Given the logarithm of the weights, L, and observation, R, this computes the weighted average

logWeightedAverage.py

def logWeightedAverage(L, R):

	n = len(L)

	Lscale 	       = 0.
	sumNumerator   = 0.
	sumDenominator = 0.

	for j in range(n):
		
		if j == 0 :
			wt     = 1.0
			Lscale = L[j]
			sumNumerator   = R[j]
			sumDenominator = 1.0
		
		else:
			
			wt = np.exp(L[j] - Lscale)
			
			if wt <= 1.0:
				
				sumNumerator   = sumNumerator   + wt * R[j]
				sumDenominator = sumDenominator + wt
				
			else:	# Lscale needs to be updated to L[j]
			
				sumNumerator   = sumNumerator/wt   + R[j]
				sumDenominator = sumDenominator/wt + 1.0
				Lscale = L[j]

	weightAverage = sumNumerator/sumDenominator
	
	return weightAverage