Task $5(10$ points$).$

Slide $16$ ofslides on perceptron trainingprovides pseudocode for training a perceptron. Step $1$ of that pseudocode is to initialize weights to small random values. What would happen if you initialized all weights to zero? How would classification accuracy be affected, compared to initializing weights randomly $($would you expect it to be higher, lower, or about the same$)?$ Justify your answer.