Pseudo Feature Extraction and the Prospect of BCI modeling
Presenter: Ibrahim Almosallam, Computer Research Institute, King Abdulaziz City for Science and Technology
 
Abstract: Machine learning and pattern recognition have been proven to be very useful in a number of commercial applications.  From predicting users’ behavior, recommending products or services to providing advanced analytical tools or visualizations to aid decision-making. More often than not, data collecting and processing is more important than learning. Unlike learning however, data collecting is often time-consuming, costly and domain dependant. For example collecting data about books involves recording information such as author(s), publisher, genre … etc. Similarly in movies (director, producer, actors and genre) and in text (language, part-of-speech and tense) and so on. As demonstrated, in some cases an expert knowledge is even required. In this seminar an overview of  “Pseudo Feature Extraction” will be given which is a universal learning algorithm aimed at revealing the most relevant features irrespective of the data domain, not from painstaking data collecting, labeling or experts’ opinions but rather from the abundance data on user-item interaction with the data. However, using such data has its own set backs because humans are often inconsistent in their interaction. A research proposal will be given in how to augment the user-item interaction data with BCI sensory data to better account for human inconsistency that are due to several psychological factors.
 
 
Bio: Ibrahim Almosallam is a researcher from the Computer Research Institute (CRI) at King Abdulaziz City for Science and Technology (KACST). Mr. Almosallam has a Master of Science in Computer Science from the University of Missouri-Columbia and a Bachelor of Science in Computer Science from King Fahd University of Petroleum and Minerals. His Masters’ Thesis was in Machine learning and Data Mining and it has been his research focus since then. He has been working at KACST for three years in problems such as Automatic Speech Recognition, Text-to-Speech, Social Network Analysis and Natural Language Processing. Mr. Almosallam also collaborated with the machine-learning group at Stanford University as a visiting researcher (2011-2012) under the supervision of Dr. Andrew Ng. Mr. Almosallam has delivered several tutorials
and short-courses in machine learning and social network analysis.
 
Date: April 15, 2013
Time: 1:00 - 2:00 pm
Location: Room 117, Building 20 for Skerg members; Broadcasting to B20, room 37 for students