Figure 4From: Parallel mutual information estimation for inferring gene regulatory networks on GPUsIllustration of unwinding the recursive Eqn. (8) into k dependent steps using k = 3 and R = 10 for computing one row of a WM. S is the shared memory array. In each step, 12 threads work together to compute the values of S. After 3 steps, the first 10 values in S is the final result.Back to article page