Recently deep reinforcement learning (DRL) has achieved outstanding success on solving many difficult and large-scale RL problems. However the high sample cost required for effective learning often makes DRL unaffordable in resource-limited applications. With the aim of improving sample efficiency and learning performance, we will develop a new DRL algorithm in this paper that seamless integrates entropy-induced and bootstrap-induced techniques for efficient and deep exploration of the learning environment. Specifically, a general form of Tsallis entropy regularizer will be utilized to drive entropy-induced exploration based on efficient approximation of optimal action-selection policies. Different from many existing works that rely on action dithering strategies for exploration, our algorithm is efficient in exploring actions with clear exploration value. Meanwhile, by employing an ensemble of Q-networks under varied Tsallis entropy regularization, the diversity of the ensemble can be further enhanced to enable effective bootstrap-induced exploration. Experiments on Atari game playing tasks clearly demonstrate that our new algorithm can achieve more efficient and effective exploration for DRL, in comparison to recently proposed exploration methods including Bootstrapped Deep Q-Network and UCB Q-Ensemble.

While modern machine translation has relied on large parallel corpora, a recent line of work has managed to train Neural Machine Translation (NMT) systems from monolingual corpora only (Artetxe et al., 2018c; Lample et al., 2018). Despite the potential of this approach for low-resource settings, existing systems are far behind their supervised counterparts, limiting their practical interest. In this paper, we propose an alternative approach based on phrase-based Statistical Machine Translation (SMT) that significantly closes the gap with supervised systems. Our method profits from the modular architecture of SMT: we first induce a phrase table from monolingual corpora through cross-lingual embedding mappings, combine it with an n-gram language model, and fine-tune hyperparameters through an unsupervised MERT variant. In addition, iterative backtranslation improves results further, yielding, for instance, 14.08 and 26.22 BLEU points in WMT 2014 English-German and English-French, respectively, an improvement of more than 7-10 BLEU points over previous unsupervised systems, and closing the gap with supervised SMT (Moses trained on Europarl) down to 2-5 BLEU points. Our implementation is available at https:// github.com/artetxem/monoses.

Most textual entailment models focus on lexical gaps between the premise text and the hypothesis, but rarely on knowledge gaps. We focus on filling these knowledge gaps in the Science Entailment task, by leveraging an external structured knowledge base (KB) of science facts. Our new architecture combines standard neural entailment models with a knowledge lookup module. To facilitate this lookup, we propose a fact-level decomposition of the hypothesis, and verifying the resulting sub-facts against both the textual premise and the structured KB. Our model, NSnet, learns to aggregate predictions from these heterogeneous data formats. On the SciTail dataset, NSnet outperforms a simpler combination of the two predictions by 3% and the base entailment model by 5%.

We propose KDSL, a new word sense disambiguation (WSD) framework that utilizes knowledge to automatically generate sense-labeled data for supervised learning. First, from WordNet, we automatically construct a semantic knowledge base called DisDict, which provides refined feature words that highlight the differences among word senses, i.e., synsets. Second, we automatically generate new sense-labeled data by DisDict from unlabeled corpora. Third, these generated data, together with manually labeled data and unlabeled data, are fed to a neural framework conducting supervised and unsupervised learning jointly to model the semantic relations among synsets, feature words and their contexts. The experimental results show that KDSL outperforms several representative state-of-the-art methods on various major benchmarks. Interestingly, it performs relatively well even when manually labeled data is unavailable, thus provides a potential solution for similar tasks in a lack of manual annotations.

We aim to predict whether a primary school student will perform in the `below standard' band of a national standardized test. We exploit a data set containing test performance on the National Assessment Program - Literacy and Numeracy (NAPLAN); a test given annually to all Australian school students in grades 3, 5, 7, and 9. We separate the analysis into students in grade 5 and above, for which previous achievement may be used as a predictor; and students in grade 3, which must rely on family- and school-level predictors only. We train and compare a set of classifiers for reading and numeracy learning areas respectively. The classifiers achieve good predictive power in terms of area under the ROC curve, suggesting that it is feasible for schools to more accurately screen a large number of students for academic risk.

Distant supervision is a popular method for performing relation extraction from text that is known to produce noisy labels. Most progress in relation extraction and classification has been made with crowdsourced corrections to distant-supervised labels, and there is evidence that indicates still more would be better. In this paper, we explore the problem of propagating human annotation signals gathered for open-domain relation classification through the CrowdTruth methodology for crowdsourcing, that captures ambiguity in annotations by measuring inter-annotator disagreement. Our approach propagates annotations to sentences that are similar in a low dimensional embedding space, expanding the number of labels by two orders of magnitude. Our experiments show significant improvement in a sentence-level multi-class relation classifier.

A machine-learning study that analysed hundreds of thousands of earthquakes beat the standard method at predicting the location of aftershocks. Scientists say that the work provides a fresh way of exploring how changes in ground stress, such as those that occur during a big earthquake, trigger the quakes that follow. It could also help researchers to develop new methods for assessing seismic risk. "We've really just scratched the surface of what machine learning may be able to do for aftershock forecasting," says Phoebe DeVries, a seismologist at Harvard University in Cambridge, Massachusetts. She and her colleagues report their findings on 29 August in Nature.

Scientists have used artificial intelligence to predict how cancers will progress and evolve, which could help doctors design the most effective treatment for each patient. A team led by the Institute of Cancer Research, London (ICR) and the University of Edinburgh developed a new technique known as Revolver (Repeated evolution of cancer), which picks out patterns in DNA mutation within cancers and uses the information to forecast future genetic changes. They said the ever-changing nature of tumours is one of the biggest challenges of treatment – with cancers often evolving to a drug-resistant form. Parents think children should be taught signs of cancer, poll finds'Exciting' cancer drug combination shrinks tumours and stops growth Parental cancer has lifetime impact on children's education and earnin Vaping causes DNA mutations which could lead to cancer, says study'Exciting' cancer drug combination shrinks tumours and stops growth Parental cancer has lifetime impact on children's education and earnin But if doctors can predict how a tumour will evolve, they could intervene earlier to stop cancer in its tracks before it has had a chance to evolve or develop resistance, increasing the patient's chances of survival. The team also found a link between certain sequences of repeated tumour mutations and survival outcome.

SYDNEY – A robot submarine able to hunt and kill the predatory crown-of-thorns starfish that are devastating the Great Barrier Reef was unveiled by Australian researchers on Friday. Scientists at Queensland University of Technology (QUT) said the robot, named the RangerBot and developed with a grant from Google, would serve as a "robo reef protector" for the vast World Heritage site off Australia's northeastern coast. The RangerBot has an eight-hour battery life and computer vision capabilities allowing it to monitor and map reef areas at scales not previously possible. "RangerBot is the world's first underwater robotic system designed specifically for coral reef environments, using only robot-vision for real-time navigation, obstacle avoidance and complex science missions," said Matthew Dunbabin, the QUT professor who unveiled the submarine. "This multifunction ocean drone can monitor a wide range of issues facing coral reefs including coral bleaching, water quality, pest species, pollution and siltation."

Whether you do your shopping online or in store, your retail experience is the latest battleground for the artificial intelligence (AI) and machine learning revolution. Major Australian retailers have begun to realise they have a lot to gain from getting their AI strategy right, with Woolworths currently openly recruiting a head of AI and machine learning to be supported by a team of data scientists. The newly developed Woolworths division WooliesX aims to bring together a diverse group of teams -- including technology, customer digital experience, e-commerce, financial services and digital customer experience. To understand the opportunities and threats for all major retailers, it's useful to understand why artificial intelligence is back on the agenda. Two crucial things have changed since the initial forays into AI decades ago: data and computing power.