def _reward(self) -> float:
"""
Notes
-----
Logged parameters:
* energy (float): the energy of the current conformer
"""
config = self.config
self.backup_mol.AddConformer(self.conf, assignId=True)
energy = get_conformer_energy(self.mol)
self.step_info['energy'] = energy
self.backup_energys.append(energy)
self._prune_conformers()
total_reward = np.log(
np.exp(-1.0 * (np.array(self.backup_energys) - config.E0) /
(KB * config.tau)).sum() / config.Z0)
if not np.isfinite(total_reward):
total_reward = np.finfo(np.float64).eps
rew = total_reward - self.total_reward
return rew
def _reward(self):
energy = get_conformer_energy(self.mol)
self.step_info['energy'] = energy # log energy
reward = 1. if energy < self.energy_thresh else 0.
if energy < self.energy_thres:
self.confs_below_threshold += 1
self.episode_info[
'confs_below_threshold'] = self.confs_below_threshold
return reward
def _reward(self) -> float:
"""Returns :math:`e^{-1 * energy}` where :math:`energy` is the
energy of the current conformer of the molecule.
Notes
-----
Logged parameters:
* energy (float): the energy of the current conformer
"""
energy = get_conformer_energy(self.mol)
reward = np.exp(-1. * energy)
self.step_info['energy'] = energy
return reward
def _reward(self) -> float:
"""
Notes
-----
Logged parameters:
* energy (float): the energy of the current conformer
"""
config = self.config
self.backup_mol.AddConformer(self.conf, assignId=True)
energy = get_conformer_energy(self.mol)
self.step_info['energy'] = energy
reward = np.exp(-1. * (energy - config.E0) /
(KB * config.tau)) / config.Z0
if self._done():
reward -= self._pruning_penalty()
return reward
def _reward(self) -> float:
"""
Notes
-----
Logged parameters:
* energy (float): the energy of the current conformer
* repeat (int): total number of repeated actions so far in the episode
"""
config = self.config
energy = get_conformer_energy(self.mol)
self.step_info['energy'] = energy
if tuple(self.action) in self.seen:
self.repeats += 1
self.episode_info['repeat'] = self.repeats
return 0.
else:
self.seen.add(tuple(self.action))
reward = np.exp(-1. * (energy - config.E0) /
(KB * config.tau)) / config.Z0
return reward