analysis.py

import numpy as np

models_config = []
models_config.append(['GCN', 3, -1])
models_config.append(['SAGE', 3, -1])
models_config.append(['GAT', 3, -1])
models_config.append(['GIN', 3, -1])
models_config.append(['PGNN_fast', 1, 2])
models_config.append(['PGNN_fast', 2, 2])
models_config.append(['PGNN_fast', 1, -1])
models_config.append(['PGNN_fast', 2, -1])

fname_missing = []
for task in ['link', 'link_pair']:
    if task == 'link':
        datasets_name = ['Cora','grid','communities','ppi']
    else:
        datasets_name = ['communities', 'email', 'protein']
    for dataset_name in datasets_name:
        for model_config in models_config:
            results = []
            for repeat in range(10):
                fname = 'results/{}_{}_{}_layer{}_approximate{}_repeat{}.txt'.format(
                    task, model_config[0], dataset_name, model_config[1], model_config[2], repeat)
                try:
                    with open(fname,'r') as f:
                        result = f.read()
                        results.append(float(result))
                except:
                    fname_missing.append(fname)
            results = np.array(results)
            print('{}\t\t\t\t\t{}\t{}\t{}'.format(fname,np.mean(results).round(4), np.std(results).round(4), len(results)))