Consider a dataset with four negative examples (-2, 0), (0, -2), (1, -1), (-1, 1) and four positive examples (2,0), (1, 1), (0, 2), (1,1). Recall that perceptron is trained on a sequence of examples. On each example, the weights are updated if perceptron makes a mistake in classifying that example. Starting with w = [00], use the perceptron algorithm to learn on the data points in the above order (from negative examples to positive examples). Provide values of the weights after each iteration of update using each data point. Find an ordering of examples in this dataset on which perceptron make at most two mistakes during training, or explain that no such sequence exists.