CODE FOR S. Satpathi, H. Gupta, S. Liang, and R. Srikant `The Role of Regularization in Overparameterized Neural Networks', Submitted to CDC 2020
DATASET - first two classes of MNIST. Digit 0 and 1.
MNISTCompleixty CDC2020.ipynb has code showing satisfaction of theoritical assumptions and the theoritical bounds on test and training error. Fig. 1 and Table 1 in the paper
MNIST regularizer CDC2020.ipynb has code showing the tradeoff betweeen test and training loss for different degrees on regularisation. Fig. 2 and Table 2 in the paper