Train a linear regression model using full-batch gradient descent under three different gradient clipping strategies, and compare their loss curves.
