How to Think Like Einstein

Learning from Neural Nets

Deep learning and neural nets are revolutionizing the ability of computers to recognize patterns from data. Because the structure of neural nets is so similar to our own brains, they are also helping us to better understand how we think. Neural nets use arrays of artificial neurons to evaluate data. The neurons are not programmed.  Instead, they are trained with data that is similar to what they will be analyzing. 

Neural nets can be trained with rules that don't always work, just like our brains. In deep learning, this is sometimes referred to as over fitting. One of the biggest challenges in deep learning is mitigating over fitting. Without these mitigation strategies, neural nets do well with data like their training data but not on other data.  Many over-fitting mitigation strategies are similar to rule breaking strategies in How to Think Like Einstein.