Many objects in the real world are difficult to describe by a single numerical vector of a fixed length, whereas describing them by a set of vectors is more natural. Therefore, Multiple instance learning (MIL) techniques have been constantly gaining on importance throughout last years. MIL formalism represents each object (sample) by a set (bag) of feature vectors (instances) of fixed length where knowledge about objects (e.g., class label) is available on bag level but not necessarily on instance level. Many standard tools including supervised classifiers have been already adapted to MIL setting since the problem got formalized in late nineties. In this work we propose a neural network (NN) based formalism that intuitively bridges the gap between MIL problem definition and the vast existing knowledge-base of standard models and classifiers. We show that the proposed NN formalism is effectively optimizable by a modified back-propagation algorithm and can reveal unknown patterns inside bags. Comparison to eight types of classifiers from the prior art on a set of 14 publicly available benchmark datasets confirms the advantages and accuracy of the proposed solution.

We consider the problem of maximizing an unknown function over a compact and convex set using as few observations as possible. We observe that the optimization of the function essentially relies on learning the induced bipartite ranking rule of f. Based on this idea, we relate global optimization to bipartite ranking which allows to address problems with high dimensional input space, as well as cases of functions with weak regularity properties. The paper introduces novel meta-algorithms for global optimization which rely on the choice of any bipartite ranking method. Theoretical properties are provided as well as convergence guarantees and equivalences between various optimization methods are obtained as a by-product. Eventually, numerical evidence is given to show that the main algorithm of the paper which adapts empirically to the underlying ranking structure essentially outperforms existing state-of-the-art global optimization algorithms in typical benchmarks.

Bayesian optimization has become a successful tool for hyperparameter optimization of machine learning algorithms, such as support vector machines or deep neural networks. Despite its success, for large datasets, training and validating a single configuration often takes hours, days, or even weeks, which limits the achievable performance. To accelerate hyperparameter optimization, we propose a generative model for the validation error as a function of training set size, which is learned during the optimization process and allows exploration of preliminary configurations on small subsets, by extrapolating to the full dataset. We construct a Bayesian optimization procedure, dubbed Fabolas, which models loss and training time as a function of dataset size and automatically trades off high information gain about the global optimum against computational cost. Experiments optimizing support vector machines and deep neural networks show that Fabolas often finds high-quality solutions 10 to 100 times faster than other state-of-the-art Bayesian optimization methods or the recently proposed bandit strategy Hyperband.

Clustering is an underspecified task: there are no universal criteria for what makes a good clustering. This is especially true for relational data, where similarity can be based on the features of individuals, the relationships between them, or a mix of both. Existing methods for relational clustering have strong and often implicit biases in this respect. In this paper, we introduce a novel similarity measure for relational data. It is the first measure to incorporate a wide variety of types of similarity, including similarity of attributes, similarity of relational context, and proximity in a hypergraph. We experimentally evaluate how using this similarity affects the quality of clustering on very different types of datasets. The experiments demonstrate that (a) using this similarity in standard clustering methods consistently gives good results, whereas other measures work well only on datasets that match their bias; and (b) on most datasets, the novel similarity outperforms even the best among the existing ones.

Juergen Schmidhuber taught a computer to park a car. He's also showing that same machine how to trade stocks and detect flaws in steel production. Unrelated as these tasks may appear, Schmidhuber thinks a seemingly random training regimen is key to creating artificial intelligence that can solve any problem. Schmidhuber's AI theories tend to carry weight. In 1997, he co-authored a seminal paper that laid the groundwork for modern AI systems.

Governments and politicians have failed to respond to the current refugee crisis, and that's being made painfully prominent by the fact that 65 million people had been forced from their homes last year. However, where governments fail, private citizens and ingenuity can help. Joshua Browder's DoNotPay has proved this with a robot lawyer who gives refugees free legal advice. According to the Guardian, the chatbot aids refugees through Facebook Messenger by asking them a series of question to determine which forms they have to hand in and whether they are eligible for asylum protection. The robot lawyer also uses the information to automatically fill out forms and send them in on behalf of his'clients'.

The Volkwagen Group is driving forward amid a lawsuit following the scandal involving cheated emissions tests, planning to unveil new technologically advanced vehicles at the Geneva Motor Show, including a completely self-driving car. The show begins Thursday, March 9, and will last 10 days in Geneva, Switzerland, but the Volkswagen Group unveiled their newest concept a bit early. As a step in the direction of innovation -- and a fresh image -- Volkswagen Group CEO Matthias Müller announced that the company is aiming to have more than 30 solely battery powered vehicles offered by 2025. The company also announced the work they've done on their first self-driving concept vehicle called "Sedric," a project the company plans to continue investing in. Sedric will be the model for a Level 5 automated vehicle, meaning it will be completely automated and fully electric. It has no cockpit, no pedals and no steering wheel, according to a press release.

You may have heard of artificial intelligence (AI), which is usually defined as the science of making computers do things that would require intelligence if completed by humans. However, for many this seems like a visionary technology, not ready for day-to-day use within your business. So you may be surprised to learn that AI could already be benefitting your business. "Many service providers to small businesses are already leveraging AI's capabilities," reveals Dr. Andy Pardoe, founder of Informed.AI and homeAI.info. "AI has never been more widely used, with many of the largest technology companies providing integrated AI platforms. This democratisation of AI is allowing small businesses to more easily add advanced data analytics and machine learning to their processes."

Roboticists at Disney Research are investigating how to improve the quality of human-robot interactions by studying how speech patterns affect engagement with a creepy anthropomorphic bot that imitates its playmates' speech. It's a natural enough line of research: computer-generated speech and interaction patterns are so wooden across the board that little is expected of them. But pairs of people -- for example, kids playing with each other -- who are in tune verbally (in "prosodic synchrony") tend to be more engaged and successful in what they're doing. For the study, the team paired kids with a robot and had them play a little platforming game where one player tells the character to go and the other tells it to jump. The system they created listened to the child's voice and extracted some basic properties: loudness, word length, and frequency (i.e.

MIT researchers have developed a way for humans to control robots with their mind. The new system can detect when a person notices a robot making a mistake, and classifies these brain waves almost instantly to provide feedback. In tests with a robot named Baxter, the researchers found that the new technique helped the humanoid to make correct choices during an object-sorting task – and it could one day allow humans to wordlessly communicate with robots. MIT researchers have developed a way for humans to control robots with their mind. As the robot, Baxter, attempts to sort objects between'Paint' or'Wire' bins, the system looks for brain signals from the human observer known as'error-related potentials.' Once an error has been detected, machine-learning algorithms can sort these brain waves in just 10 to 30 milliseconds to provide feedback.