Ecse

This paper compares various techniques for improving the frequency response of silicon photodiodes fabricated in mainstream CMOS technology for fully integrated optical receivers. The three presented photodiodes, Spatially Modulated Light detectors, Double, and Interrupted P-Finger photodiodes, aim at reducing the low speed diffusive component of the photo generated current. For the first photodiode, Spatially Modulated Light (SML) detectors, the low speed current component is canceled out by converting it to a common mode current driving a differential transimpedance amplifier. The Double Photodiode (DP) uses two depletion regions to increase the fast drift component, while the Interrupted-P Finger Photodiode (IPFPD) redirects the low speed component towards a different contact from the main fast terminal of the photodiode. Extensive device simulations using 130 nm CMOS technology-parameters are presented to compare their performance using the same technological platform. Finally a new type of photodiode that uses triple well CMOS technology is introduced that can achieve a bandwidth of roughly 10 GHz without any process modification or high reverse bias voltages that would jeopardize the photodetector and subsequent transimpedance amplifier reliability.

Clustering suffers from the curse of dimensionality, and similarity functions that use all input features with equal relevance may not be effective. We introduce an algorithm that discovers clusters in subspaces spanned by different combinations of dimensions via local weightings of features. This approach avoids the risk of loss of information encountered in global dimensionality reduction techniques, and does not assume any data distribution model. Our method associates to each cluster a weight vector, whose values capture the relevance of features within the corresponding cluster. We experimentally demonstrate the gain in perfomance our method achieves, using both synthetic and real data sets. In particular, our results show the feasibility of the proposed technique to perform simultaneous clustering of genes and conditions in microarray data.

Clustering suffers from the curse of dimensionality, and similarity functions that use all input features with equal relevance may not be effective. We introduce an algorithm that discovers clusters in subspaces spanned by different combinations of dimensions via local weightings of features. This approach avoids the risk of loss of information encountered in global dimensionality reduction techniques, and does not assume any data distribution model. Our method associates to each cluster a weight vector, whose values capture the relevance of features within the corresponding cluster. We experimentally demonstrate the gain in perfomance our method achieves with respect to competitive methods, using both synthetic and real datasets. In particular, our results show the feasibility of the proposed technique to perform simultaneous clustering of genes and conditions in gene expression data, and clustering of very high-dimensional data such as text data.

Temporal data mining aims at finding patterns in historical data. Our work proposes an approach to extract temporal patterns from data to predict the occurrence of target events, such as computer attacks on host networks, or fraudulent transactions in financial institutions. Our problem formulation exhibits two major challenges: 1) we assume events being characterized by categorical features and displaying uneven inter-arrival times; such an assumption falls outside the scope of classical time-series analysis, 2) we assume target events are highly infrequent; predictive techniques must deal with the class-imbalance problem. We propose an efficient algorithm that tackles the challenges above by transforming the event prediction problem into a search for all frequent eventsets preceding target events. The class imbalance problem is overcome by a search for patterns on the minority class exclusively; the discrimination power of patterns is then validated against other classes. Patterns are then combined into a rule-based model for prediction. Our experimental analysis indicates the types of event sequences where target events can be accurately predicted.

Periodic behavior is common in real-world applications. However in many cases, periodicities are partial in that they are present only intermittently. The authors study such intermittent patterns, which they refer to as p-patterns. The formulation of p-patterns takes into account imprecise time information (e.g., due to unsynchronized clocks in distributed environments), noisy data (e.g., due to extraneous events), and shifts in phase and/or periods. We structure mining for p-patterns as two sub-tasks: (1) finding the periods of p-patterns and (2) mining temporal associations. For (2), a level-wise algorithm is used. For (1), we develop a novel approach based on a chi-squared test, and study its performance in the presence of noise. Further we develop two algorithms for mining p-patterns based on the order in which the aforementioned sub-tasks are performed: the period-first algorithm and the association-first algorithm. Our results show that the association-first algorithm has a higher tolerance to noise; the period-first algorithm is more computationally efficient and provides flexibility as to the specification of support levels. In addition, we apply the period-first algorithm to mining data collected from two production computer networks, a process that led to several actionable insights

We describe algorithms and an architecture for a real-time problem determination system that uses online selection of most-informative measurements -the approach called herein active probing. Probes are end-to-end test transactions which gather information about system components. Active probing allows probes to be selected and sent on-demand, in response to one's belief about the state of the system. At each step the most informative next probe is computed and sent. As probe results are received, belief about the system state is updated using probabilistic inference. This process continues until the problem is diagnosed. We demonstrate through both analysis and simulation that the active probing scheme greatly reduces both the number of probes and the time needed for localizing the problem when compared with non-active probing schemes.

In this paper, we propose a new clustering algorithm, IFD 1 , based on a cluster model of data coefficients D and feature coefficients F. The coefficients denote the degree (or weights) of the data and features associated with the clusters. Clustering is performed via an iterative optimization procedure to mutually reinforce the relationships between the coefficients. The mutually reinforcing optimization exploits the duality of the data and features and enable a simultaneous clustering of both data and features. We have shown the convergence property of the clustering algorithm and discussed its connections with various existential approaches. Extensive experimental results on both synthetic and real data sets show the effectiveness of IFD algorithm.

Learning to predict significant events from sequences of data with categorical features is an important problem in many application areas. We focus on events for system management, and formulate the problem of prediction as a classification problem. We perform co-occurrence analysis of events by means of Singular Value Decomposition (SVD) of the examples constructed from the data. This process is combined with Support Vector Machine (SVM) classification, to obtain efficient and accurate predictions. We conduct an analysis of statistical properties of event data, which explains why SVM classification is suitable for such data, and perform an empirical study using real data.

Monitoring systems are able to capture thousands of events from a computer network. Some of those events may be particularly informative to ensure the correct operation of an application. We assume the user is interested in a specific class of events, called target events (e.g., communication link is down). We propose a system that generates a set of rules correlating each target event with events occurring previous to the target event within a specified time window interval. Such rules can be extremely helpful in elucidating the origin of target events. We conduct experiments to assess the accuracy of the induced rules for different types of target events in a real-world network environment. Our results show the accuracy of the induced rules generally above 80% when the time window interval is at least 20 minutes wide. Such results give strong empirical support to the validity of our approach.

As distributed systems continue to grow in size and complexity, scalable and cost-efficient techniques are needed for performing tasks such as problem determination and fault diagnosis. In this paper, we address these tasks using probes, or test transactions, which replace traditional "passive" event-correlation techniques with a more active, real-time information-gathering approach. We provide a theoretical foundation and a set of practical techniques for implementing efficient probing strategies-the main issue is minimizing the cost of probing while maximizing the diagnostic accuracy of the probe set. We show that finding an optimal probe set is NP-hard and devise polynomial-time approximation algorithms that demonstrate excellent empirical performance, even on large networks. We also implement an active, on-line probing strategy that yields a significant reduction in the probe set size.

In this paper, we address the problem of efficient diagnosis in real-time systems capable of on-line information gathering, such as sending "probes" (i.e., test transactions, such as "traceroute" or "ping") in order to identify network faults and evaluate performance of distributed computer systems. We use a Bayesian network to model probabilistic relations between the problems (faults, performance degradation) and symptoms (probe outcomes). Due to intractability of exact probabilistic inference in large systems, we investigated approximation techniques, such as a local-inference scheme called mini-buckets(Dechter & Rish 1997). Our empirical study demonstrates advantages of local approximations for large diagnostic problems: the approximation is very efficient and "degrades gracefully" with noise; also, the approximation error gets smaller on networks with higher confidence (probability) of the exact diagnosis. Since the accuracy of diagnosis depends on how much information the probes can provide about the system states, the second part of our work is focused on the probe selection task. Small probe sets are desirable in order to minimize the costs imposed by probing, such as additional network load and data management requirements. Our results show that, although finding the optimal collection of probes is expensive for large networks, efficient approximation algorithms can be used to find a nearly-optimal set.

We consider the use of probing technology for fault diagnosis in computer networks. Probes are test transactions that can be actively selected and sent through the network. The use of probing technology for costeffective diagnosis requires addressing two issues: a planning phase in which the probes are selected, followed by a diagnosis phase in which problem determination is performed using the results of the probes. The planning phase requires selecting a small but effective subset of all the possible probes. The diagnosis phase requires making inferences about the state of the network from the probe results in an environment of noise and uncertainty. This work addresses the probing problem using methods from artificial intelligence-we call the resulting approach intelligent probing. The probes are selected by reasoning about the interactions between the probe paths. Although finding the optimal probe set is prohibitively expensive for large networks, we implement algorithms which find near-optimal probe sets in linear time. In the diagnosis phase we use a Bayesian network approach and use a local-inference approximation scheme that avoids the intractability of exact inference for large networks. Our results show that the quality of this approximate inference "degrades gracefully" under increasing uncertainty and increases as the quality of the probe set increases.

Real-time problem diagnosis in large distributed computer systems and networks is a challenging task that requires fast and accurate inferences from potentially huge data volumes. In this paper, we propose a cost-efficient, adaptive diagnostic technique called active probing. Probes are end-to-end test transactions that collect information about the performance of a distributed system. Active probing uses probabilistic reasoning techniques combined with information-theoretic approach, and allows a fast online inference about the current system state via active selection of only a small number of most-informative tests. We demonstrate empirically that the active probing scheme greatly reduces both the number of probes (from 60% to 75% in most of our real-life applications), and the time needed for localizing the problem when compared with non-adaptive (pre-planned) probing schemes. We also provide some theoretical results on the complexity of probe selection, and the effect of "noisy" probes on the accuracy of diagnosis. Finally, we discuss how to model the system's dynamics using Dynamic Bayesian networks, and an efficient approximate approach called sequential multifault; empirical results demonstrate clear advantage of such approaches over "static" techniques that do not handle system's changes.

To obtain classification systems with both good generalizatìon performance and efficiency in space and time, we propose a learning method based on combinations of weak classifiers, where weak classifiers are linear classifiers (perceptrons) which can do a little better than making random guesses. A randomized algorithm is proposed to find the weak classifiers. They are then combined through a majority vote. As demonstrated through systematic experiments, the method developed is able to obtain combinations of weak classifiers with good generalization performance and a fast training time on a variety of test problems and real applications. Theoretical analysis on one of the test problems investigated in our experiments provides insights on when and why the proposed method works. In particular, when the strength of weak classifiers is properly chosen, combinations of weak classifiers can achieve a good generalization performance with polynomial space-and time-complexity.

Real-time problem diagnosis in large distributed computer systems and networks is a challenging task that requires fast and accurate inferences from potentially huge data volumes. In this paper, we propose a cost-efficient, adaptive diagnostic technique called active probing. Probes are end-to-end test transactions that collect information about the performance of a distributed system. Active probing uses probabilistic reasoning techniques combined with information-theoretic approach, and allows a fast online inference about the current system state via active selection of only a small number of most-informative tests. We demonstrate empirically that the active probing scheme greatly reduces both the number of probes (from 60% to 75% in most of our real-life applications), and the time needed for localizing the problem when compared with non-adaptive (pre-planned) probing schemes. We also provide some theoretical results on the complexity of probe selection, and the effect of "noisy" probes on the accuracy of diagnosis. Finally, we discuss how to model the system's dynamics using Dynamic Bayesian networks, and an efficient approximate approach called sequential multifault; empirical results demonstrate clear advantage of such approaches over "static" techniques that do not handle system's changes.