Yesterday

1. TETAM PhD Seminars
   • Start time: 10:00am, Friday, February 22nd
   • End time: 11:30am, Friday, February 22nd
   • Where: AVS Conference Room, BM

   • Title: Exploring Temporal Accumulative Features for Sign Language Recognition
     Speaker: Ahmet Alp Kındıroğlu

     Title: The Determination of Optimum Radiation Treatment Plans
     Speaker: Pınar Dursun

     http://tetam.boun.edu.tr/sites/default/files/2019-02-22_TETAM_PhD_Seminars_vol-1.jpg

     http://tetam.boun.edu.tr/sites/default/files/tetam_doktora_seminerleri_hafta_2019_spring.jpg

   • View this event in Google Calendar

Tuesday, February 26th

1. CmpE Seminar: General Overview of Deep Learning And Its Applications in Self Driving Cars and Speech processing by Cevahir Parlak
   • Start time: 12:00pm, Tuesday, February 26th
   • End time: 01:00pm, Tuesday, February 26th
   • Where: AVS Conference Room, BM

   • Abstract:
     Major Architectures of Deep Networks
     There are three major architectures of deep networks and how we use the smaller networks to build them.
     Unsupervised Pretrained Networks (UPNs)
     o Autoencoders and Variational AutoEncoders
     o Deep Belief Networks (DBNs)
     o Generative Adversarial Networks (GANs)

     Convolutional Neural Networks (CNNs)
     The goal of a CNN is to learn higher-order features in the data via convolutions. They are well suited to object recognition with images and consistently top image classification competitions. They can identify faces, individuals, street signs, and many other aspects of visual data. CNNs overlap with text analysis via optical character recognition, but they are also useful when analyzing words as discrete textual units. They’re also good at analyzing sound.

     Recurrent Neural Networks (RNNs)
     Recurrent Neural Networks are in the family of feed-forward neural networks. They are different from other feed-forward networks in their ability to send information over time-steps. Historically, these networks have been difficult to train, but more recently, advances in research (optimization, network architectures, parallelism, and graphics processing units [GPUs]) have made them more approachable for the practitioner. LSTM networks are the new way to handle time series data such as speech.

     Androcars (Robocars, Self Driving Cars, Autonomous Cars) in recent decades gained
     substantial advances in parallel to the developments in computer vision
     and machine learning applications. In particular, we saw striking developments
     in the last twenty years. Today Androcars are running their test drives on the public roads.

     In the last few years advances in the deep learning paved the way to run speech recognition classification without using human-engineered features and gained great success. Classification without using man-made features produced very good results even sometimes higher than the results of mfcc or mel filter banks. LSTM and CNNs can be used for speech recognition applications.

   • View this event in Google Calendar