... FPGA

Field-Programmable Gate Array

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... industry.

Although they are not necessarily useless as components in a less conventional architecture, such as certain artificial neural networks: Thanks to Ed Rietman for pointing this out.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... approach.

Thanks to Dario Floreano for pointing this out.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... PSU,

Hewlett Packard HP E3631A.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... shield,

This wiring will be eliminated in the Mk. II Evolvatron.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... conditions.

Another possibility would be the circulation of a refrigerated fluid through the heatsink.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...$15^{\circ}$C.

Other investigators are cautioned to beware condensation problems (corrosion, mould), especially in causing bad connections in socketed chips.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... FPGAs.

Here, evolution time is dominated by the time taken for each fitness evaluation. Rather arbitrarily, an evaluation consisted of exposure to five half-second bursts of each tone [9], taking a total of 5s (whether evaluated on just one FPGA, or on several simultaneously). Hence, the original 5000 generations took around 5s$\times$50$\times$5000 $\simeq 2$ weeks, and the 300 generations reported here $\simeq 21$ hours. Possibly, much shorter evaluations could be made adequate.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.