Layton Hayes
Friday, March 2, 2018
Memory-Augmented Neural Networks
Memory Augmented Neural Networks (MANNs) are a relatively new class of architectures that improve the memory performance of neural networks significantly beyond that of traditional Recurrent Neural Networks. They have exciting applications for learned planning, meta learning, and one-shot learning. They also achieve state-of-the-art performance on the bAbI dataset, which is a task suite designed to help advance towards general AI.
Layton is a Master's student in the AI institute. He spent last summer interning at Amazon and climbing mountains in the PNW, and this summer he'll be at Facebook and climbing the Sierra Nevadas. His research interests focus on Deep Reinforcement Learning, MANNs, and Human-Machine Interface. His real interests focus on space-future, dank memes, and building superhero gadgets.