A task-driven approach to time scale detection in dynamic networks
Author:
Benjamin Fish, Department of Mathematics, Statistics and Computer Science, University of Illinois at Chicago
Abstract:
For any stream of time-stamped edges that form a dynamic network, an important choice is the aggregation granularity that an analyst uses to bin the data. Picking such a windowing of the data is often done by hand, or left up to the technology that is collecting the data. However, the choice can make a big difference in the properties of the dynamic network. Finding a good windowing is the time scale detection problem. In previous work, this problem is often solved with an unsupervised heuristic. As an unsupervised problem, it is difficult to measure how well a given windowing algorithm performs. In addition, we show that there is little correlation between the quality of a windowing across different tasks. Therefore the time scale detection problem should not be handled independently from the rest of the analysis of the network. Given this, in accordance with standard supervised machine learning practices, we introduce new windowing algorithms that automatically adapt to the task the analyst wants to perform by treating windowing as a hyperparameter for the task, rather than using heuristics. This approach measures the quality of the windowing by how well a given task is accomplished on the resulting network. This also allows us, for the first time, to directly compare different windowing algorithms to each other, by comparing how well the task is accomplished using that windowing algorithm. We compare this approach to previous approaches and several baselines on real data.
More on http://www.kdd.org/kdd2017/
KDD2017 Conference is published on http://videolectures.net/
Видео A task-driven approach to time scale detection in dynamic networks канала KDD2017 video
Benjamin Fish, Department of Mathematics, Statistics and Computer Science, University of Illinois at Chicago
Abstract:
For any stream of time-stamped edges that form a dynamic network, an important choice is the aggregation granularity that an analyst uses to bin the data. Picking such a windowing of the data is often done by hand, or left up to the technology that is collecting the data. However, the choice can make a big difference in the properties of the dynamic network. Finding a good windowing is the time scale detection problem. In previous work, this problem is often solved with an unsupervised heuristic. As an unsupervised problem, it is difficult to measure how well a given windowing algorithm performs. In addition, we show that there is little correlation between the quality of a windowing across different tasks. Therefore the time scale detection problem should not be handled independently from the rest of the analysis of the network. Given this, in accordance with standard supervised machine learning practices, we introduce new windowing algorithms that automatically adapt to the task the analyst wants to perform by treating windowing as a hyperparameter for the task, rather than using heuristics. This approach measures the quality of the windowing by how well a given task is accomplished on the resulting network. This also allows us, for the first time, to directly compare different windowing algorithms to each other, by comparing how well the task is accomplished using that windowing algorithm. We compare this approach to previous approaches and several baselines on real data.
More on http://www.kdd.org/kdd2017/
KDD2017 Conference is published on http://videolectures.net/
Видео A task-driven approach to time scale detection in dynamic networks канала KDD2017 video
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Designing AI at Scale to Power Everyday LifeEstimation of Recent Ancestral Origins of Individuals on a Large Scalestruc2vec: Learning Node Representations from Structural IdentityLearning to Generate Rock Descriptions from Multivariate Well Logs with Hierarchical AttentionA Local Algorithm for StructurePreserving Graph CutEmbedJoin: Efficient Edit Similarity Joins via EmbeddingsMulti-Aspect Streaming Tensor CompletionPlanning Bike Lanes based on SharingBikes' TrajectoriesTripoles: A New Class of Relationships in Time Series DataIndustrial Machine LearningOptimal Reserve Price for Online Ads Trading Based on Inventory IdentificationMulti-Aspect Streaming Tensor CompletionDispatch with Confidence: Integration of machine learningLearning certifiably optimal rule lists for categorical dataInterpretable Predictions of Tree-based Ensembles via Actionable Feature TweakingHinDroid: An Intelligent Android Malware Detection SystemRevisiting power-law distributions in spectra of real world networksKDD Business LunchTraffic sensor health monitoring using spatiotemporal graphical modelingA Practical Algorithm for Solving the Incoherence Problem of Topic Models In Industrial ApplicationsDetecting Network Effects: Randomizing Over Randomized Experiments