Usually, ICDATA hosts several tutorials and invited talks focused on data science topics, such as Conformal Prediction, Big Data Analytics etc.
As soon as tutorials/talks are negotiated with speakers and confirmed, they will be published here.
All workshops, talks, tutorials etc. at CSCE are free for all CSCE attendees.
In 2022, the following events under the umbrella of ICDATA are planned (as of May 2022; maybe others will be added later):
Invited Talk 1
Professor, Computer & Information Science Department, Fordham University, USA
|Topic/Title||Data Science in Education: Extracting Knowledge from Admissions and Course Enrollment Data|
|Date & Time||July 25, 2022 |
01:40pm – 02:40pm
|Description||Data science has made great inroads into many areas of our daily life. However, applications to education have lagged most other fields, although this is beginning to change. In this talk I will first provide an overview of how data science can aid in education, and then describe the educational data mining research that I have been involved in that utilizes undergraduate student course-grade data and graduate admissions data. I will show how this data can be used to perform descriptive data mining to better understand how students sequence their courses, and how courses can be grouped based on student course co-enrollments or similar patterns in student performance. This data can also be used to identify strong and weak instructors based on the performance of their students in future courses, quantify the impact of different course sequences on student performance, and predict student grade performance based on admissions data.|
|Short Bio||Dr. Gary M. Weiss is a Professor of Computer and Information Science at Fordham University, and serves as the program director for the MS program in Computer Science. He received his B.S. degree in Computer Science from Cornell University, M.S. degree in Computer Science from Stanford University, and doctorate in Computer Science from Rutgers University. He joined Fordham in 2004 after working for nearly two decades at AT&T Bell Labs. His main research area is data mining and machine learning. His research focused on learning in the presence of class imbalance until 2010, at which point he founded the Wireless Sensor Data Mining Lab and focused on using smartphone and smartwatch sensor data to perform activity recognition and behavioral biometrics. Since 2019 he has been working with two of his colleagues in the area of educational data mining, with the goal of improving higher education with a data-intensive approach. Dr. Weiss has published over ninety papers and book chapters in the area of data science. He is currently Co-PI on a $4M National Science Foundation scholarship grant to educate the next generation of cybersecurity practitioners, which involves incorporating data science training into the cybersecurity curriculum.|
Invited Talk 2
Booz Allen Hamilton
|Topic/Title||"Developing Effective Extremism Detection Systems at Scale"|
|Date & Time||July 26|
06:20pm – 07:20pm
|Location||Galleria Ballrooms D & E|
|Description||The threat of online extremism continues to grab mainstream media headlines. Despite increased resources on enhancing trust and safety of social media networks, extremist material continues to proliferate. Concurrently, extremist groups have adopted privacy-preserving technologies such as the darknet as a means of maintaining a stable presence for the purposes of recruitment, propaganda dissemination, training, and the planning and execution of violence. Artificially intelligent systems can enable performant countermeasures but developing and deploying such systems is incredibly challenging. Namely, such systems must be able to operate in multilingual contexts and be resistant to common countermeasures employed by extremists to evade word filters and other common moderation and investigation techniques. Moreover, such systems need to maintain a low level of false positives to be useful in high-velocity deployments such as social media networks or investigative tools. In this talk, we begin by defining extremist online activity and how it differs from other potentially unwanted content such as hate speech or profanity. With an understanding of the problem domain, we will review state-of-the-art natural language model architectures and identify techniques to extend them to operate more effectively against extremist texts. Finally, we will discuss some ideas to improve existing work to address the challenges posed by extremist texts, such as alternative embedding strategies, zero-shot classification, and dataset development methodologies.|
|Short Bio||Andrew Johnston is a Lead Technologist with Booz Allen Hamilton providing proactive cybersecurity consulting. In addition to his work with Booz Allen, Andrew is an Adjunct Professor with Fordham University's Graduate School of Arts and Sciences. Andrew Johnston's primary area of research is in the application of deep learning to problems within the cyber and national security spaces. Prior to his work with Booz Allen, Andrew was a proactive cybersecurity consultant with Mandiant serving both commercial and government customers. Andrew Johnston formerly worked with the FBI in the Cyber and Counterterrorism divisions.|