Modeling multivariate time series has long been a subject that has attracted researchers from a diverse range of fields including economics, finance, and traffic. A basic assumption behind multivariate time series forecasting is that its variables depend on one another but, upon looking closely, it is fair to say that existing methods fail to fully exploit latent spatial dependencies between pairs of variables. In recent years, meanwhile, graph neural networks (GNNs) have shown high capability in handling relational dependencies. GNNs require well-defined graph structures for information propagation which means they cannot be applied directly for multivariate time series where the dependencies are not known in advance. In this paper, we propose a general graph neural network framework designed specifically for multivariate time series data. Our approach automatically extracts the uni-directed relations among variables through a graph learning module, into which external knowledge like variable attributes can be easily integrated. A novel mix-hop propagation layer and a dilated inception layer are further proposed to capture the spatial and temporal dependencies within the time series. The graph learning, graph convolution, and temporal convolution modules are jointly learned in an end-to-end framework. Experimental results show that our proposed model outperforms the state-of-the-art baseline methods on 3 of 4 benchmark datasets and achieves on-par performance with other approaches on two traffic datasets which provide extra structural information.

There are a wide range of applications that involve multi-modal data, such as cross-modal retrieval, visual question-answering, and image captioning. Such applications are primarily dependent on aligned distributions of the different constituent modalities. Existing approaches generate latent embeddings for each modality in a joint fashion by representing them in a common manifold. However these joint embedding spaces fail to sufficiently reduce the modality gap, which affects the performance in downstream tasks. We hypothesize that these embeddings retain the intra-class relationships but are unable to preserve the inter-class dynamics. In this paper, we present a novel framework COBRA that aims to train two modalities (image and text) in a joint fashion inspired by the Contrastive Predictive Coding (CPC) and Noise Contrastive Estimation (NCE) paradigms which preserve both inter and intra-class relationships. We empirically show that this framework reduces the modality gap significantly and generates a robust and task agnostic joint-embedding space. We outperform existing work on four diverse downstream tasks spanning across seven benchmark cross-modal datasets.

Delaying the reopening of primary schools in England on 1 June by two weeks could halve the risk to each child of being exposed to an infectious classmate, according to a report by the Independent Scientific Advisory Group for Emergencies, a recently-formed group of scientists that is seeking to provide alternative advice to the UK government. The group say that modelling suggests that waiting until September would reduce this risk further, to less than the risk to children of road traffic accidents. The group is chaired by former government chief scientific advisor David King and is separate from the official SAGE committee that advises the UK government. "The crucial factor allowing school reopening around the world has been the presence of well-functioning local test, trace and isolate protocols – something that is now accepted will not be in place in England by early June," the report says. It adds that before schools can reopen, it is important to confirm that daily new ...

A deep-learning based hybrid strategy for short-term load forecasting is presented. The strategy proposes a novel tree-based ensemble method Warm-start Gradient Tree Boosting (WGTB). Current strategies either ensemble submodels of a single type, which fail to take advantage of statistical strengths of different inference models. Or they simply sum the outputs from completely different inference models, which doesn't maximize the potential of ensemble. WGTB is thus proposed and tailored to the great disparity among different inference models in accuracy, volatility and linearity. The complete strategy integrates four different inference models (i.e., auto-regressive integrated moving average, nu support vector regression, extreme learning machine and long short-term memory neural network), both linear and nonlinear models. WGTB then ensembles their outputs by hybridizing linear estimator ElasticNet and nonlinear estimator ExtraTree via boosting algorithm. It is validated on the real historical data of a grid from State Grid Corporation of China of hourly resolution. The result demonstrates the effectiveness of the proposed strategy that hybridizes statistical strengths of both linear and nonlinear inference models.

Customer experience automation vendor, ActiveCampaign, has secured US$100 million in its latest series B investment round. Key investors this time include Susquehanna and Silversmith Capital Partners, which was the sole sponsor in the series A funding round in 2016. ActiveCampaign said it will use the funding to develop its CXA category through product innovations to advance beyond legacy marketing automation, traditional CRM and service technology, while continuing international expansion and building on its customer success team and partner ecosystem. ActiveCampaign is pitching its platform across the spectrum of small, midsize, and enterprise businesses and has 90,000 customers spanning 161 countries. Since its last funding round, the company said it had increased annual recurring revenue six-fold to more than $90 million, opened new offices in several locations including Sydney, and grown its employee base to 550 staff.

Farmers in New Zealand have used a four-legged robot to herd sheep, patrol fields and perform other agricultural tasks. The feats were carried out as part of a demonstration of Spot – a robotic dog developed by Massachusetts-based engineering firm Boston Dynamics. Equipped with software developed by robotics company Rocos, Spot was controlled remotely to shepherd sheep across a mountainside. "The age of autonomous robots is upon us," claimed Rocos chief executive David Inggs. "Our customers are augmenting their human workforces to automate physical processes that are often dull, dirty, or dangerous. Organisations can now design, schedule and manage inspection missions remotely."

Researchers in Australia have achieved a world record internet speed of 44.2 terabits per second, allowing users to download 1,000 HD movies in a single second. A team from Monash, Swinburne and RMIT universities used a'micro-comb' optical chip containing hundreds of infrared lasers to transfer data across existing communications infrastructure in Melbourne. The highest commercial internet speed anywhere in the world is currently in Singapore, where the average download speed is 197.3 megabits per second (mbps). In Australia, the average download speed is 43.4 mbps – one million-times slower than the speeds achieved in the latest test. "There's a bit of a global race on at the moment to get this technology to a commercial stage, as the'micro-comb' at its heart is useful in a really broad range of existing technologies," Dr Bill Corcoran from Monash University, told The Independent.

The New Zealand software company Rocos is training a Boston Dynamics-designed robot called Spot to work on farms to help'relieve the strain of worker shortages, and create precision in farming'.

Ceibal Foundation is organizing the 3rd edition of the EdTech Winter School in partnership with ANII (Agencia Nacional de Investigación e Innovación) and with the support of the International Development Research Centre -IDRC-. The EdTech Winter School is a multi-stakeholder initiative organized within the framework of the Education Sector Fund "Digital Inclusion: Education with New Horizons" created with ANII and ADELA (Alliance for the Digitalization of Education in Latin America) supported by the International Development Research Centre (IDRC). In this context and for the past three years, the Winter School focused in creating a stimulating learning environment to present and discuss key challenges, research trends and opportunities; to foresee new horizons in education, learning and teaching practices enhanced by digital technologies. This year's edition "Human Centered Technologies for Education" aims to assess, analyze and explore the changes, opportunities and challenges that technology-driven transformations are creating for education worldwide. Advances in areas as automation, artificial intelligence, robotics, Big Data, among others, are shaping society in ways that could not be foreseen a few years ago.

Artificial Intelligence and Machine Learning (AI and ML) technologies have come a long way since its first inception. Who would have thought that we would have a working model of actual computer-based assistants that can do things like manage our schedules? Who would have thought that we could even use these assistants to manage our homes? These things can even be used to diagnose cancer patients, something impossible without doctors even five years ago. AWS is at the forefront of AI and ML technology. As one of the world's largest technology innovators, they would naturally be at an advantage to feed enough data to the technology and accelerate their development.