A Visualization Tool for Analyzing Training Dynamics of Neural Networks- Group 1

2021-2022 Summer
Faculty Department of Project Supervisor: 
Faculty of Engineering and Natural Sciences
Number of Students: 

The deep learning models benefit from large amount of labeled data. Often time the relationship of the inputs to the results is hidden due to the black-box nature of the model. Analyzing the training dynamics across the epochs can reveal which examples are more informative, redundant or of poor quality.
This work will involve analyzing input examples’ training dynamics in deep learning models. The students are expected to build visualization tools that would aid the investigation.
The students with creativity and good programming skills are welcome to apply.

Related Areas of Project: 
Computer Science and Engineering
Electronics Engineering

About Project Supervisors

Öznur Taştan
Faculty of Engineering and Natural Science