Print Email Facebook Twitter On the Regularization of Convolutional Neural Networks and Transformers under Distribution Shifts Title On the Regularization of Convolutional Neural Networks and Transformers under Distribution Shifts Author Assini, Leo (TU Delft Electrical Engineering, Mathematics and Computer Science) Contributor Bohmer, Wendelin (mentor) Poulsen, C.B. (graduation committee) Degree granting institution Delft University of Technology Programme Computer Science and Engineering Project CSE3000 Research Project Date 2022-06-24 Abstract The use of Transformers outside the realm of natural language processing is becoming more and more prevalent. Already in the classification of data sets such as CIFAR-100 it has shown to be able to perform just as well as the much more established Convolutional Neural Network. This paper investigates the possible out-of-distribution capabilities of the multi-head attention mechanism, through the classification of the MNIST data set with added backgrounds. Additionally, various regularization techniques are applied to increase the generalization capabilities even more. Regularization is shown to be an important tool to improve out-of-distribution accuracy, though it might imply some trade offs for in-distribution settings. To reference this document use: http://resolver.tudelft.nl/uuid:a5eb41ea-c6b4-4035-a482-fa9bd74d42f7 Part of collection Student theses Document type bachelor thesis Rights © 2022 Leo Assini Files PDF Thesis_LeoAssini.pdf 5.18 MB Close viewer /islandora/object/uuid:a5eb41ea-c6b4-4035-a482-fa9bd74d42f7/datastream/OBJ/view