Trust-region methods that constrain steps to a radius where the model is trusted

Original label: Trust-region methods that constrain steps to a radius where the model is trusted