Search

Spiking Neural Networks (SNNs) have shown great promise in processing spatiotemporal information compared to Artificial Neural Networks (ANNs). However, there remains a performance gap between SNNs and ANNs, which impedes the practical application of SNNs. With intrinsic event-triggered property and temporal dynamics, SNNs have the potential to effectively extract spatiotemporal features from event streams. To leverage the temporal potential of SNNs, we propose a self-attention-based temporal-channel joint attention SNN (STCA-SNN) with end-to-end training, which infers attention weights along both temporal and channel dimensions concurrently. It models global temporal and channel information correlations with self-attention, enabling the network to learn 'what' and 'when' to attend simultaneously. Our experimental results show that STCA-SNNs achieve better performance on N-MNIST (99.67%), CIFAR10- DVS (81.6%), and N-Caltech 101 (80.88%) compared with the state-of-the-art SNNs. Meanwhile, our ablation study demonstrates that STCA-SNNs improve the accuracy of event stream classification tasks. [ABSTRACT FROM AUTHOR]

Event cameras are asynchronous sensors that operate in a completely different way from traditional cameras.Rather than catching pictures at a steady rate,event cameras measure light changes (called events) separately for every pixel.As a sequence,it alleviates the problems of traditional cameras in complex light conditions and scenes where objects move at high speed.With the development of convolutional neural networks,learning-based pattern recognition methods have made great progress in visual tasks such as optical flow estimation and target recognition by converting the output of the event camera into a pseudo-ima-ge representation.However,such methods abandon the temporal correlation between the event streams,so that the texture of the pseudo image is not clear enough,and it is difficult to extract the features.The key to solving this problem lies in how to model relevant information between events in the sample.Therefore,a neural network framework based on event stream partition algorithm is proposed,which explicitly integrates the temporal information of event streams.The framework divides the incoming stream of events into several parts,and a weight distribution network assigns different weights to each piece of the streams.Then,the framework uses convolutional neural network to fuse temporal information and extract advanced features.Finally,the input sample is classified.We thoroughly validate the proposed framework on object recognition.Comparison experiments on N-Caltech101 and N-cars datasets show that the proposed framework has a significant improvement in classification accuracy compared with the most advanced existing algorithms.

Abstract Critical systems that produce big data streams can require human operators to monitor these event streams for changes of interest. Automated systems which oversee many tasks can still have a need for the ‘human-in-the-loop’ operator to evaluate whether an intervention is required due to a lack of suitable training data initially offered to the system which would allow a correct course of actions to be taken. In order for an operator to be capable of reacting to real-time events, the visual depiction of the event data must be in a form which captures essential associations and is readily understood by visual inspection. A similar requirement can be found during inspections on activity protocols in a large organization where a code of correct conduct is prescribed and there is a need to oversee whether the activity traces match the expectations, with minimal delay. The methodology presented here addresses these concerns by providing an adaptive window sizing measurement for subsetting the data, and subsequently produces a set of network diagrams based upon event label co-occurrence networks. With an intuitive method of network construction the amount of time required for operators to learn how to monitor complex event streams of big datasets can be reduced.

Business processes are dynamic and change due to diverse factors. While existing approaches aim to detect drifts in the process structure, Tesseract looks for temporal drifts in activity interim times. This orthogonal view on the process extends the traditional data cube of events – case id, activities and timestamps – by a fourth dimension and improves the operational support by a visualization of temporal drifts in real-time. Insights about temporal deviations lead to an augmented awareness of imminent failures or improved service times. The detection of related structural concept drifts can be improved by early warning, as operation times of critical parts often increase before they catastrophically fail. • Utilization of event interim times for process analysis. • Adaptation of a scalable and noise adapting trend detection method from text mining. • Anytime drift detection in events streams for operational support. • Visualization for supervised domain expert interaction. • Application of Tesseract in the remaining time prediction. [ABSTRACT FROM AUTHOR]

The aim of process discovery, originating from the area of process mining, is to discover a process model based on business process execution data. A majority of process discovery techniques relies on an event log as an input. An event log is a static source of historical data capturing the execution of a business process. In this paper, we focus on process discovery relying on online streams of business process execution events. Learning process models from event streams poses both challenges and opportunities, i.e. we need to handle unlimited amounts of data using finite memory and, preferably, constant time. We propose a generic architecture that allows for adopting several classes of existing process discovery techniques in context of event streams. Moreover, we provide several instantiations of the architecture, accompanied by implementations in the process mining toolkit ProM (). Using these instantiations, we evaluate several dimensions of stream-based process discovery. The evaluation shows that the proposed architecture allows us to lift process discovery to the streaming domain. [ABSTRACT FROM AUTHOR]

This paper introduces the Hawkes skeleton and the Hawkes graph. These objects summarize the branching structure of a multivariate Hawkes point process in a compact, yet meaningful way. We demonstrate how the graph-theoretic vocabulary (ancestor sets, parent sets, connectivity, walks, walk weights, etc.) is very convenient for the discussion of multivariate Hawkes processes. For example, we reformulate the classic eigenvalue-based subcriticality criterion of multitype branching processes in graph terms. Next to these more terminological contributions, we show how the graph view can be used for the specification and estimation of Hawkes models from large, multitype event streams. Based on earlier work, we give a nonparametric statistical procedure to estimate the Hawkes skeleton and the Hawkes graph from data. We show how the graph estimation can then be used for specifying and fitting parametric Hawkes models. Our estimation method avoids the a priori assumptions on the model from a straightforward MLE-approach and is numerically more flexible than the latter. Our method has two tuning parameters: one controlling numerical complexity, and the other controlling the sparseness of the estimated graph. A simulation study confirms that the presented procedure works as desired. We pay special attention to computational issues in the implementation. This makes our results applicable to high-dimensional event-stream data such as dozens of event streams and thousands of events per component. [ABSTRACT FROM AUTHOR]

Process mining is a sub-field of data mining that focuses on analyzing timestamped and partially ordered data. This type of data is commonly called event logs. Each event is required to have at least three attributes: case ID, task ID/name, and timestamp to apply process mining techniques. Thus, any missing information need to be supplied first. Traditionally, events collected from different sources are manually correlated. While this might be acceptable in an offline setting, this is infeasible in an online setting. Recently, several use cases have emerged that call for applying process mining in an online setting. In such scenarios, a stream of high-speed and high-volume events continuously flow, e.g. IoT applications, with stringent latency requirements to have insights about the ongoing process. Thus, event correlation must be automated and occur as the data is being received. We introduce an approach that correlates unlabeled events received on a stream. Given a set of start activities, our approach correlates unlabeled events to a case identifier. Our approach is probabilistic. That implies a single uncorrelated event can be assigned to zero or more case identifiers with different probabilities. Moreover, our approach is flexible. That is, the user can supply domain knowledge in the form of constraints that reduce the correlation space. This knowledge can be supplied while the application is running. We realize our approach using complex event processing (CEP) technologies. We implemented a prototype on top of Esper, a state of the art industrial CEP engine. We compare our approach to baseline approaches. The experimental evaluation shows that our approach outperforms the throughput and latency of the baseline approaches. It also shows that using real-life logs, the accuracy of our approach can compete with the baseline approaches. • Online event correlation for low-level process events that leverage CEP technology. • A flexible approach using CEP rules can be added, removed, or modified at runtime. • Realization on top of Esper engine (a state-of-the-art industrial CEP engine). • Evaluation on real-life and synthetic logs and comparison with baseline approaches. • Achieving higher throughput, accuracy, and latency compared to the state-of-the-art. [ABSTRACT FROM AUTHOR]

Information about social networks can often be collected as event stream data. However, most methods in social network analysis are defined for static network snapshots or for panel data. We propose an actor oriented Markov process framework to analyze the structural dynamics in event streams. Estimated parameters are similar to what is known from exponential random graph models or stochastic actor oriented models as implemented in SIENA. We apply the methodology on a question and answer web community and show how the relevance of different kinds of one- and two-mode network structures can be tested using a new software. [Copyright &y& Elsevier]

This article focuses on data aggregation in vehicular ad hoc networks. In such networks, sensor data are usually produced and exchanged between vehicles in order to warn or inform the drivers when an event is detected (e.g., accident, emergency braking, parking space released, vehicle with non-functioning brake lights, etc.). In the following, we present a solution to aggregate and store these data in order to have a history of past events. We therefore use Flajolet-Martin sketches. Our goal is to generate additional knowledge to assist drivers by providing them useful information even if no event is transmitted by vehicles in the vicinity. [ABSTRACT FROM AUTHOR]

Frequent episode mining is a wide range framework of data mining from sequential data with many applications, which is a totally short-ordered collection of event-types and unearths temporal correlations without information loss over event streams. While offering substantial benefits, directly releasing frequent episodes to the public will enormously threaten the individual's privacy. However, there is little work so far concentrating on privately frequent episode mining. In this paper, we investigate the privacy problem in mining frequent episodes from event streams due to continuous releases in successive windows and propose a real-time differentially private frequent episode mining algorithm over event streams to avoid the privacy leakage with ω -event privacy guarantee. To obtain private frequent episodes, we propose a sample-based perturbation approach, which improves the accuracy of selecting frequent episodes based on sampling databases. To reduce the privately mining time and avoid repeatedly privacy budget allocation to coincident window of adjacent releases as much as possible, we present an incremental perturbation approach according to the judgment in dissimilarity calculation mechanism. Meanwhile, in order to protect data collected from any ω successive timestamps over event streams, we employ an adaptive ω -event privacy mechanism on the basis of the dynamicity of episodes. Finally, experimental results on real-world datasets demonstrate the effectiveness and efficiency of our algorithm. [ABSTRACT FROM AUTHOR]

Advances in clock synchronization techniques allow an approximated global time in ubiquitous environments. This paper presents an event stream-based online algorithm that fuses the data reported from the processors in such a network to detect time-based predicates. The algorithm has low space, time, and message complexities. The paper also considers the detection of simultaneous events as a special case. [ABSTRACT FROM AUTHOR]

This paper presents event stream-based online algorithms that fuse the data reported from processes to detect causality-based predicates of interest. The proposed algorithms have the following features. 1) The algorithms are based on logical time, which is useful to detect "cause and effect" relationships in an execution. 2) The algorithms detect properties that can be specified using predicates under a rich palette of time modalities. Specifically, for a conjunctive predicate ø, the algorithms can detect the exact fine-grained time modalities between each pair of intervals, one interval at each process, with low space, time, and message complexities. The main idea used to design the algorithms is that any "cause and effect" interaction can be decomposed as a collection of interactions between pairs of system components. The detection algorithms, which leverage the pairwise interaction among the processes, incur a low overhead and are, hence, highly scalable. The paper then shows how the algorithms can deal with mobility in mobile ad hoc networks. [ABSTRACT FROM AUTHOR]

Event log anomaly detection aims at identifying anomalous information in the logs generated by the execution of business processes. While several techniques for detecting trace-level anomalies in event logs in offline settings, i.e., when event logs are processed as a batch, have appeared recently in the literature, such techniques are currently lacking for online settings, i.e., when events are processed as a stream. Event log anomaly detection in online settings can be crucial for discovering anomalies in process execution as soon as they occur and, consequently, allowing to take early corrective actions. Moreover, it is also crucial for creating models that can adapt to concept drift in the process generating the events. This paper describes a novel approach to event log anomaly detection in process event streams: we define a general framework in which different anomaly detection methods can be plugged in and we propose and evaluate our own method based on statistical leverage. The leverage is an information-theoretic measure that has been used extensively in statistics to identify outliers and it has been adapted in this paper to the specific scenario of event streams. The proposed approach has been evaluated on artificial and real event streams and also on artificial event streams characterised by concept drift. • A general framework for online anomaly detection. • Adaptation of an information-theoretic measure (leverage) to online anomaly detection. • Extensive evaluation on real-life and artificial event logs. • Additional evaluation and discussion with event logs characterised by concept drift. [ABSTRACT FROM AUTHOR]

In this letter, we present an event stream processing system that can evaluate a pattern query for a data sequence with predicates. We propose a pattern query language and develop a pattern query processing system. In our system, we propose novel techniques for run-time aggregation and negation processing and apply our system to stream data generated from vehicles to monitor unusual driving patterns. [ABSTRACT FROM AUTHOR]

Process mining aims at gaining insights into business processes by analyzing the event data that is generated and recorded during process execution. The vast majority of existing process mining techniques works offline, i.e. using static, historical data, stored in event logs. Recently, the notion of online process mining has emerged, in which techniques are applied on live event streams, i.e. as the process executions unfold. Analyzing event streams allows us to gain instant insights into business processes. However, most online process mining techniques assume the input stream to be completely free of noise and other anomalous behavior. Hence, applying these techniques to real data leads to results of inferior quality. In this paper, we propose an event processor that enables us to filter out infrequent behavior from live event streams. Our experiments show that we are able to effectively filter out events from the input stream and, as such, improve online process mining results. • A new online filter is presented, which filters anomalies from event streams. • The filter uses an evolving ensemble of probabilistic non-deterministic automata. • Experiments show that filter parameterization leads to different accuracy levels. • Additionally, the paper presents a noise-oriented taxonomy of event data. [ABSTRACT FROM AUTHOR]

Critical systems that produce big data streams can require human operators to monitor these event streams for changes of interest. Automated systems which oversee many tasks can still have a need for the 'human-in-the-loop' operator to evaluate whether an intervention is required due to a lack of suitable training data initially offered to the system which would allow a correct course of actions to be taken. In order for an operator to be capable of reacting to real-time events, the visual depiction of the event data must be in a form which captures essential associations and is readily understood by visual inspection. A similar requirement can be found during inspections on activity protocols in a large organization where a code of correct conduct is prescribed and there is a need to oversee whether the activity traces match the expectations, with minimal delay. The methodology presented here addresses these concerns by providing an adaptive window sizing measurement for subsetting the data, and subsequently produces a set of network diagrams based upon event label co-occurrence networks. With an intuitive method of network construction the amount of time required for operators to learn how to monitor complex event streams of big datasets can be reduced. [ABSTRACT FROM AUTHOR]