Eureqa how run ends

[red333]

The original (Cornell University) Eureqa program could be run forever, and the program would keep searching for improved formulas at every level of complexity. I can understand why Data Robot wanted the runtime to be constrained. There is a Data Robot fine tuning parameter, EUREQA_max_generations, that I thought determined how long the program ran. But it appears that that is not the case. Even if I set that to a very large number, the program stops earlier. I'm attaching part of the log for a run that shows that there is some metric that is satisfied (hundred percent of the metric) for a particular complexity level. What is this metric? What determines how long the program runs? And how can I make it so that it will run longer and find more solutions for different levels of complexity?

Thanks, Richard

[08-22-2023 18:14:34] 95/100% Metric: 0.018413111756295467, Complexity: 19.0, Expression: "Target = 2490.89755462728*exp(0.00431257495352471*Lpp) - 2486.98837566945 - 11.3066533514062*Lpp - 0.0559574818023985*cos(2.74491770302305e-5 - 0.2618*MLT)"

[08-22-2023 18:14:35] 96/100% Metric: 0.018220045384136872, Complexity: 19.0, Expression: "Target = 2466.2870190519*exp(0.00431257495352471*Lpp) - 2462.38169379372 - 11.1983403478849*Lpp - 0.0604628316532552*cos(0.143864212707825 - 0.2618*MLT)"

[08-22-2023 18:14:35] 97/100% Metric: 0.018220045384136872, Complexity: 19.0, Expression: "Target = 2466.2870190519*exp(0.00431257495352471*Lpp) - 2462.38169379372 - 11.1983403478849*Lpp - 0.0604628316532552*cos(0.143864212707825 - 0.2618*MLT)"

[08-22-2023 18:14:36] 98/100% Metric: 0.018220045384136872, Complexity: 19.0, Expression: "Target = 2466.2870190519*exp(0.00431257495352471*Lpp) - 2462.38169379372 - 11.1983403478849*Lpp - 0.0604628316532552*cos(0.143864212707825 - 0.2618*MLT)"

[08-22-2023 18:14:36] 99/100% Metric: 0.018220045384136872, Complexity: 19.0, Expression: "Target = 2466.2870190519*exp(0.00431257495352471*Lpp) - 2462.38169379372 - 11.1983403478849*Lpp - 0.0604628316532552*cos(0.143864212707825 - 0.2618*MLT)"

[08-22-2023 18:14:37] 100/100% Metric: 0.018220045384136872, Complexity: 19.0, Expression: "Target = 2466.2870190519*exp(0.00431257495352471*Lpp) - 2462.38169379372 - 11.1983403478849*Lpp - 0.0604628316532552*cos(0.143864212707825 - 0.2618*MLT)"

[08-22-2023 18:14:38] 'Eureqa Generalized Additive Model (3000000 Generations) (Least-Squares Loss)': completed fitting.

[jenD · Accepted solution]

Hi Richard. This has generated a really interesting conversation internally. Here's what we are thinking...can I ask you to check the value of the timeout setting and increase it manually to something like 1000000. You can find the argument, EUREQA_timeout_sec, in the Advanced Tuning tab. To read up on it, see the in-app model docs (you need to be logged in) here . Then, can you report back whether the searching continues beyond the solution where you’re currently stuck?

Thanks....jen

[red333]

Yes, that did allow it to run longer and find more solutions. Thanks.