This study investigates the use of the SWIM (Sampling Where It Matters) sampling framework within DeepONet for solving partial differential equations. Unlike conventional neural network training methods that employ an optimizer like SGD or Adam, SWIM is a sampling strategy utilized for selecting the weights and biases of neural networks using the training data. In this thesis, we explore an iterative sampling algorithm to examine the compatibility of SWIM with DeepONet. The iteratively sampled DeepONet is then trained and tested on PDEs, with the results and methods being thoroughly discussed.