this is my respository for EE 298 Deep Learning class
For our class assignment,I used:
a tolerance value, 𝛿 = 0.01 a varying learning rate,epsilon, an initial value of learning rate, 𝝐 = 0.0001, if it is a 3rd degree polynomial and an initial value of learning rate, 𝝐 = 0.001, if the polynomial degree is less than 3, if the current loss function is greater than the previous loss function, the learning rate is halved to avoid overshooting
The two initial values of learning rate are based on trial and error experimentation.
The results are in the txt files: linreg_results_2_4_neg1.txt linreg_results_2_5.txt linreg_results_10_neg10_10_neg10.txt linreg_results_2_2_4_neg1.txt
For the given input: 2 4 -1 The recommended learning rate,𝝐, is 𝝐 = 1.5625e-05 given tolerance value, 𝛿 = 0.01