October 2017 - Dm-meeting - mailman.cs.vt.edu

Thursday, Nov. 2nd, @ 11:00 am, Torg 3160 A, Liangzhe Chen
by Sorour Ekhtiari Amiri 31 Oct '17

31 Oct '17

Hi everyone, Please join us for our next DM-Meeting. ****DM-Meeting Fall 2017**** WHO: Liangzhe Chen WHEN: Thursday, Nov. 2nd, 2017, @ 11:00 am WHERE: Torg 3160 A WHAT: Liangzhe will give a prep talk for his CIKM paper. Title: HotSpots: Failure Cascades on Heterogeneous Critical Infrastructure Networks Abstract: Critical Infrastructure Systems such as transportation, water and power grid systems are vital to our national security, economy, and public safety. Recent events, like the 2012 hurricane Sandy, show how the interdependencies among different CI networks lead to catastrophic failures among the whole system. Hence, analyzing these CI networks, and modeling failure cascades on them becomes a very important problem. However, traditional models either do not take multiple CIs or the dynamics of the system into account, or model it simplistically. In this paper, we study this problem using a heterogeneous network viewpoint. We first construct heterogeneous CI networks with multiple components using national level datasets. Then we study novel failure maximization problems on these networks, to compute critical nodes in such systems. We then provide HotSpots, a scalable and effective algorithm for these problems, based on careful transformations. Finally, we conduct extensive experiments on real CIS data from multiple US states, and show that our method HotSpots outperforms non-trivial baselines, gives meaningful results and that our approach gives immediate benefits in providing situational-awareness during large-scale failures. Best regards, Sorour

1 0

Thursday, Oct. 26, @ 11:00 am, Torg 3160 A, Bijaya Adhikari
by Sorour Ekhtiari Amiri 23 Oct '17

23 Oct '17

Hi everyone, Please join us for our next DM-Meeting. ****DM-Meeting Fall 2017**** WHO: Bijaya Adhikari WHEN: Thursday, Oct. 26th, 2017, @ 11:00 am WHERE: Torg 3160 A WHAT: Bijaya will give a prep talk for his ICDM Workshop paper. Title: Distributed Representations of Subgraphs Abstract: There has been a surge in research interest in learning feature representation of networks in recent times. Researchers, motivated by the recent successes of embeddings in natural language processing and advances in deep learning, have explored various means for network embedding. Network embedding is useful as it can exploit off-the-shelf machine learning algorithms for network mining tasks like node classification and link prediction. However, most recent works focus on learning feature representation of nodes, which are ill-suited to tasks such as community detection which are intuitively dependent on subgraphs. In this work, we formulate a novel subgraph embedding problem based on an intuitive property of subgraphs and propose SubVec, an unsupervised scalable algorithm to learn feature representations of arbitrary subgraphs. We demonstrate usability of features learned by SubVec by leveraging them for community detection problem, where SubVec significantly outperforms nontrivial baselines. We also conduct case-studies in two distinct domains to demonstrate wide applicability of SubVec. Best regards, Sorour

1 0

Thursday, Oct. 19, @ 11:00 am, Torg 3160 A, Liangzhe Chen
by Sorour Ekhtiari Amiri 16 Oct '17

16 Oct '17

Hi everyone, Please join us for our next DM-Meeting. ****DM-Meeting Fall 2017**** WHO: Liangzhe Chen WHEN: Thursday, Oct. 19th, 2017, @ 11:00 am WHERE: Torg 3160 A WHAT: Liangzhe will talk about the following two papers, Title: Analyzing Invariants in Cyber-Physical Systems using Latent Factor Regression Abstract: The analysis of large-scale data logged from complex cyber-physical systems, such as microgrids, often entails the discovery of invariants capturing functional as well as operational relationships underlying such large systems. We describe a latent factor approach to infer invariants underlying system variables and how we can leverage these relationships to monitor a cyber-physical system. In particular, we illustrate how this approach helps rapidly identify outliers during system operation. Title: Unifying Dependent Clustering and Disparate Clustering for Non-homogeneous Data Abstract: Modern data mining settings involve a combination of attribute valued descriptors over entities as well as specified relationships between these entities. We present an approach to cluster such non-homogeneous datasets by using the relationships to impose either dependent clustering or disparate clustering constraints. Unlike prior work that views constraints as boolean criteria, we present a formulation that allows constraints to be satisfied or violated in a smooth manner. This enables us to achieve dependent clustering and disparate clustering using the same optimization framework by merely maximizing versus minimizing the objective function. We present results on both synthetic data as well as several real-world datasets Best regards, Sorour

1 0

Fwd: [Urbcomp-faculty] UrbComp Seminar Series featuring Mohamed Sarwat - Oct. 24 at 2:30 - 3:30 pm
by B. Aditya Prakash 16 Oct '17

16 Oct '17

FYI. Sounds interesting. Folks should try attending it. cheers, - Aditya -------- Forwarded Message -------- Subject: [Urbcomp-faculty] UrbComp Seminar Series featuring Mohamed Sarwat - Oct. 24 at 2:30 - 3:30 pm Date: Mon, 16 Oct 2017 18:55:14 +0000 From: Urbcomp-faculty via Hawrami, Wanawsha <urbcomp-faculty(a)cs.vt.edu> Reply-To: Hawrami, Wanawsha <wanah92(a)vt.edu> To: urbcomp-faculty(a)cs.vt.edu <urbcomp-faculty(a)cs.vt.edu> Dear Students and Colleagues, Please join us for the UrbComp Seminar Series featuring Mohamed Sarwat, Assistant Professor in the in the Department of Computer Science at Arizona State University on *Tuesday, Oct. 24 at 2:30 - 3:30 pm* in room 3-024 of the Virginia Tech Research Center - Arlington and Kelly Hall room 310. We look forward to seeing you there!

1 0

Thursday, Oct. 12, @ 11:00 am, Torg 3160 A, Sorour E. Amiri
by Sorour Ekhtiari Amiri 11 Oct '17

11 Oct '17

Hi everyone, Please join us for our next DM-Meeting. ****DM-Meeting Fall 2017**** WHO: Sorour E. Amiri WHEN: Thursday, Oct. 12th, 2017, @ 11:00 am WHERE: Torg 3160 A WHAT: I will talk about the following two papers, Title: Discriminative Embeddings of Latent Variable Models for Structured Data Abstract: Kernel classifiers and regressors designed for structured data, such as sequences, trees and graphs, have significantly advanced a number of interdisciplinary areas such as computational biology and drug design. Typically, kernels are designed beforehand for a data type which either exploit statistics of the structures or make use of probabilistic generative models, and then a discriminative classifier is learned based on the kernels via convex optimization. However, such an elegant two-stage approach also limited kernel methods from scaling up to millions of data points, and exploiting discriminative information to learn feature representations. We propose, structure2vec, an effective and scalable approach for structured data representation based on the idea of embedding latent variable models into feature spaces, and learning such feature spaces using discriminative information. Interestingly, structure2vec extracts features by performing a sequence of function mappings in a way similar to graphical model inference procedures, such as mean field and belief propagation. In applications involving millions of data points, we showed that structure2vec runs 2 times faster, produces models which are 10, 000 times smaller, while at the same time achieving the state-of-the-art predictive performance. Title: Learning Combinatorial Optimization Algorithms over Graphs Abstract: The design of good heuristics or approximation algorithms for NP-hard combinatorial optimization problems often requires significant specialized knowledge and trial-and-error. Can we automate this challenging, tedious process, and learn the algorithms instead? In many real-world applications, it is typically the case that the same optimization problem is solved again and again on a regular basis, maintaining the same problem structure but differing in the data. This provides an opportunity for learning heuristic algorithms that exploit the structure of such recurring problems. In this paper, we propose a unique combination of reinforcement learning and graph embedding to address this challenge. The learned greedy policy behaves like a meta-algorithm that incrementally constructs a solution, and the action is determined by the output of a graph embedding network capturing the current state of the solution. We show our framework can be applied to a diverse range of optimization problems over graphs, and learns effective algorithms for the Minimum Vertex Cover, Maximum Cut and Traveling Salesman problems. Best regards, Sorour

1 0

Thursday, Oct. 5, @ 11:00 am, Torg 3160 A, Xinfeng Xu
by Sorour Ekhtiari Amiri 04 Oct '17

04 Oct '17

Hi everyone, Please join us for our next DM-Meeting. ****DM-Meeting Fall 2017**** WHO: Xinfeng Xu WHEN: Thursday, Oct. 5th, 2017, @ 11:00 am WHERE: Torg 3160 A WHAT: Xinfeng will talk about the following two papers Title: Learning Tree-Structured Detection Cascades for Heterogeneous Networks of Embedded Devices Abstract: In this paper, we present a new approach to learning cascaded classifiers for use in computing environments that involve networks of heterogeneous and resource-constrained, low-power embedded compute and sensing nodes. We present a generalization of the classical linear detection cascade to the case of tree-structured cascades where different branches of the tree execute on different physical compute nodes in the network. Different nodes have access to different features, as well as access to potentially different computation and energy resources. We concentrate on the problem of jointly learning the parameters for all of the classfiers in the cascade given a fixed cascade architecture and a known set of costs required to carry out the computation at each node. To accomplish the objective of joint learning of all detectors, we propose a novel approach to combining classier outputs during training that better matches the hard cascade setting in which the learned system will be deployed. This work is motivated by research in the area of mobile health where energy effcient real time detectors integrating information from multiple wireless on-body sensors and a smart phone are needed for real-time monitoring and the delivery of just-in-time adaptive interventions. We evaluate our framework on mobile sensor-based human activity recognition and mobile health detector learning problems. Title: Deep Embedding Forest: Forest-based Serving with Deep Embedding Features Abstract: Deep Neural Networks (DNN) have demonstrated superior ability to extract high level embedding vectors from low level features. Despite the success, the serving time is still the bo.leneck due to expensive run-time computation of multiple layers of dense matrices. GPGPU, FPGA, or ASIC-based serving systems require additional hardware that are not in the mainstream design of most commercial applications. In contrast, tree or forest-based models are widely adopted because of low serving cost, but heavily depend on carefully engineered features. .is work proposes a Deep Embedding Forest model that bene.ts from the best of both worlds. .e model consists of a number of embedding layers and a forest/tree layer. .e former maps high dimensional (hundreds of thousands to millions) and heterogeneous low-level features to the lower dimensional (thousands) vectors, and the la.er ensures fast serving. Built on top of a representative DNN model called Deep Crossing [21], and two forest/tree-based models including XGBoost and LightGBM, a two-step Deep Embedding Forest algorithm is demonstrated to achieve on-par or slightly be.er performance as compared with the DNN counterpart, with only a fraction of serving time on conventional hardware. A.er comparing with a joint optimization algorithm called partial fuzzification, also proposed in this paper, it is concluded that the two-step Deep Embedding Forest has achieved near optimal performance. Experiments based on large scale data sets (up to 1 billion samples) from a major sponsored search engine proves the efficacy of the proposed model. Best regards, Sorour

1 0