def get_prior_value()

in Experiments/PolicyNetworks.py [0:0]

• 22 lines of code
• 8 McCabe index (conditional complexity)

	def get_prior_value(self, elapsed_t, max_limit=5):

		skill_time_limit = max_limit-1

		if self.args.data=='MIME' or self.args.data=='Roboturk' or self.args.data=='OrigRoboturk' or self.args.data=='FullRoboturk' or self.args.data=='Mocap':
			# If allowing variable skill length, set length for this sample.				
			if self.args.var_skill_length:
				# Choose length of 12-16 with certain probabilities. 
				lens = np.array([12,13,14,15,16])
				# probabilities = np.array([0.1,0.2,0.4,0.2,0.1])
				prob_biases = np.array([[0.8,0.],[0.4,0.],[0.,0.],[0.,0.4]])				

				max_limit = 16
				skill_time_limit = 12

			else:
				max_limit = 20
				skill_time_limit = max_limit-1	

		prior_value = torch.zeros((1,2)).to(device).float()
		# If at or over hard limit.
		if elapsed_t>=max_limit:
			prior_value[0,1]=1.

		# If at or more than typical, less than hard limit:
		elif elapsed_t>=skill_time_limit:
	
			if self.args.var_skill_length:
				prior_value[0] = torch.tensor(prob_biases[elapsed_t-skill_time_limit]).to(device).float()
			else:
				# Random
				prior_value[0,1]=0. 

		# If less than typical. 
		else:
			# Continue.
			prior_value[0,0]=1.

		return prior_value