Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Xin Li

Committee Member

Shuo Wang

Committee Member

Donald Adjeroh

Committee Member

Katerina Goseva-Popstojanova

Committee Member

Yuxin Liu

Committee Member

Yong-Lak Park


The intertwined history of artificial intelligence and neuroscience has significantly impacted their development, with AI arising from and evolving alongside neuroscience. The remarkable performance of deep learning has inspired neuroscientists to investigate and utilize artificial neural networks as computational models to address biological issues. Studying the brain and its operational mechanisms can greatly enhance our understanding of neural networks, which has crucial implications for developing efficient AI algorithms. Many of the advanced perceptual and cognitive skills of biological systems are now possible to achieve through artificial intelligence systems, which is transforming our knowledge of brain function. Thus, the need for collaboration between the two disciplines demands emphasis. It's both intriguing and challenging to study the brain using computer science approaches, and this dissertation centers on exploring computational mechanisms related to face perception.

Face recognition, being the most active artificial intelligence research area, offers a wealth of data resources as well as a mature algorithm framework. From the perspective of neuroscience, face recognition is an important indicator of social cognitive formation and neural development. The ability to recognize faces is one of the most important cognitive functions. We first discuss the problem of how the brain encodes different face identities. By using DNNs to extract features from complex natural face images and project them into the feature space constructed by dimension reduction, we reveal a new face code in the human medial temporal lobe (MTL), where neurons encode visually similar identities. On this basis, we discover a subset of DNN units that are selective for facial identity. These identity-selective units exhibit a general ability to discriminate novel faces. By establishing coding similarities with real primate neurons, our study provides an important approach to understanding primate facial coding. Lastly, we discuss the impact of face learning during the critical period. We identify a critical period during DNN training and systematically discuss the use of facial information by the neural network both inside and outside the critical period. We further provide a computational explanation for the critical period influencing face learning through learning rate changes. In addition, we show an alternative method to partially recover the model outside the critical period by knowledge refinement and attention shifting.

Our current research not only highlights the importance of training orientation and visual experience in shaping neural responses to face features and reveals potential mechanisms for face recognition but also provides a practical set of ideas to test hypotheses and reconcile previous findings in neuroscience using computer methods.