Department of Physics, Stanford University, Stanford, CA 94305, USA.
Neuron. 2014 Sep 17;83(6):1253-5. doi: 10.1016/j.neuron.2014.08.006.
Experience with complex systems more primitive than the brain teaches important lessons about big data in biology. Chief among them is that physical laws, relationships among measured things that are always true, emerge out of chaos, not the other way around. Correct prediction (as opposed to incorrect prediction) from large data sets requires understanding of these laws. The reason is that the same processes that make them also make the system wildly error-intolerant if the errors are too large. This instability routinely causes computer simulations of even primitive systems to fail by enabling mistakes to cascade into ever worsening falsehoods. The more complex and sophisticated the system is, the more intolerant to errors it becomes.
对大脑等复杂系统的研究为生物学中的大数据提供了重要的经验教训。其中最重要的是，物理定律，即始终成立的测量事物之间的关系，是从混沌中产生的，而不是相反。要从大数据集中进行正确的预测（与错误的预测相对），就需要理解这些定律。原因是，如果错误太大，使这些定律产生的相同过程也会使系统对错误的容忍度变得极低。即使是对原始系统的计算机模拟，这种不稳定性也会导致错误级联成越来越严重的错误，从而导致模拟失败。系统越复杂和精细，它对错误的容忍度就越低。
