I had some crazy ideas recently after I took Andrew Ng’s course. I consider that maybe the universe is like deep neural networks, which is not organized by low-dimensional space but organized by high dimensional space. Indeed, humans do not easily observe the high dimensional space. The universe is like a deep neural network’s high-dimensional space is not observed by humans’ intuition but not means it does not exist.
我上了吴安德的课程后，最近有一些疯狂的想法。我认为宇宙也许就像是深度神经网络，不是由低维空间组织的，而是由高维空间组织的。 的确，人类不容易观察高维空间。 宇宙就像一个深层神经网络的高维空间，不是人类的直觉所观察到的，但这并不意味着它不存在。
In this case, the whole universe is like a non-convex function, and there are many local optimization points. From mathematics, we know that the derivate of local optimization points is 0. In other words, if some matters exist in those local optimization points, the gradient (rate of change) is 0. Thus, those matters in the local optimization point will not change, but some matters are not in the local optimization point will change a lot. If this idea is right, many scientists can observe some things in some local optimization points. It may easily dig more secrets from the universe.
在这种情况下，整个宇宙就像一个非凸函数并且有许多局部优化点。 从数学上，我们知道局部优化点的导数为0。换句话说，如果局部优化点中存在某些物质，则梯度（变化率）为0。因此，局部优化点中的那些物质不会变化，但是有些事情不在本地优化点上会改变很多。 如果这个想法正确，那么许多科学家可以在某些局部优化点中观察到一些东西。 它很容易从宇宙中挖掘更多的秘密。
The above is a crazy idea. You can treat it like a joke if you like.
以上是一个疯狂的主意，如果您愿意，您可以像个玩笑一样对待。