Description

1. Your task is to find a linear approximation of the function 1 + x, x ∈ [0,1]. Your homework should contain the following steps:
a) Generate N = 10000 random numbers from [0,1]:
x1,x2,…,xN ∈ [0,1],
√
and then obtain their labels: yi = 1 + xi, i = 1,2,…,N.
b) Do linear regression on your generated data using stochastic gradient descent (implement yourself).
c) Do linear regression on your generated data using mini-batch gradient descent (implement yourself). d*) Do linear regression on your generated data using tensorflow.
e) Sketch the graphs of all approximations on one graph.
f) Compare all solutions with the first degree Taylor approximation of√

the function 1 + x.
2*. Pass the following tutorial.
Remarks:
1. Exercises with asterisks are supplementary and will not be graded.
2. Don’t forget about train, validation and test sets.
3. Use jupyter notebook for writing your code.
4. You can use google for any question, but don’t do copies of others’ codes.
5. You can ask me whatever you want and whenever you want.
1