решить домашнее задание по началам Machine Learning Сделать домашнее задание. Задание дано на английском (файл .ipynb) Perceptron - digit classification: The idea is to use a perceptron for digit classification. Important: In the following we focus on the discussion if the dataset is linear separable. In particular, we will not split the dataset into training and test dataset, which have to be done in the real world. We have all helper functions: online training, batch training, plot picture function, perceptron error, perceptron class and other. Data: We are using a data set that is included in scikit learn. Task 1: Investigate which pairs of digits are linear separable. Task 2: Influence on epochs We consider only the sub datasets of a pair of digits that are linear separable (this means the one you found in Task 1). What is the influence of learning rate, shuffling, and online/batch learning on the number of needed epochs? Thereby, consider the total number of epochs for all sub datasets (this means the sum). If useful show also plots of your results. Task 3: Influence of learning rate for reduce_1_digit function. Investigate the influence of learning rate on epochs and/or convergence for all ten possible cases of reduce_1_digit. Discuss the impact of shuffling. Give also plots of your results. Task 4: Find optimal weights Consider now the case where a perceptron classifies one digit from the whole dataset. Update the online learning algorithm in a way that it returns the best weights within a given number of epochs. Use the updated algorithm to find good solutions for all the different 10 cases. Give for each case the number of wrong classifications.