BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Events//NONSGML v1.0//EN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTART;TZID="Pacific Time (US & Canada)":20201112T100000
DTEND;TZID="Pacific Time (US & Canada)":20201112T120000
SUMMARY:Amirhossein Sayyafan – Preliminary Exam
LOCATION:Online
DESCRIPTION:Advisor:  Dr. Ben Belzer & Dr. Krishnamoorthy Sivakumar\n\nDegree: Ph.D. Electrical & Computer Engineering\n\nAbstract: Recording in multiple dimensions on magnetic hard drives has been a challenge in the hard disk drive industry. My research focus has been on developing some algorithms and approaches to more accurately read data from one and two-dimensional hard drives, which are densely packed with information. The objective of reading approaches for magnetic recording is to detect the highest density of information possible with an acceptable error rate. These approaches will help us to have fast and accurate data reading of hard drives and increase the number of bits per square inch. We utilize some methods in signal processing, channel decoding, and machine learning to cancel out the undesired media noise. The machine learning techniques being used are deep neural networks (DNNs) that have led to great success in applications such as image understanding. So far, we have developed a DNN media noise predictor turbo detection system for one and two dimensional magnetic recordings to enhance the efficiency of hard drives and achieve the highest areal density among the published figures for magnetic media with grain densities of 11.4 Teragrains/in^2.\n\nDespite the great success of DNNs in various applications, they have massive computational complexity. We are pursuing to optimize the deep neural network performance by using information-theoretic methods. Currently, we are working on optimizing the information bottleneck between the layers of DNNs to drive an improved training algorithm which is superior in terms of computational resource allocation and performance.
BEGIN:VALARM
ACTION:DISPLAY
DESCRIPTION:REMINDER
TRIGGER;RELATED=START:-PT00H15M00S
END:VALARM
END:VEVENT
END:VCALENDAR