The Spiral Discovery Network as an Automated General-Purpose Optimization Tool

Figure 4

Cycles of SDN from left to right, top to bottom. Performance is indicated at the top of the plots, with the sizes of the evaluated locations inversely proportional to the value of the loss function (and scaled per plot). The figure shows that the smallest loss value, in a 2% vicinity of the minimum, was found as early as in the 7th cycle, without recourse to any kind of gradient value.