How long can a person go without sleep? The world record is apparently 11 days. However, when Randy Gardner set that record in 1965, he may have been awake during the time, but he was basically'cognitively dysfunctional'. If perhaps Gardner went beyond two weeks he would likely have died. Some animals appear to be awake all the time. Cetaceans) need to remain awake because they need to periodically come up to the surface to breathe oxygen.

Google has been testing its latest machine learning system, called DeepMind, to explore the limits of game theory, using AI as a mirror to analyze how we respond to cooperative and competitive situations. The team found that in highly competitive environments, DeepMind will use "highly aggressive" tactics to either win the game or cause as much damage it can to its opponent. The premise for this game was collecting apples, and when apples became scarce, the AI would turn on each other and try to steal as many as possible. When testing the same principle in another game where the AI were taught to cooperate with one another, however, the AI behaved more socially, showing that the AI evaluated its environment to decide the optimal strategy for survival. Using this knowledge, Google can now construct environments and learning scenarios that help AI develop cooperative behaviors.

A new paper in the Proceedings of the National Academy of Sciences (PNAS) reports how a cutting-edge artificial intelligence technique called deep learning can automatically identify, count and describe animals in their natural habitats. Photographs that are automatically collected by motion-sensor cameras can then be automatically described by deep neural networks. The result is a system that can automate animal identification for up to 99.3 percent of images while still performing at the same 96.6 percent accuracy rate of crowdsourced teams of human volunteers. "This technology lets us accurately, unobtrusively and inexpensively collect wildlife data, which could help catalyze the transformation of many fields of ecology, wildlife biology, zoology, conservation biology and animal behavior into'big data' sciences. This will dramatically improve our ability to both study and conserve wildlife and precious ecosystems," says Jeff Clune, the senior author of the paper.

At Connect() developer event last year, Microsoft announced the preview of Azure Databricks service for highest-performance streaming analytics projects. Azure Databricks is an Apache Spark-based analytics platform that delivers one-click setup, streamlined workflows and an interactive workspace. Earlier this year, Microsoft announced the general availability of this service. At the Spark AI Summit yesterday, Microsoft announced a couple of new enhancements to Azure Databricks. First, they are adding support for GPU enabled VMs.

Recently, I gave a talk at the O'Reilly AI conference in Beijing about some of the interesting lessons we've learned in the world of NLP. While there, I was lucky enough to attend a tutorial on Deep Reinforcement Learning (Deep RL) from scratch by Unity Technologies. I thought that the session, led by Arthur Juliani, was extremely informative and wanted to share some big takeaways below. In our conversations with companies, we've seen a rise of interesting Deep RL applications, tools and results. In parallel, the inner workings and applications of Deep RL, such as AlphaGo pictured above, can often seem esoteric and hard to understand.

This study targets to automatically annotate on arrhythmia by deep network. The investigated types include sinus rhythm, asystole (Asys), supraventricular tachycardia (Tachy), ventricular flutter or fibrillation (VF/VFL), ventricular tachycardia (VT). Methods: 13s limb lead ECG chunks from MIT malignant ventricular arrhythmia database (VFDB) and MIT normal sinus rhythm database were partitioned into subsets for 5-fold cross validation. These signals were resampled to 200Hz, filtered to remove baseline wandering, projected to 2D gray spectrum and then fed into a deep network with brand-new structure. In this network, a feature vector for a single time point was retrieved by residual layers, from which latent representation was extracted by variational autoencoder (VAE). These front portions were trained to meet a certain threshold in loss function, then fixed while training procedure switched to remaining bidirectional recurrent neural network (RNN), the very portions to predict an arrhythmia category. Attention windows were polynomial lumped on RNN outputs for learning from details to outlines. And over sampling was employed for imbalanced data. The trained model was wrapped into docker image for deployment in edge or cloud. Conclusion: Promising sensitivities were achieved in four arrhythmias and good precision rates in two ventricular arrhythmias were also observed. Moreover, it was proven that latent representation by VAE, can significantly boost the speed of convergence and accuracy.

This paper proposes and implements an intuitive and pervasive solution for neonatal EEG monitoring assisted by sonification and deep learning AI that provides information about neonatal brain health to all neonatal healthcare professionals, particularly those without EEG interpretation expertise. The system aims to increase the demographic of clinicians capable of diagnosing abnormalities in neonatal EEG. The proposed system uses a low-cost and low-power EEG acquisition system. An Android app provides single-channel EEG visualization, traffic-light indication of the presence of neonatal seizures provided by a trained, deep convolutional neural network and an algorithm for EEG sonification, designed to facilitate the perception of changes in EEG morphology specific to neonatal seizures. The multifaceted EEG interpretation framework is presented and the implemented mobile platform architecture is analyzed with respect to its power consumption and accuracy.

Meta-learning enables a model to learn from very limited data to undertake a new task. In this paper, we study the general meta-learning with adversarial samples. We present a meta-learning algorithm, ADML (ADversarial Meta-Learner), which leverages clean and adversarial samples to optimize the initialization of a learning model in an adversarial manner. ADML leads to the following desirable properties: 1) it turns out to be very effective even in the cases with only clean samples; 2) it is model-agnostic, i.e., it is compatible with any learning model that can be trained with gradient descent; and most importantly, 3) it is robust to adversarial samples, i.e., unlike other meta-learning methods, it only leads to a minor performance degradation when there are adversarial samples. We show via extensive experiments that ADML delivers the state-of-the-art performance on two widely-used image datasets, MiniImageNet and CIFAR100, in terms of both accuracy and robustness.

Cyber attacks are growing in frequency and severity. Over the past year alone we have witnessed massive data breaches that stole personal information of millions of people and wide-scale ransomware attacks that paralyzed critical infrastructure of several countries. Combating the rising cyber threat calls for a multi-pronged strategy, which includes predicting when these attacks will occur. The intuition driving our approach is this: during the planning and preparation stages, hackers leave digital traces of their activities on both the surface web and dark web in the form of discussions on platforms like hacker forums, social media, blogs and the like. These data provide predictive signals that allow anticipating cyber attacks. In this paper, we describe machine learning techniques based on deep neural networks and autoregressive time series models that leverage external signals from publicly available Web sources to forecast cyber attacks. Performance of our framework across ground truth data over real-world forecasting tasks shows that our methods yield a significant lift or increase of F1 for the top signals on predicted cyber attacks. Our results suggest that, when deployed, our system will be able to provide an effective line of defense against various types of targeted cyber attacks.

We propose a new model for digital pathology segmentation, based on the observation that histopathology images are inherently symmetric under rotation and reflection. Utilizing recent findings on rotation equivariant CNNs, the proposed model leverages these symmetries in a principled manner. We present a visual analysis showing improved stability on predictions, and demonstrate that exploiting rotation equivariance significantly improves tumor detection performance on a challenging lymph node metastases dataset. We further present a novel derived dataset to enable principled comparison of machine learning models, in combination with an initial benchmark. Through this dataset, the task of histopathology diagnosis becomes accessible as a challenging benchmark for fundamental machine learning research.