Distributed Learning Group
Due to the data explosion in the modern age of information technology, there exist enough data to learn complicated models such as deep neural networks and probabilistic models. However, due to the storage and computational expenses, the data cannot be stored and processed in a single node. Communication restrictions and exponential time complexity of learning algorithms are the main challenges for centralized learning. Distributed learning aims to solve these challenges. The learning problem usually can be represented as a statistical or optimization problem defined on a set of distributed data. Proposed methods include approximating the objective function, parallelizing sequential algorithms and so forth. Our research in this group mainly focuses on distributed learning, especially on learning with probabilistic models. Density estimation, parallel estimation of the parameters of Markov random fields and distributed Markov Chain Monte-Carlo methods are among the problems we consider.
Brain Signal Processing Team
The brain is one of the biggest and oldest unsolved puzzles for human beings and there are too many unanswered questions about it. According to historians, the earliest recorded reference to the brain is “The Edwin Smith Surgical Papyrus”. This papyrus was written in the 17th century BC and describes the symptoms, diagnosis, and prognosis of two patients. From that time up to now, studying about the brain has evolved into a popular interdisciplinary area which contains diverse disciplines of neuroscience, cognitive science, psychology, computer science, mathematics, physics, philosophy and etc. Here, the Brain Signal Processing Team is recently established to join the growing trend of brain studies. In this team, as computer scientists, we are dealing with brain modeling and information processing. Currently, the team’s focus is on attention modeling and developing methods to cluster and classify EEG signals to improve brain-computer interfaces performance.
Information retrieval is the activity of obtaining information relevant to user’s query, Searches can be based on full-text or other content-based indexing. Information retrieval is used to reduce information overload and find the best result in a domain. One the best practices in achieving these goals are specified searches. In this lab expert retrieval, author retrieval, opinion mining is being considered specifically, and with user behavior analytic information retrieval is being personalized to increase user satisfaction. This lab is researching and promoting the use of personalized search in specific domain retrievals.
Intelligence System Laboratory (Room 505) Computer Engineering Faculty, Sharif University of Technology, Azadi Ave., Tehran, Iran Phone Number: +982166166674