Feng Xia will present her MSE Thesis talk "Exploring Early Exiting Strategies for Deep Neural Networks" on Monday, April 29, 2024 at 3pm in PSH 311. Advisor: Tom Griffiths; Reader: Matt Weinberg All are invited to attend. Please see abstract below. Abstract: Abstract: Deep neural networks (DNNs) are renowned for their accuracy across a spectrum of machine learning tasks but often suffer from prolonged inference times due to their depth. To address this, early exiting strategies have been proposed, allowing predictions to be made at intermediate layers, thus reducing inference time. This paper explores several methodologies to enhance early exit mechanisms within a single multi-exit DNN architecture on an image classification task. First, we improve the task transferability of traditional confidence measures by converting entropy to probabilities. Second, we observe that the total uncertainty used by conventional confidence measures does not consistently reflect true model uncertainty, especially for ambiguous images. To address this, we utilize a Dirichlet framework for neural networks, which assumes the network outputs a Dirichlet distribution of class probabilities rather than a point estimate. We introduce a novel training mechanism that includes both original training data and artificially created ambiguous data by blending training images. This approach allows more ambiguous data to exit early compared to the previous approach. Lastly, we propose two alternative perspectives to determine if an intermediate prediction is sufficient for an early exit. While our experiments focus on a specific neural network architecture and dataset, our methodologies are designed to be independent of architectures and tasks, rendering potential for wider applicability in deep neural networks. CS Grad Calendar Link: https://calendar.google.com/calendar/event?action=TEMPLATE https://calendar.google.com/calendar/event?action=TEMPLATE&tmeid=NXJna3B2c2xsMGU4ZGlyaTZxdHY4ZmZzYzYgYWNnMDc5YmxzbzRtczNza2tmZThwa2lyb2dAZw&tmsrc=acg079blso4ms3skkfe8pkirog%40group.calendar.google.com &tmeid=NXJna3B2c2xsMGU4ZGlyaTZxdHY4ZmZzYzYgYWNnMDc5YmxzbzRtczNza2tmZThwa2lyb2dAZw&tmsrc=acg079blso4ms3skkfe8pkirog%40group.calendar.google.com