2017 IEEE North Jersey Advanced Communications Symposium

#IEEE #North #Jersey #Advanced #Communications #Symposium #COMSOC #IT #VTS #Computer #5G #AI #Machine #Learning #Deep #Big #Data #Security.

The 2017 IEEE North Jersey Advanced Communications Symposium (NJACS-2017) will be held at the Babbio Center, Stevens Institute of Technology, in Hoboken, NJ, on Saturday, September 23, 2017. The symposium consists of several keynote presentations and a parallel poster session. The symposium program will cover advanced topics in AI, big data, machine learning, deep learning, and applications. The posters will be presented by graduate students and postdocs. Poster presentations will be on display all day and special dedicated exhibition times are scheduled for all attendees. There will be plenty of opportunities to interact with presenters and network with peers.  For all questions about poster presentations, poster cash prizes, poster registration, deadlines, contact Prof. Hong Zhao ( zhao@fdu.edu).

Symposium Program

Registration

This event has limited seating and registration will close once event reaches capacity.

CEUs (continuing education units) are available for this event as a separate fee of $9 payable at registration desk.

• Date: 23 Sep 2017
• Time: 10:00 AM to 06:00 PM
• All times are (GMT-05:00) US/Eastern
• Add Event to Calendar
  iCal
  Google Calendar

• Stevens Institute of Technology
• 525 River Street
• Hoboken, New Jersey
• United States 07030
• Building: Babbio Center
• Click here for Map

---

• Contact Event Host

• For more information, please contact: Yu-Dong Yao, Symposium Chair, yyao@stevens.edu

  Adriaan van Wijngaarden,  Program Chair, avw@ieee.org

   

• Starts 12 August 2017 01:00 AM
• Ends 23 September 2017 06:00 AM
• All times are (GMT-05:00) US/Eastern
• Admission fee ?

• 

Deep Learning: Autoencoder  - Prof. Rensheng Wang, Stevens

Abstract  - Autoencoders play a fundamental role in unsupervised learning and in deep architectures for transfer learning and other tasks.  Here we introduce a general framework of autoencoders in different applications, including the discussions about the learning complexity, their critical points and some variations.

Rensheng Wang is an adjunct professor with Stevens.  His research interests lie in the areas of machine learning, data mining,  big-data processing platform, predictive modeling and recommender systems.   He obtained his Ph.D degree from the Electrical & Computer Engineering department in Stevens Institute of Technology.

Differential Privacy Preservation in Deep Learning - Prof. Hai Phan, NJIT

Abstract - In recent years, advances in deep learning have enabled a dizzying array of applications such as data analytics, signal and information processing, and autonomous systems. This presents an obvious threat to privacy in new deep learning systems and models, which are being developed and deployed. In this talk, I will review the current picture of security and privacy in deep learning. Then, I will introduce our approaches to preserve differential privacy in deep learning and to uncover vulnerabilities of differentially private deep neural networks. Future directions in privacy and security in deep learning will be discussed as well.

Hai Phan is an Assistant Professor at the Ying Wu College of Computing, New Jersey Institute of Technology. He holds a Ph.D. in computer science, awarded in 2013 by the French National Centre for Scientific Research (CNRS) - University Montpellier 2 (UM2). His topics of interest primarily focus on data science, machine learning, deep learning, and privacy and security, especially for health informatics, social network analysis, and spatio-temporal data mining. His contributions have been published in at least 30 publications at leading and top-tier venues, including AAAI, ICDM, ACM Multimedia, Machine Learning Journal, ACM CIKM, and IEEE Intelligent Systems.

Convolutional Neural Networks for Figure Extraction in Historical Bell Labs Records - Dr. Chun-Nam Yu, Nokia Bell Labs

Abstract - We present a novel method of extracting figures and images from pages in scanned documents, especially from historical technical documents. Our approach is based on convolutional neural networks, which have been successfully applied to many computer vision problems. One major challenge in training neural networks is obtaining large amount of labeled data. We show that we can automatically generate training data for our convolutional neural networks by using PDF figure extractors applied to modern technical publications, and the learned models transfer very well to the problem of extracting figures in historical technical documents. On a test set consisting of modern journal papers and conference proceedings, our convolutional neural networks achieve precision and recall close to 85% in identifying figures. On the test set of historical technical documents from the Bell Labs Records, our models obtain precision and recall above 80% in identifying figures. This is joint work with Caleb Levy and Iraj Saniee.

Chun-Nam Yu is a Member of Technical Staff at at Bell Labs. He received his BA degree in Mathematics and Computer Science from Oxford University in 2004, the MS degree in Computer Science in 2008, and the PhD degree in Computer Science in 2010 from Cornell University. He was a postdoctoral fellow at the Alberta Innovates Centre of Machine Learning (AICML) at the University of Alberta from 2010 to 2012. His research interests include structured output learning, graphical models, kernel methods, optimization, and biomedical applications. His work at Bell Labs focus on smart grid load forecasting, telecom customer experience modeling, streaming data processing, and deep learning.

Adversarial Perturbations in Deep Neural Networks: Attack and Defense - Mr. Huaxia Wang, Stevens/Nokia Bell Labs

Abstract - Deep neural networks have been shown to perform well on many classical machine learning problems, especially in image classification task. However, researchers have found that neural networks are easily get fooled, they are surprisingly sensitive to small perturbations. Carefully crafted input images (adversarial examples) can force a well-trained neural network to provide adversary-selected outputs. This talk provides a review of recent work on the topic of adversarial examples in deep neural networks, including attack/defense models.

Huaxia Wang received the B.Eng. degree in information engineering from Southeast University, Nanjing, China, in 2012. He is currently pursuing the Ph.D. degree with the Electrical and Computer Engineering Department, Stevens Institute of Technology, Hoboken, NJ, USA. From 2016 to 2017, he was a Research Intern with the Mathematics of Networks and Systems Research Department, Nokia Bell Labs, Murray Hill, NJ, USA. His current research interests include wireless communications, cognitive radio networks, and deep learning.