@misc{indiciaed23dd40dbe83,
  title = {Leaky ReLUs That Differ in Forward and Backward Pass Facilitate Activation Maximization in Deep Neural Networks},
  author = {Christoph Linse and Erhardt Barth and Thomas Martinetz},
  year = {2024},
  doi = {10.1109/ijcnn60899.2024.10650881},
  url = {https://arxiv.org/abs/2410.16958},
  note = {Source identifier: 2410.16958}
}