z[j]=kw[k]x[j+k],u[m]=z[sm] for some stride parameter s>0. That is, u takes every s samples of z. (a) (3 points) If x is 250 milliseconds of single channel audio sampled at 20kHz, how many samples are in x ? (b) (3 points) If x has length 1000, w has length 10 and s=4, how many output samples are in u assuming the convolution is only computed on the valid samples. (c) (4 points) Given a gradient J/u[m], how do you compute J/w[k]