highly complex and large-scale data.
The volumes alone make it impractical for data analysts to analyze and ex-
plore the data without the assistance of highly sophisticated computational
tools. As mentioned earlier, the data mining and analysis capabilities of ex-
isting applications have not achieved their fullest potential. This shortfall is
attributed mainly to the algorithmic complexity of existing data mining ap-
plications. For instance, the complexity of a decision tree classification tool
