Proposes a differentiable Forth interpreter, integrating program sketches with neural slots, enabling complex sequence learning with end-to-end training.

Matko Bošnjak, Tim Rocktäschel, Jason Naradowsky et al.

Introduces Adaptive Computation Time (ACT) for RNNs, enabling dynamic step adjustment, significantly improving performance on synthetic tasks.

Alex Graves

Proposed a connection importance learning-based pruning method; reduced AlexNet parameters by 9× and VGG-16 by 13× without accuracy loss.

Song Han, Jeff Pool, John Tran et al.

Proposes a hybrid data-model parallel training method for CNNs, achieving over 6x speedup on 8 GPUs with minimal accuracy loss.

Alex Krizhevsky