Google CEO Sundar Pichai was obviously excited when he spoke to developers about a blockbuster result from his machine-learning lab earlier this month. Researchers had figured out how to automate some of the work of crafting machine-learning software, something that could make it much easier to deploy the technology in new situations and industries. But the project had already gained a reputation among AI researchers for another reason: the way it illustrated the vast computing resources needed to compete at the cutting edge of machine learning. A paper from Google's researchers says they simultaneously used as many as 800 of the powerful and expensive graphics processors that have been crucial to the recent uptick in the power of machine learning (see "10 Breakthrough Technologies 2013: Deep Learning"). They told MIT Technology Review that the project had tied up hundreds of the chips for two weeks solid--making the technique too resource-intensive to be more than a research project even at Google.

The Google-owned computer algorithm AlphaGo is retiring from playing humans in the ancient Chinese game of Go after defeating the world's top player this week. AlphaGo defeated 19-year-old world number one Ke Jie of China on Saturday to sweep a three-game series that was closely watched as a measure of how far artificial intelligence (AI) has come. Ke Jie anointed the program as the new'Go god' after his defeat. AlphaGo last year became the first computer programme to beat an elite player in a full Go match, and its successes have been hailed as groundbreaking due to the game's complexity. Go has an incomputable number of moves, putting a premium on human-like'intuition' and strategy.

Data preprocessing is a fundamental part of any machine learning project and often more time is spent on the data preparation than on the actual machine learning. While some preprocessing tasks are problem specific many others such as partitioning data into training and test folds, stratifying samples or building mini-batches are generic. The following Canonical Pipeline shows the processing steps common for deep-learning in vision. A Reader reads sample data stored in text files, Excel or Pandas tables. The Splitter then partitions data into training, validation and test folds and performs stratification if needed.

We investigate the generalizability of deep learning based on the sensitivity to input perturbation. We hypothesize that the high sensitivity to the perturbation of data degrades the performance on it. To reduce the sensitivity to perturbation, we propose a simple and effective regularization method, referred to as spectral norm regularization, which penalizes the high spectral norm of weight matrices in neural networks. We provide supportive evidence for the abovementioned hypothesis by experimentally confirming that the models trained using spectral norm regularization exhibit better generalizability than other baseline methods.

Magnetic resonance image (MRI) reconstruction is a severely ill-posed linear inverse task demanding time and resource intensive computations that can substantially trade off {\it accuracy} for {\it speed} in real-time imaging. In addition, state-of-the-art compressed sensing (CS) analytics are not cognizant of the image {\it diagnostic quality}. To cope with these challenges we put forth a novel CS framework that permeates benefits from generative adversarial networks (GAN) to train a (low-dimensional) manifold of diagnostic-quality MR images from historical patients. Leveraging a mixture of least-squares (LS) GANs and pixel-wise $\ell_1$ cost, a deep residual network with skip connections is trained as the generator that learns to remove the {\it aliasing} artifacts by projecting onto the manifold. LSGAN learns the texture details, while $\ell_1$ controls the high-frequency noise. A multilayer convolutional neural network is then jointly trained based on diagnostic quality images to discriminate the projection quality. The test phase performs feed-forward propagation over the generator network that demands a very low computational overhead. Extensive evaluations are performed on a large contrast-enhanced MR dataset of pediatric patients. In particular, images rated based on expert radiologists corroborate that GANCS retrieves high contrast images with detailed texture relative to conventional CS, and pixel-wise schemes. In addition, it offers reconstruction under a few milliseconds, two orders of magnitude faster than state-of-the-art CS-MRI schemes.

Partially observable environments present an important open challenge in the domain of sequential control learning with delayed rewards. Despite numerous attempts during the two last decades, the majority of reinforcement learning algorithms and associated approximate models, applied to this context, still assume Markovian state transitions. In this paper, we explore the use of a recently proposed attention-based model, the Gated End-to-End Memory Network, for sequential control. We call the resulting model the Gated End-to-End Memory Policy Network. More precisely, we use a model-free value-based algorithm to learn policies for partially observed domains using this memory-enhanced neural network. This model is end-to-end learnable and it features unbounded memory. Indeed, because of its attention mechanism and associated non-parametric memory, the proposed model allows us to define an attention mechanism over the observation stream unlike recurrent models. We show encouraging results that illustrate the capability of our attention-based model in the context of the continuous-state non-stationary control problem of stock trading. We also present an OpenAI Gym environment for simulated stock exchange and explain its relevance as a benchmark for the field of non-Markovian decision process learning.

We propose a new equilibrium enforcing method paired with a loss derived from the Wasserstein distance for training auto-encoder based Generative Adversarial Networks. This method balances the generator and discriminator during training. Additionally, it provides a new approximate convergence measure, fast and stable training and high visual quality. We also derive a way of controlling the trade-off between image diversity and visual quality. We focus on the image generation task, setting a new milestone in visual quality, even at higher resolutions. This is achieved while using a relatively simple model architecture and a standard training procedure.

In this paper, we propose gcForest, a decision tree ensemble approach with performance highly competitive to deep neural networks. In contrast to deep neural networks which require great effort in hyper-parameter tuning, gcForest is much easier to train. Actually, even when gcForest is applied to different data from different domains, excellent performance can be achieved by almost same settings of hyper-parameters. The training process of gcForest is efficient and scalable. In our experiments its training time running on a PC is comparable to that of deep neural networks running with GPU facilities, and the efficiency advantage may be more apparent because gcForest is naturally apt to parallel implementation. Furthermore, in contrast to deep neural networks which require large-scale training data, gcForest can work well even when there are only small-scale training data. Moreover, as a tree-based approach, gcForest should be easier for theoretical analysis than deep neural networks.

After AlphaGo's historic victory against South Korean grandmaster Lee Sedol in March 2016, Go teacher Jianlun Qian felt a sense of impending crisis. He fretted about the demise of the game brought about by AI. Now that a more powerful AlphaGo has beaten the world's number one player, though, Qian feels differently. At DeepMind's Go summit in Wuzhen last week, Qian, a teacher at the local Go association's training center, contemplated a very different future in which humans and AI can complement each other. "I'm indifferent to the results now," said Qian, who coaches about 50 preschoolers in Wuzhen.

Keras is a high-level neural networks library, written in Python and capable of running on top of either TensorFlow or Theano. It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research. Keras is compatible with: Python 2.7-3.5. A model is understood as a sequence or a graph of standalone, fully-configurable modules that can be plugged together with as little restrictions as possible.