Source code for demo_cli

import argparse



def main():
    
    parser = argparse.ArgumentParser()
    
    # dataset synthesis / loading settings
    parser.add_argument('--dataset', type=str, default='speechcommands',
                        choices=['speechcommands', 'mnist', 'mante'])
    parser.add_argument('--top_n', type=int, default=10)
    parser.add_argument('--nbatch', type=int, default=100)
    parser.add_argument('--seqlen', type=int, default=200000)
    parser.add_argument('--stoch', type=str, default='stim', choices=['stim', 'delay', 'none'], )
    parser.add_argument('--dim_ring', type=int, default=6)
    parser.add_argument('--sigma', type=float, default=1)
    
    # input preprocessing
    parser.add_argument('--center', action='store_true')
    parser.add_argument('--unit_std', action='store_true')
    parser.add_argument('--orthgonalize', action='store_true')
    parser.add_argument('--whiten', action='store_true')
    parser.add_argument('--zero_phase', action='store_true')
    parser.add_argument('--skip_normed_trial', action='store_true')

    # tau extraction methods
    parser.add_argument('--skip_spaced', action='store_true')
    parser.add_argument('--num_tau0', type=int, default=10)
    parser.add_argument('--filt', action='store_true')
    parser.add_argument('--first_unstream', action='store_true')
    parser.add_argument('--flattening_method', type=str, default='flat')
    
    # network and heterogeneity setup
    parser.add_argument('--Nnet', type=int, default=10)
    parser.add_argument('--N', type=int, default=250)
    parser.add_argument('--p', type=float, default=.1)
    parser.add_argument('--J', type=float, default=1.)
    parser.add_argument('--Ju', type=float, default=1.)
    parser.add_argument('--af', type=str, default='tanh')
    parser.add_argument('--h', type=str, default='uniform')
    parser.add_argument('--h_is_additive', action='store_true')
    parser.add_argument('--dt', type=float, default=1.)
    # parser.add_argument('--tau_min', type=float, default=0.)
    # parser.add_argument('--tau_max', type=float, default=1.)
    
    # pooling method
    parser.add_argument('--pooling', type=list, default=['go', 'avg', 'snapshot'])
    parser.add_argument('--go_length', type=int, default=10)
    parser.add_argument('--nsnapshot', type=int, default=10)
    
    # training settings
    parser.add_argument('--njobs', type=int, default=2)
    parser.add_argument('--cv', type=int, default=0)
    parser.add_argument('--Creg', type=float, nargs='+', default=[0.01, 0.1, 1., 10, 100])
    parser.add_argument('--clf_name', type=str, default='logreg')
    
    # plotting
    parser.add_argument('--skip_plot_uvx', action='store_true')
    parser.add_argument('--skip_plot_man', action='store_true')
    
    # naming
    parser.add_argument('--name', type=str, default='test_cli')
    parser.add_argument('--seed', type=int, default=1234)
    
    
    args = parser.parse_args()


    demo_cfg = dict(
        # dataset synthesis / loading settings
        dataset_name=args.dataset,
        top_n = args.top_n,
        Nbatch=args.nbatch, 
        seq_len=args.seqlen,
        dim_ring=args.dim_ring,
        sigma = args.sigma,
        
        # input preprocessing
        normed_trial= not args.skip_normed_trial,
        center=args.center, 
        unit_std=args.unit_std, 
        orthgonalize=args.orthgonalize, 
        zero_phase=args.zero_phase, 
        whiten=args.whiten,  
        
        # tau extraction methods
        num_tau0 = args.num_tau0,
        filt=args.filt,
        first_unstream=args.first_unstream,
        flattening_method=args.flattening_method,

        # network and heterogeneity set up
        N = args.N,
        p = args.p,
        J = args.J,
        Ju= args.Ju,
        af=args.af, 
        dt=args.dt,

        h = args.h, 
        h_is_multiplicative = not args.h_is_additive,
        # tau_min = args.tau_min, 
        # tau_max = args.tau_max,
        
        # pooling
        pooling_funcs = args.pooling,
        golength = args.go_length,
        nsnapshot = args.nsnapshot,
        
        # training
        n_jobs=args.njobs,
        cv=args.cv,
        Creg = args.Creg,
        clf_name = args.clf_name,

        # naming, seed, and the nubmer of realizations
        prefix = args.name,
        seed = args.seed,
        Nnet = args.Nnet,      

        # plotting
        plot_uvx = not args.skip_plot_uvx, 
        plot_man = not args.skip_plot_man,
        )

    # machine learning demos
    if args.dataset in ['mnist', 'speechcommands']:
        import ml 

        if not args.skip_spaced:
            ml.run_spaced_tau0_multinet(**demo_cfg)
        else:
            ml.run_multinet(**demo_cfg)

    # neuroscience demos
    elif args.dataset == 'mante':
        import ns 
        if args.stoch == 'stim':
            demo_cfg['stim_period'] = [50, 10, 250]
            demo_cfg['delay_period']= 100
        elif args.stoch == 'delay':
            demo_cfg['stim_period'] = 100
            demo_cfg['delay_period']= [50, 10, 250]
        elif args.stoch == 'none':
            demo_cfg['stim_period'] = 100
            demo_cfg['delay_period']= 1
            

        if not args.skip_spaced:
            ns.run_spaced_tau0_multinet(**demo_cfg)
        else:
            ns.run_multinet(**demo_cfg)

    else:
        raise ValueError(f'Dataset name {args.dataset} not recognized.')



if __name__ == '__main__':
    main()