There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

A new aspect of cell membrane structure is presented, based on the dynamic clustering of sphingolipids and cholesterol to form rafts that move within the fluid bilayer. It is proposed that these rafts function as platforms for the attachment of proteins when membranes are moved around inside the cell and during signal transduction.

Functional rafts in cell membranes

We will review some of the major results in random graphs and some of the more challenging open problems. We will cover algorithmic and structural questions. We will touch on newer models, including those related to the WWW.

Random graphs

An analytic solution of the waiting time distribution for the discrete-time GI/G/1 queue

Wavelength converters are effective to reduce connection blocking probability in dynamic WDM networks. Full complete conversion, where all the output wavelengths of each node are equipped with converters, provides the best performance. However, it requires huge amount of wavelength converters, which are expected to be quite expensive in the foreseeable future. On the other hand, partial conversion leads to much less cost by reducing the number of converters in each node. To approximate the best performance provided by full complete conversion using partial conversion, two questions have to be answered: a) How many converters does the network need? b) What are the optimal converter locations? For the first time, this paper presents the answers using a heuristic algorithm based on an analytical model. In particular, we develop an algorithm to deal with the case in real networks where the accurate traffic estimation is not available. Our approach is shown to be both efficient and effective: the calculation in either NSFNet or EON needs no more than 30 seconds using an ordinary PC; and the designed networks with partial conversion ability approximate the best performance very well under both uniform and non-uniform traffic.

How many wavelength converters do we need

By introducing video-quality adaptation mechanisms into intermediate network equipments using active network technologies, we can provide users with video distribution services taking into account client heterogeneity in terms of available bandwidth, performance of client systems, and user’s preferences about video quality. In this paper, we implement the low-pass filter, a quality adjustment technique for real-time multicasting of MPEG-2 video, on an Intel IXP1200 network processor-based network node. We applied the filter to video streams passing through the node and evaluated its practicality and applicability in term of accuracy of video rate adaptation, variation of video quality, and filtering throughput. From the result of evaluation experiments, we demonstrate that the implemented video-quality adjustment mechanism has sufficient rate adaptation capability, and that the low-pass filter is able to accelerate with parallel processing.

Implementation and Evaluation of Video-Quality Adjustment for Heterogeneous Video Multicast

The authors analyze the flow enforcement algorithm suitable for regulating the flow of the bursty traffic in asynchronous transfer mode (ATM) networks (for broadband packet switching). The flow enforcement algorithm has two objectives: to force the input traffic to conform to the traffic parameters which are specified by the source itself at its connection time; and to obtain greater performance gain at the cell multiplexer by regulating the flow to some extent. The authors reveal the tradeoff relationship between the regulation level of the cell flow and the improvement of the link performance. The authors first analyze the flow enforcement mechanism to derive the delay distribution and the interdeparture time distribution for cells at the flow enforcer. For the input traffic, they allow the general independent arrivals of cells to represent the traffic burstiness larger than a Poisson distribution. The obtained interdeparture time distributions from the flow enforcer are then used to approximately analyze the performance of the cell multiplexer. For this purpose, a Markov-modulated Poisson process (MMPP) approximation method is extended to investigate the performance of individual traffic stream in the superposed arriving traffic streams.<<ETX>>

Analysis of Flow Enforcement Algorithm for Bursty Traffic in ATM Networks

Overlay networks achieve new functionality and enhance network performance by allowing routing to be controlled at the application layer. However, these approaches result in degradations of underlying networks due to the selfish behavior of overlay networks. In this paper, we investigate the stability of virtual network topology (VNT) control under the overlay networks that perform dynamic routing updates. We reveal that the dynamics of routing on overlay networks causes a high fluctuation in the traffic demand matrix, which leads to significant instability of VNT control. To overcome the instability induced by the overlay routing, we introduce hysteresis to the VNT control. Simulation results indicate that the hysteresis mechanism improves the network stability, but cannot always improve the network performance. We therefore extend the hysteresis mechanism and show that the proposed method improves both the network stability and the performance when the amount of traffic for overlay network is not large.

Masayuki Murata

Papers

An analytic solution of the waiting time distribution for the discrete-time GI/G/1 queue

How many wavelength converters do we need

Implementation and Evaluation of Video-Quality Adjustment for Heterogeneous Video Multicast

Analysis of Flow Enforcement Algorithm for Bursty Traffic in ATM Networks

On the stability of virtual network topology control for overlay routing services