We characterize the performance of standard and micro genetic algorithms when applied to focusing inside of an opaque media using feedback from a fluorescent guidestar particle. Using this feedback modality we find that the algorithms optimize more quickly (but to a lower enhancement) than when using reflective feedback, with the underlying mechanism being the fluorescence signal's multimodal nature and lower signal-to-noise ratio. We also find that both algorithms' performance decrease at very large numbers of bins due to decoherence effects related to a bin-dependent iteration time. To mitigate this effect we implement multithreaded functionality in the genetic algorithms and find that for our specific computer we obtain a 3.3x improvement in speed utilizing multithreading. These results demonstrate the usefulness of both algorithms for focusing inside of opaque media, which has applications in biological imaging and the study of subsurface chemical reactions in heterogeneous materials.

中文翻译：

---

用于在不透明介质内聚焦的遗传算法

我们使用来自荧光引导星粒子的反馈来表征标准和微遗传算法在应用于不透明介质内部聚焦时的性能。使用这种反馈模式，我们发现算法比使用反射反馈时优化得更快（但增强程度较低），其潜在机制是荧光信号的多模态性质和较低的信噪比。我们还发现，由于与 bin 相关的迭代时间相关的退相干效应，两种算法在大量 bin 下的性能都会下降。为了减轻这种影响，我们在遗传算法中实现了多线程功能，并发现对于我们的特定计算机，我们利用多线程将速度提高了 3.3 倍。