A new type of neural network is more interpretable

This new type of network learns functions rather than linear weights, allowing researchers to understand their behavior better. These networks could be the antidote to "black box" artificial intelligence, which could be particularly useful in helping scientists discover new laws.

Read more here: External Link