Best practices in measuring energy consumption in population-based metaheuristics

JJ Merelo, Cecilia Merelo-Molina

Universidad de Granada, Zenzorrito

Modern Taurokathapsia

We need to tame the bull of energy consumption in EAs

Meet the Brave New Algorithm

Portada del libro

Literature-inspired metaphor

Stratified population evolutionary algorithm

α caste → with itself, crossover + mutation

β only with α → crossover + mutation

γ, δ, ε → only mutation; γ → hillclimbing

It's only fitting that we use Greek letters

Crossover was invented in Crete

Representation of the young minotaur

After a couple of failed attempts

kidnapping of Europe

☝️Take it easy

⏱️ Decrease sampling frequency

different
                                sampling frequency

✌️ Improve little by little

📏 But always measure after trying to improve

Microoptimization

Ⅲ Adjust your perception of reality

🪏 to find out what's happened

Timeline of baseline plots

Baseline shouldn't move too much...

But it does

Baseline comparison for different parameter
                                combination

🎯 Ⅳ Localize your measures

🔗 Don't separate baseline and workload measurements

W+B using the new baseline

We can compare now!

comparación de versiones

Conclusions

Taming the energy bull requires measuring it first

The Brave New Algorithm provides an interesting framework for applying it to evolutionary algorithms

Follow best practices to make the measured workload stand out

Let literature

and Greek history and philosophy

be your guide

Σας ευχαριστώ πολύ!

And check out https://jj.github.io/brave-new-green-algorithm for additional content and code