initial J = 4.3377   optimal J* = 6.5963

                            method  delta  gap@10  J after 50  gap closed  J drops  backtracks   max KL
      natural + line search (TRPO)  0.003   0.105      6.1104       0.785        0          14   0.0030
      natural + line search (TRPO)   0.03   0.482      6.5963       1.000        0           5   0.0299
  Euclidean gradient + line search  0.003   0.101      5.9810       0.728        0          21   0.0030
  Euclidean gradient + line search   0.03   0.411      6.5963       1.000        0           7   0.0299
           natural, no line search  0.003   0.210      6.2765       0.858        0           0   0.0030
           natural, no line search   0.03   0.619      6.5963       1.000        0           0   0.0303
Euclidean gradient, no line search  0.003   0.203      6.2504       0.847        0           0   0.0030
Euclidean gradient, no line search   0.03   0.606      6.5963       1.000        0           0   0.0301
