DISSERTATION DEFENSE Department of Computer Science and Engineering, University of South Carolina Candidate: Shizhong Han Advisor: Dr. Yan Tong Abstract Recognizing facial action units (AUs) from spontaneous facial expression is a challenging problem, because of subtle facial appearance changes, free head movements, occlusions, and limited AU-coded training data. Most recently, convolutional neural networks (CNNs) have shown promise on facial AU recognition. However, CNNs are often overfitted and do not generalize well to unseen subject due to limited AU-coded training images. In order to improve the performance of facial AU recognition, we developed two novel CNN frameworks, by substituting the traditional decision layer and convolutional layer with the incremental boosting layer and adaptive convolutional layer respectively, to recognize the AUs from static image. First, in order to handle the limited AU-coded training data and reduce the overfitting, we proposed a novel Incremental Boosting CNN (IB-CNN) to integrate boosting into the CNN via an incremental boosting layer that selects discriminative neurons from the lower layer and is incrementally updated on successive mini-batches. In addition, a novel loss function that accounts for errors from both the incremental boosted classifier and individual weak classifiers was proposed to fine-tune the IB-CNN. Experimental results on four benchmark AU databases have demonstrated that the IB-CNN yields significant improvement over the traditional CNN and the boosting CNN without incremental learning, as well as outperforming the state-of-the-art CNN-based methods in AU recognition. The improvement is more impressive for the AUs that have the lowest frequencies in the databases. Second, all current CNNs use predefined and fixed convolutional filter size. However, AUs activated by different facial muscles cause facial appearance changes at different scales and thus favor different filter sizes. The traditional strategy is to experimentally select the best filter size for each AU in each convolutional layer, but it suffers from expensive training cost, especially when the networks become deeper and deeper. We proposed a novel Optimized Filter Size CNN (OFS-CNN), where the filter sizes and weights of all convolutional layers are learned simultaneously from the training data along with learning convolutional filters. Specifically, the filter size is defined as a continuous variable, which is optimized by minimizing the training loss. Experimental results on four AU-coded databases and one spontaneous facial expression database outperforms traditional CNNs with fixed filter sizes and achieves state-of-the-art recognition performance. Furthermore, the OFS-CNN also beats traditional CNNs using the best filter size obtained by exhaustive search and is capable of estimating optimal filter size for varying image resolution.

Csilla Farkas Professor, Department of Computer Science and Engineering University of South Carolina, Columbia Abstract: Are you living in a smart home? Are you using smart devices to monitor your health? Is your organization considering to increase automation for sensing and controlling operations? While the concept of Internet of Things (IoT) may mean different things to different people, there is a common theme: the need for cybersecurity. The key security challenges are focused on three areas: 1) device vulnerabilities, 2) communication security and trust, and 3) data integrity, security, and privacy. In this talk I present a semantics-based approach to support IoT data integration and security. Bio: Csilla Farkas is a Professor in the Department of Computer Science and Engineering and Director of the Center for Information Assurance Engineering at the University of South Carolina. Dr. Farkas’ research interests include information security, data inference problem, financial and legal analysis of cyber crime, and security and privacy on the Semantic Web. silla Farkas received her PhD from George Mason University, Fairfax. In her dissertation she studied the inference and aggregation problems in multilevel secure relational databases. She received a MS in computer science from George Mason University and BS degrees in computer science and geology from SZAMALK, Hungary and Eotvos Lorand University, Hungary, respectively.

This week at our Weekly ACM/Software Engineering Meeting, Jeremy Day is giving a talk about the History and Usage of vi. Food may be provided. More Information

Come join us for our first meeting of the semester in Amoco Hall! We'll be discussing events throughout the school year, voting on future meeting times, and watching a demonstration on how to turn your pets into mobile hacking machines. More Details

I would like to invite you to attend this week's CSCE 791 seminar. These seminars highlight research being performed in our department and across the world. All CSCE 791 seminars are open to anybody who wishes to attend - not just students registered for the course. Friday, September 1, 2:20 - 3:10 PM Swearingen room 2A14 Speaker: Dr. Duncan Buell, University of South Carolina Abstract: First year students at most large universities take required courses whose purpose is to teach them to write prose essays and make arguments. We have acquired more than 7000 pairs of draft-and-final essays from USC and have been analyzing them. We are not trying to do “machine grading” of essays as an AI project. Rather, we are trying to identify features of writing that can be quantified and thus processed with programs as a big-data analysis. We are interested in the extent to which students revise their draft essays to become final versions. And we are interested in comparing our student writing against other genres of writing. For this last we use the Corpus of Contemporary American English (COCA) as source data. The COCA is a corpus of more than 500 million words of text separated into genres of academic writing, magazine writing, transcripts of spoken English and interviews, and such. Our eventual goal is to situate student writing relative to other genres and thu s to help with improving the pedagogy of teaching writing; knowing what the students are actually writing now is key to knowing how to get them to write formal prose effectively. Programming is done in Python. Part of speech tagging is done using the CLAWS package from the University of Lancaster in the UK. Sentence parsing is done using the package from Dan Jurafsky’s lab at Stanford. Bio: Duncan A. Buell is a Professor in the Department of Computer Science and Engineering at the Unviversity of South Carolina. His Ph.D. is in mathematics from the University of Illinois at Chicago (1976). He was from 2000 to 2009 the department chair at USC, and in 2005-2006 was interim dean. He has done research in document retrieval, computational number theory, and parallel computing, and has more recently turned to digital humanities as one of the emerging “marketplace” applications for computing. He is engaged with First Year English at USC on the analysis of freshman English essays, searching for an understanding of actual student writing in an effort to improve pedagogy for first year English instruction. He has team taught four times with Dr. Heidi Rae Cooley on the presentation of unacknowledged history on mobile devices, and he and Dr. Cooley are actively engaged in ways to go beyond text to fully enable the use of visual media in mobile applications that present human ities content, especially content that might normally remain unacknowledged by institutional authority.

Welcome to join Women in Computing tonight. We welcome everyone - all genders and majors! The event will be held tonight in Faculty Lounge at Swearingen 1A03, from 7:00 - 8:30 pm. Today’s major agenda is to give an information session for Grace Hopper Conference. Department of Computer Science and Engineering will be sending a group of students to the 2017 Grace Hopper Celebration of Women in Computing which will take place on Wednesday, October 4 through Friday, October 6, in Orlando, Florida. http://ghc.anitaborg.org/. The deadline of application is tonight. We will talk about the application and selection process. The club has invited some previous attendees to attend the meeting and share their experience. More info about WiC.

Click here for more details.

Click here for more details.

It’s that time of year. Students enrolled in the cross-College Critical Interactives class (CSCE 571, FAMS 511, MART 591, FAMS 710, CSCE 790) have been developing the 2017 version of Ward One App, which they will present and demonstrate on Tuesday, April 18, 2017, at 4:30 - 5:30pm in Booker T. Washington Auditorium. We invite you join us for the event. The presentation and demonstration will be followed by Q&A.