The problem with the entropy method is that it's applicable only to classification tasks. So, we can't apply our algorithm for regression or reinforcement learning tasks. To overcome this, we'll see one more algorithm, called inequality minimization TAML. It's as simple as the entropy method. In this method, we try to minimize the inequality. There are several inequality measures used in economics to measure income distribution, wealth distribution, and more. In our meta learning setting, we can use these economic inequality measures to minimize our task bias. So, the bias of the model toward a task can be minimized by minimizing the inequality over the losses of all of the sampled tasks in a batch.