Figure 7

Maximal probability reconstruction vs. gradient descent reconstruction. Maximal probability reconstruction vs. gradient descent reconstruction of trajectories of the Boolean XOR (left) and Boolean NOR (right) networks. The gradient descent algorithm is given an inaccurate structure. The rows correspond to the time points and columns to the network nodes. For display purposes, only a prefix of the trajectory is shown. The yellow color represents mistakes, i.e., values different than the real Boolean values, and orange represents a correct value. In each of the two comparisons, the maximal probability reconstruction is presented to the left of the gradient descent reconstruction. Overall, the gradient descent is more accurate than the maximal probability reconstruction despite the imperfect structures that are given to it as input. The XOR network's trajectory reconstruction is not affected by the error in structure, while the NOR network's reconstruction is slightly less accurate. The percentages of incorrect reconstructed values for maximal probability reconstruction are 17.6% (XOR) and 18% (NOR), and for the gradient descent reconstruction, 6.7% (XOR) and 3% (NOR).