Neural networks discover meaningful representations of the data through the process of learning. Here, the authors explore how these representations are affected by scaling the network output or modifying the activation functions.
- Alexander van Meegen
- Haim Sompolinsky
