Generative adversarial networks (GANs) are pow- erful generative models based on providing feed- back to a generative network via a discriminator network. However, the discriminator usually as- sesses individual samples. This prevents the dis- criminator from accessing global distributional statistics of generated samples, and often leads to mode dropping: the generator models only part of the target distribution. We propose to feed the discriminator with mixed batches of true and fake samples, and train it to predict the ratio of true samples in the batch. The latter score does not depend on the order of samples in a batch. Rather than learning this invariance, we introduce a generic permutation-invariant discriminator ar- chitecture. This architecture is provably a uni- versal approximator of all symmetric functions. Experimentally, our approach reduces mode col- lapse in GANs on two synthetic datasets, and obtains good results on the CIFAR10 and CelebA datasets, both qualitatively and quantitatively.

Motivation: Drug discovery demands rapid quantification of compound-protein interaction (CPI). However, there is a lack of methods that can predict compound-protein affinity from sequences alone with high applicability, accuracy, and interpretability. Results: We present a seamless integration of domain knowledges and learning-based approaches. Under novel representations of structurally-annotated protein sequences, a semi-supervised deep learning model that unifies recurrent and convolutional neural networks has been proposed to exploit both unlabeled and labeled data, for jointly encoding molecular representations and predicting affinities. Our representations and models outperform conventional options in achieving relative error in IC50 within 5-fold for test cases and 10-fold for protein classes not included for training. Performances for new protein classes with few labeled data are further improved by transfer learning. Furthermore, an attention mechanism is embedded to our model to add to its interpretability, as illustrated in case studies for predicting and explaining selective drug-target interactions.

We propose a computationally efficient estimator, formulated as a convex program, for a broad class of non-linear regression problems that involve difference of convex (DC) non-linearities. The proposed method can be viewed as a significant extension of the "anchored regression" method formulated and analyzed in [9] for regression with convex non-linearities. Our main assumption, in addition to other mild statistical and computational assumptions, is availability of a certain approximation oracle for the average of the gradients of the observation functions at a ground truth. Under this assumption and using a PAC-Bayesian analysis we show that the proposed estimator produces an accurate estimate with high probability. As a concrete example, we study the proposed framework in the bilinear regression problem with Gaussian factors and quantify a sufficient sample complexity for exact recovery. Furthermore, we describe a computationally tractable scheme that provably produces the required approximation oracle in the considered bilinear regression problem.

ABSTRACT This paper studies the effect of various hyper-parameters and their selection for the best performance of the deep learning model proposed in [1] for distributed attack detection in the Internet of Things (IoT). The findings show that there are three hyper-parameters that have more influence on the best performance achieved by the model. As a consequence, this study shows that the model's accuracy as reported in the paper is not achievable, based on the best selections of parameters, which is also supported by another recent publication [2]. INTRODUCTION Diro and Chilamkurti [1] have introduced a distributed deep neural network model for intrusion detection in the IoT environment. The primary principle behind the proposed model is to train the deep neural network model using multiple nodes in a distributed computing environment while parameter sharing and optimization is done through a coordinating master node.

We present a neural sequence model designed specifically for symbolic music. The model is based on a learned edit distance mechanism which generalises a classic recursion from computer science, leading to a neural dynamic program. Repeated motifs are detected by learning the transformations between them. We represent the arising computational dependencies using a novel data structure, the edit tree; this perspective suggests natural approximations which afford the scaling up of our otherwise cubic time algorithm. We demonstrate our model on real and synthetic data; in all cases it outperforms a strong stacked long short-term memory benchmark.

A group of Amazon shareholders is asking CEO Jeff Bezos to stop selling and marketing facial recognition technology to governments after civil liberties groups warned of the potential for abuse. Earlier this year, a group of advocacy organisations led by the American Civil Liberties Union (ACLU) published a report detailing how Amazon was marketing its Rekognition tool to American law enforcement agencies. In addition to touting the technology as helping to find suspects, Amazon has said it could be used to preemptively identify "persons of interest" and prevent crimes. A letter signed by 19 shareholders - and provided to The Independent by the ACLU - urges Mr Bezos to halt the tool's expansion until those concerns can be addressed. Amazon supplier investigated over'mistreatment' of workers in China How Alexa recorded a family's conversation then sent it to someone Amazon told to stop selling facial recognition tools to police Amazon supplier investigated over'mistreatment' of workers in China How Alexa recorded a family's conversation then sent it to someone Furnishing police and sheriff's departments with the tool would bolster "government surveillance infrastructure technology" and could drive down Amazon's value, the letter warned. It also echoed concerns about the potential for misuse. "While Rekognition may be intended to enhance some law enforcement activities, we are deeply concerned it may ultimately violate civil and human rights", the letter said.

Players as young as 13 are being exposed to gambling like behaviours through video game'loot boxes', researchers say. Experts are calling on regulators to impose greater controls on the boxes which give players random digital rewards, with similar tactics to a casino. These include weapons or costumes that can be acquired through gameplay or by spending real cash. Gamers favour them because they give a competitive advantage or power and some also have the potential to be sold on for money. Young players are being exposed to gambling-like behaviours through video game'loot boxes', researchers say.

Siri, do you believe in God? Siri, do you believe in God? Siri, I insist, do you believe in God? "I would ask that you address your spiritual questions to someone more qualified to comment. She – is it a she? – has a point: artificial intelligences (AI) like Siri are less situated than humans to answer questions about religion and spirituality. Existential angst, ethical inquiries, theological considerations: these belong exclusively to the domain of Homo Sapiens. But some futurists and tech experts predict a not-so-distant future in which AI, having achieved a certain indistinguishability from humans, will be truly intelligent. At that point, they claim, AI will experience the world in ways not too unlike the ways that we experience it – emotionally, intelligently, and spiritually. When that day comes, I'll have a new question for her. "Siri, do you have a soul?" A consideration of AI's religious status can be found in some of the earliest discussions of modern computing.

The expectation that artificial intelligence and automation will replace a significant number of human workers in the next decade is undisputed. What kinds of jobs and how soon, however, is still being worked out. In 2015, a CEDA report suggested 40 per cent of jobs in Australia were highly "susceptible to computerisation" in the next 15 years. Last year consultancy AlphaBeta said three million Australian jobs (around a third of all jobs) were at risk by 2030. The issue is a global one.

And for people worried their jobs might soon be gone, that might be scary -- but the threat could actually be much worse. It might affect our entire race. The modern-day dolls that not only look like women but have a similar feel, and they react like real people in conversations too. For those who find it's too hard to form a real relationship, will they just turn to a robot as a long-term solution? Armed with questions and recording equipment, my fear took me to Realbotixs in San Diego to meet Harmony, a robot with artificial intelligence (and a working vagina).