Learning a generative model from partial data (data with missingness) is a challenging area of machine learning research. We study a specific implementation of the Auto-Encoding Variational Bayes (AEVB) algorithm, named in this paper as a Variational Auto-Decoder (VAD). VAD is a generic framework which uses Variational Bayes and Markov Chain Monte Carlo (MCMC) methods to learn a generative model from partial data. The main distinction between VAD and Variational Auto-Encoder (VAE) is the encoder component, as VAD does not have one. Using a proposed efficient inference method from a multivariate Gaussian approximate posterior, VAD models allow inference to be performed via simple gradient ascent rather than MCMC sampling from a probabilistic decoder. This technique reduces the inference computational cost, allows for using more complex optimization techniques during latent space inference (which are shown to be crucial due to a high degree of freedom in the VAD latent space), and keeps the framework simple to implement. Through extensive experiments over several datasets and different missing ratios, we show that encoders cannot efficiently marginalize the input volatility caused by imputed missing values. We study multimodal datasets in this paper, which is a particular area of impact for VAD models.

Large-scale distributed training of neural networks is often limited by network bandwidth, wherein the communication time overwhelms the local computation time. Motivated by the success of sketching methods in sub-linear/streaming algorithms, we propose a sketching-based approach to minimize the communication costs between nodes without losing accuracy. In our proposed method, workers in a distributed, synchronous training setting send sketches of their gradient vectors to the parameter server instead of the full gradient vector. Leveraging the theoretical properties of sketches, we show that this method recovers the favorable convergence guarantees of single-machine top-$k$ SGD. Furthermore, when applied to a model with $d$ dimensions on $W$ workers, our method requires only $\Theta(kW)$ bytes of communication, compared to $\Omega(dW)$ for vanilla distributed SGD. To validate our method, we run experiments using a residual network trained on the CIFAR-10 dataset. We achieve no drop in validation accuracy with a compression ratio of 4, or about 1 percentage point drop with a compression ratio of 8. We also demonstrate that our method scales to many workers.

We propose a new low-cost machine-learning-based methodology which assists designers in reducing the gap between the problem and the solution in the design process. Our work applies reinforcement learning (RL) to find the optimal task-oriented design solution through the construction of the design action for each task. For this task-oriented design, the 3D design process in product design is assigned to an action space in Deep RL, and the desired 3D model is obtained by training each design action according to the task. By showing that this method achieves satisfactory design even when applied to a task pursuing multiple goals, we suggest the direction of how machine learning can contribute to the design process. Also, we have validated with product designers that this methodology can assist the creative part in the process of design.

Apple has confirmed its next major event, with a major launch later this month. The event will happen on 25 March, the company announced, alongside a teaser that made heavy reference to films. The announcement does not give away any details of what the company is getting ready to launch at the event. But it appears to confirm one major rumour about what is coming. We'll tell you what's true.

Today, in these vision centers, technicians take eye scans and send them to doctors in Madurai for review. Automated diagnosis can streamline and expand the process, reaching more people in more places -- the kind of "McDonaldization" espoused by Dr. V. The technology still faces regulatory hurdles in India, in part because of the difficulty of navigating the country's bureaucracy. And though Google's eye system is now certified for use in Europe, it is still awaiting approval in the United States. Luke Oakden-Rayner, the director of medical imaging research at the Royal Adelaide Hospital in Australia, said these systems might even need new regulatory frameworks because existing rules weren't always sufficient. "I am not convinced that people care enough about the safety of these systems," he said.

Details about the highly anticipated augmented reality (AR) game Harry Potter: Wizards Unite, have been revealed by the makers, who also created Pokémon Go. Developer Niantic partnered with Warner Brothers Interactive to make the Harry Potter game, using its AR expertise to bring wizards and mythical beasts into the real world. Images of the gameplay and several teaser videos were shared through the game's website, although its release date was not revealed. We'll tell you what's true. You can form your own view.

Tim Berners-Lee helped invent the world wide web 30 years ago. And he has consistently pointed out that the original dream that gave rise to it is under threat. It is exactly 30 years since Sir Tim submitted a paper to his colleagues at CERN, suggesting a way of sharing data across networks, under the title "Information Management: A Proposal". The humble title belies the importance of what was contained inside, which included a complete sketch for the networked information system that would on to become the internet we know today. That would take some time, however.

The government will soon announce the start of a wide-ranging plan to ban people from viewing pornography unless they prove their identity. The plans, referred to as porn blocks, will attempt to stop children from looking at adult content by establishing age verification procedures. The most popular of those procedures will require people to register and show official documents and then log in whenever they look at pornographic websites. Numerous reports have suggested that the plan will go into place on 1 April. But the government says a final date is not yet set, though it will be announced soon.

Russian companies are leaving European and U.S. firms behind with some 30% employing artificial intelligence (AI) technology, a Microsoft survey of business leaders has shown, according to The Straits Times. The tech giant's survey also revealed that Russian executives are at least 7.7 percent ahead when it comes to adopting the technology. In France, for example, the level of AI use amounts to 10 percent, Microsoft said. The survey, titled "Business Leaders in the Age of AI" was published last week. Some 800 executives from eight countries – France, Germany, Italy, the Netherlands, Russia, Switzerland, the UK and the US – took part in the survey, which was conducted back in January among companies with more than 250 employees.

Tumblr's ban on pornography and adult content has led to a fifth of its users deserting the platform, figures reveal. The ban, which came into effect on 17 December, provoked a backlash from users who claimed it would penalise sex-positive, LGBT and NSFW art communities. Visits to the Tumblr website fell from 521 million in December to 437 million in January, according to data from web analytics firm SimilarWeb. We'll tell you what's true. You can form your own view.