- Why does deep and cheap learning work so well?.
- Densely Connected Convolutional Networks.
- Stanford's Artificial Intelligence and Life in 2030.
- Visualizing and Understanding Convolutional Networks (2013).
- Machine Comprehension Using Match-LSTM and Answer Pointer.
- Deep Neural Networks for YouTube Recommendations.
- Decoupled Neural Interfaces using Synthetic Gradients. New DeepMind's paper reviewing backpropagation, using a modeled synthetic gradient in place of true backpropagated error gradients. Backpropagation is a a bottleneck in DNN, so the idea is instead use an asynchronous estimator, that would be obtained by supervised training of another mini …
- Decoupled Neural Interfaces using Synthetic Gradients.
- Explaining Deep Convolutional Neural Networks on Music Classification.
- Accelerating Eulerian Fluid Simulation With Convolutional Networks.
Classic references on Deep Reinforcement Learning:
- Self-Modification of Policy and Utility Function in Rational Agents (Everitt, et. al, 2016).
- WIKIREADING: A Novel Large-scale Language Understanding Task over Wikipedia (Hewlett, et. al, 2016).
- Evaluation of General-Purpose Artificial Intelligence: Why, What & How, (Bieger, et. al, 2016).
- Learning the Curriculum with Bayesian Optimization for Task-Specific Word Representation Learning …
- Instance Normalization: The Missing Ingredient for Fast Stylization (Ulyanov, et. al, 2016).
- Learning Semantic Deformation Flows with 3D Convolutional Networks .
- Costs of extinction risk mitigation. A Cost-Benefit Analysis of the extinction risk mitigation, claiming that the annual cost of reducing the probability of human extinction by 0.01 …
About an year ago, Google published a seminal paper named ImageNet Classification with Deep Convolutional Neural Networks, together with a blog post, which became known as Inceptionism. This work unveiled not only a new way of composing hallucinating artistic pictures, but astonishing new insights on how convolutional neural networks work …more ...
This is a selection of quintessential papers for anyone starting on Deep Learning (Thanks to Joe Zimmerman):
- ImageNet Classification with Deep Convolutional Neural Networks (Krizhevsky, et al., 2014). AlexNet.
- Very Deep Convolutional Networks for large-scale image recognition (Simonyan, et al., 2014). Image classification.
- Improving neural networks by preventing co-adaptation …
- The Science of Talking with Computers
- Megan Smith: Perspectives on artificial intelligence from the White House.
- NVIDIA Deep Learning Course: Class #1 – Introduction to Deep Learning.
Mastering the Game of Go with Deep Neural Networks and Tree Search. "All games of perfect information have an optimal value function which determines the outcome of the game". , Basically:
- Values networks to evaluate board positions and policy networks to select moves.
- Trained with supervised learning from human expert …
I was seven when I first saw the shiny beige box. When my uncle finally let me touch the black screen, I was astounded with the list of unfathomable keys that “help” would print. My parents were getting divorced and my reality had just became 40 megabytes of the most …more ...