There are 7 python files in this folder:
1) data.py simply loads the LIBSVM a9a dataset
2) function.py defines a stochastic oracle and
a hessian vector product oracle for Logistic
Regression
3) optimizer.py implements the different 
optimization algorithms
4) run_help.py defines some auxillary functions
for running the experiments
5) running 'python tune.py' runs the tuning
experiment for all values of mu/lambda, M for the 
value of T specified in the code. After tuning the 
best learning rates and the best obtained loss for 
every algorithm is stored in 
'./results/T/store_mu_M.npy' for each value of R.
6) running 'python repeat.py' runs multiple
repetitions of all the algorithms for all values
of mu/lambda, M using the best learning rate for
R as stored in './results/T/store_mu_M.npy'; the
number of repetitions can be specified in the code
and all the obtained losses for each run are
stored in './results/T/Losses_mu_M.npy'
7) running 'python plot_grid_w_rep.py' stores
the generated figures 1 and 2 (in the paper) in
'figures/T/plot_w_rep.png' 