"type": "float",
"value": "data.state",
"is_index": True,
"retrieval_method": "radius-n",
"retrieval_method_params": param,
},
"act": {"type": "float", "value": "data.action", "is_index": False, "retrieval_method": "cosine"},
"delta_state": {"type": "float", "value": "data.next_state - data.state", "is_index": False},
}
model = CASML(
CbTData(
case_t_template,
rho=args.rho,
tau=args.tau,
sigma=args.sigma,
plot_reuse=False,
plot_reuse_params="original_origin",
),
ncomponents=args.ncomponents,
)
with Timer() as tm:
for j, states in enumerate(train[i]):
# Train CASML's case base and hmm with states and actions
model.fit(states, actions)
print ("Model trained in %.03f sec." % tm.time)
failed = 0
sampled = np.zeros((ntrials, d, n))
ncases = model._cb_t._counter - 1
"retrieval_method": args.retrieval_method,
"retrieval_method_params": args.retrieval_method_params
},
"act": {
"type": "float",
"value": "data.action",
"is_index": False,
"retrieval_method": "cosine",
},
"delta_state": {
"type": "float",
"value": "data.next_state - data.state",
"is_index": False,
}
}
model = CASML(CbTData(case_t_template, rho=args.rho, tau=args.tau, sigma=args.sigma),
ncomponents=args.ncomponents)
n = obs.shape[0]
action_error = -np.inf * np.ones(n)
delta_error = -np.inf * np.ones(n)
for i, states in enumerate(obs):
# Train CASML's case base and hmm with states and actions
model.fit(states, actions)
# Test model
cntr = 0
iter_ = 0
while cntr < 10:
sampled = None
try: