typo in bench_methods_run.py

[dinalzein]

There is a typo in bench_methods_run.py for methods: run_fista and run_l_bfgs.

In method run_fista :

pobj, times, d_hat, z_hat = learn_d_z(X, n_atoms, n_times_atom, func_d=update_d_block, reg=reg, n_iter=n_iter, random_state=random_state, n_jobs=1, solver_z='fista', solver_z_kwargs=dict(max_iter=2), ds_init=ds_init, verbose=verbose)

should be replaced by:

pobj, times, d_hat, z_hat, reg = learn_d_z(X, n_atoms, n_times_atom, func_d=update_d_block, reg=reg, n_iter=n_iter, random_state=random_state, n_jobs=1, solver_z='fista', solver_z_kwargs=dict(max_iter=2), ds_init=ds_init, verbose=verbose)

same modification can be applied to run_l_bfgs

[agramfort]

@dinalzein can you send a PR?

[dinalzein]

yes.

[dinalzein]

Also, for run_multichannel_gcd_fullrank function in this same class bench_methods_run.py, the uv_constraint='separate' parameter for the called method learn_d_z_multi should be replaced by uv_constraint='auto' because rank1=False.
Do I fix them together in one PR?

[agramfort]

Yes please

[dinalzein]

Done.

in the same class bench_methods_run.py, the following can be added:

isExist = os.path.exists('figures')
 if not isExist:
   os.makedirs('figures')

just before the last two lines in the main function to avoid file doesn't exist error (I didn't include it in the PR, but I can do another one).

I got an error inn bench_methods_plot.py, for the plot_barplot function,

and replacing
rect = ax.bar(left=x_positions + i * width, height=np.array(this_df['mean']), width=width, label=label, hatch=hatch, color=colors[i])

by:

rect = ax.bar(x_positions + i * width, height=np.array(this_df['mean']), width=width, label=label, hatch=hatch, color=colors[i])

worked fine for me.