ReLU
Software / App
activation function defined as the identity function for positive arguments and zero for negative arguments
Mentioned in 3 videos
Save the 3 videos on ReLU to your own pod.
Sign up free to keep building your knowledge base on ReLU as more episodes are added.
Videos Mentioning ReLU
Ep 18: Petaflops to the People — with George Hotz of tinycorp
Latent Space
George Hotz dislikes the object-oriented implementation of ReLU in PyTorch, preferring a functional approach.
deeplearning.ai's Heroes of Deep Learning: Yoshua Bengio
DeepLearningAI
Rectified Linear Unit, an activation function that worked better than expected in deep nets, surprising Bengio.
MIT 6.S094: Recurrent Neural Networks for Steering Through Time
Lex Fridman
Rectified Linear Unit, a popular activation function in neural networks, discussed in the context of potential issues like zero gradients.