Definition
These notes provide an introduction to unsupervised neural networks, in particular Kohonen self-organizing maps; together with some fundamental background material on statistical pattern recognition.
One question which seems to puzzle many of those who encounter unsupervised learning for the first time is how can anything useful be achieved when input information is simply poured into a black box with no provision of any rules as to how this information should be stored, or examples of the various groups into which this information can be placed. If the information is sorted on the basis of how similar one input is with another, then we will have accomplished an important step in condensing the available information by developing a more compact representation.
We can represent this information, and any subsequent information, in a much reduced fashion. We will know which information is more likely. This black box will certainly have learned. It may permit us to perceive some order in what otherwise was a mass of unrelated information to see the wood for the trees.
In any learning system, we need to make full use of the all the available data and to impose any constrains that we feel are justified. If we know that what groups the information must fall into, that certain combinations of inputs preclude others, or that certain rules underlie the production of the information then we must use them. Often, we do not possess such additional information. Consider two examples of experiments. One designed to test a particular hypothesis, say, to determine the effects of alcohol on driving; the second to investigate any possible connection between car accidents and the driver's lifestyle.
In the first experiment, we could arrange a laboratory-based experiment where volunteers took measured amounts of alcohol and then attempted some motor-skill activity (e.g., following a moving light on a computer screen by moving the mouse). We could collect the data (i.e., amount of alcohol vs. error rate on the computer test), conduct the customary statistical test and, finally, draw our conclusions. Our hypothesis may that the more alcohol consumed the greater the error rate we can confirm this on the basis of this experiment. Note, that we cannot prove the relationship only state that we are 99% certain (or whatever level we set ourselves) that the result is not due purely to chance.
The second experiment is much more open-ended (indeed, it could be argued that it is not really an experiment).Data is collected from a large number of drives those that have been involved in accidents and those that have not. This data could include the driver's age, occupation, health details, drinking habits, etc. From this mass of information, we can attempt to discover any possible connections. A number of conventional statistical tools exist to support this (e.g., factor analysis). We may discover possible relationships including one between accidents and drinking but perhaps many others as well. There could be a number of leads that need following up. Both approaches are valid in searching for causes underlying road accidents. This second experiment can be considered as an example of unsupervised learning.
The next section provides some introductory background material on statistical pattern recognition. The terms and concepts will be useful in understanding the later material on unsupervised neural networks. As the approach underlying unsupervised networks is the measurement of how similar (or different) various inputs are, we need to consider how the distances between these inputs are measured. This forms the basis Section Three, together with a brief description of non-neural approaches to unsupervised learning. Section Four discusses the background to and basic algorithm of Kohonen self-organizing maps. The next section details some of the properties of these maps and introduces several useful practical points. The final section provides pointers to further information on unsupervised neural networks.
Definition
Survivability In Network Systems
Contemporary large-scale networked systems that are highly distributed improve the efficiency and effectiveness of organizations by permitting whole new levels of organizational integration. However, such integration is accompanied by elevated risks of intrusion and compromise. These risks can be mitigated by incorporating survivability capabilities into an organization's systems. As an emerging discipline, survivability builds on related fields of study (e.g., security, fault tolerance, safety, reliability, reuse, performance, verification, and testing) and introduces new concepts and principles. Survivability focuses on preserving essential services in unbounded environments, even when systems in such environments are penetrated and compromised.
The New Network Paradigm: Organizational Integration
From their modest beginnings some 20 years ago, computer networks have become a critical element of modern society. These networks not only have global reach, they also have impact on virtually every aspect of human endeavor. Network systems are principal enabling agents in business, industry, government, and defense. Major economic sectors, including defense, energy, transportation, telecommunications, manufacturing, financial services, health care, and education, all depend on a vast array of networks operating on local, national, and global scales. This pervasive societal dependency on networks magnifies the consequences of intrusions, accidents, and failures, and amplifies the critical importance of ensuring network survivability.
As organizations seek to improve efficiency and competitiveness, a new network paradigm is emerging. Networks are being used to achieve radical new levels of organizational integration. This integration obliterates traditional organizational boundaries and transforms local operations into components of comprehensive, network-resident business processes. For example, commercial organizations are integrating operations with business units, suppliers, and customers through large-scale networks that enhance communication and services.
These networks combine previously fragmented operations into coherent processes open to many organizational participants. This new paradigm represents a shift from bounded networks with central control to unbounded networks. Unbounded networks are characterized by distributed administrative control without central authority, limited visibility beyond the boundaries of local administration, and lack of complete information about the network. At the same time, organizational dependencies on networks are increasing and risks and consequences of intrusions and compromises are amplified.
The Definition of Survivability
We define survivability as the capability of a system to fulfill its mission, in a timely manner, in the presence of attacks, failures, or accidents. We use the term system in the broadest possible sense, including networks and large-scale systems of systems. The term mission refers to a set of very high-level (i.e., abstract) requirements or goals.
Missions are not limited to military settings since any successful organization or project must have a vision of its objectives whether expressed implicitly or as a formal mission statement. Judgments as to whether or not a mission has been successfully fulfilled are typically made in the context of external conditions that may affect the achievement of that mission. For example, assume that a financial system shuts down for 12 hours during a period of widespread power outages caused by a hurricane.
If the system preserves the integrity and confidentiality of its data and resumes its essential services after the period of environmental stress is over, the system can reasonably be judged to have fulfilled its mission. However, if the same system shuts down unexpectedly for 12 hours under normal conditions (or under relatively minor environmental stress) and deprives its users of essential financial services, the system can reasonably be judged to have failed its mission, even if data integrity and confidentiality are preserved.
Definition
The world of mobile computing has seldom been so exciting. Not, at least, for last 3 years when all that the chip giants could think of was scaling down the frequency and voltage of the desktop CPUs, and labeling them as mobile processors. Intel Centrino mobile technology is based on the understanding that mobile customers value the four vectors of mobility: performance, battery life, small form factor, and wireless connectivity. The technologies represented by the Intel Centrino brand will include an Intel Pentium-M processor, Intel 855 chipset family, and Intel PRO/Wireless 2100 network connection .
The Intel Pentium-M processor is a higher performance, lower power mobile processor with several micro-architectural enhancements over existing Intel mobile processors. Some key features of the Intel Pentium-M processor Micro-architecture include Dynamic Execution, 400-MHz, on-die 1-MB second level cache with Advanced Transfer Cache Architecture, Streaming SIMD Extensions 2, and Enhanced Intel SpeedStep technology.
The Intel Centrino mobile technology also includes the 855GM chipset components GMCH and the ICH4-M. The Accelerated Hub Architecture is designed into the chipset to provide an efficient, high bandwidth, communication channel between the GMCH and the ICH4-M.The GMCH component contains a processor system bus controller, a graphics controller, and a memory controller, while providing an LVDS interface and two DVO ports.
The integrated Wi-Fi Certified Intel PRO/Wireless 2100 Network Connection has been designed and validated to work with all of the Intel Centrino mobile technology components and is able to connect to 802.11b Wi-Fi certified access points. It also supports advanced wireless LAN security including Cisco LEAP, 802.1X and WEP. Finally, for comprehensive security support, the Intel PRO/Wireless 2100 Network Connection has been verified with leading VPN suppliers like Cisco, CheckPoint, Microsoft and Intel NetStructure.
Pentium-M Processor
The Intel Pentium-M processor is a high performance, low power mobile processor with several micro-architectural enhancements over existing Intel mobile processors. The following list provides some of the key features on this processor:
¢ Supports Intel Architecture with Dynamic Execution
¢ High performance, low-power core
¢ On-die, 1-MByte second level cache with Advanced Transfer Cache Architecture
¢ Advanced Branch Prediction and Data Prefetch Logic
¢ Streaming SIMD Extensions 2 (SSE2)
¢ 400-MHz, Source-Synchronous processor system bus
¢ Advanced Power Management features including Enhanced Intel SpeedStep technology
¢ Micro-FCPGA and Micro-FCBGA packaging technologies
The Intel Pentium-M processor is manufactured on Intel's advanced 0.13 micron process technology with copper interconnect. The processor maintains support for MMX technology and Internet Streaming SIMD instructions and full compatibility with IA-32 software. The high performance core features architectural innovations like Micro-op Fusion and Advanced Stack Management that reduce the number of micro-ops handled by the processor. This results in more efficient scheduling and better performance at lower power.
The on-die 32-kB Level 1 instruction and data caches and the 1-MB Level 2 cache with Advanced Transfer Cache Architecture enable significant performance improvement over existing mobile processors. The processor also features a very advanced branch prediction architecture that significantly reduces the number of mispredicted branches. The processor's Data Prefetch Logic speculatively fetches data to the L2 cache before an L1 cache requests occurs, resulting in reduced bus cycle penalties and improved performance.
