eval_pale.py

# -*- coding=UTF-8 -*-\n
from eval.eval_pale import Eval_PALE
from eval.measures import *

from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter

def parse_args():
    parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter
                            , conflict_handler='resolve')
    parser.add_argument('-feat-src', required=True
                        , help='features from source network')
    parser.add_argument('-feat-end', required=True
                        , help='features from end network')
    parser.add_argument('-linkage', required=True
                        , help='linkage for test')
    parser.add_argument('-eval-type', default='mrr'
                        , help='mrr/ca/cls (MRR/Candidate selection/Classification)')
    parser.add_argument('-model', required=True
                        , help='Model file')
    parser.add_argument('-n-layer', default=5, type=int
                        , help='Number of layers')
    parser.add_argument('-model-type', required=False
                        , help='Model type: [lin/mlp]')
    parser.add_argument('-n-cands', default=9, type=int
                        , help='number of candidates')
    parser.add_argument('-output', required=True
                        , help='Output file')

    return parser.parse_args()

def main(args):
    eval_model = Eval_PALE(args.model_type)
    eval_model._init_eval(feat_src=args.feat_src,
                            feat_end=args.feat_end,
                            linkage=args.linkage
                )
    if args.eval_type=='mrr':
        eval_model.calc_mrr_by_dist(model=args.model, candidate_num=args.n_cands
                                , n_layer=args.n_layer, dist_calc=geo_distance, out_file=args.output)
    if args.eval_type=='cls':
        eval_model.eval_classes(model=args.model, candidate_num=args.n_cands
                                , n_layer=args.n_layer, dist_calc=geo_distance, out_file=args.output)

if __name__=='__main__':
    main(parse_args())