parser.add_argument("--n_splits", type=int, default=5, help="default: 5")
parser.add_argument("--rand_seed", type=int, default=0, help="default: 0")
parser.add_argument("--fit", action="store_true", help="Default: False")
return parser
def simulate_helper(hawkes, max_jumps):
simu_hawkes.reset()
simu_hawkes.max_jumps = max_jumps
simu_hawkes.simulate()
return simu_hawkes.timestamps
args = get_parser().parse_args()
set_rand_seed(args.rand_seed)
print(args)
# simulate drug event_seqs
baseline = args.baseline * np.random.random(args.n_types)
adjacency = np.diag(np.random.random(args.n_types))
if args.n_correlations > 0:
comb = list(permutations(range(args.n_types), 2))
idx = np.random.choice(range(len(comb)),
size=args.n_correlations,
replace=False)
comb = [comb[i] for i in idx]
for i, j in comb:
adjacency[i, j] = np.random.random()
if __name__ == "__main__":
args = get_parser().parse_args()
assert args.model is not None, "`model` needs to be specified."
output_path = osp.join(
args.output_dir,
args.dataset,
f"split_id={args.split_id}",
args.model,
get_hparam_str(args),
)
makedirs([output_path])
# initialization
set_rand_seed(args.rand_seed, args.cuda)
init_logging(output_path)
logger = get_logger(__file__)
logger.info(args)
export_json(vars(args), osp.join(output_path, "config.json"))
# load data
input_path = osp.join(args.input_dir, args.dataset)
data = np.load(osp.join(input_path, "data.npz"), allow_pickle=True)
n_types = int(data["n_types"])
event_seqs = data["event_seqs"]
train_event_seqs = event_seqs[data["train_test_splits"][args.split_id][0]]
test_event_seqs = event_seqs[data["train_test_splits"][args.split_id][1]]
# sorted test_event_seqs by their length
test_event_seqs = sorted(test_event_seqs, key=lambda seq: len(seq))