Playing human games such as chess and Go has long been considered to be a major benchmark of human capabilities. Computer programs have become robust chess players and, since the late 1990s, have been able to beat even the best human chess champions; though, for a long time, computers were unable to beat expert Go players -- the game of Go has proven to be especially difficult for computers. However, in 2016, a new program called AlphaGo finally won a victory over a human Go champion, only to be beaten by its subsequent versions (AlphaGo Zero and AlphaZero). AlphaZero proceeded to beat the best computers and humans in chess, shogi and Go, including all its predecessors from the Alpha family [1]. Core to AlphaZero's success is its use of a deep neural network, trained through reinforcement learning, as a powerful heuristic to guide a tree search algorithm (specifically Monte Carlo Tree Search). The recent successes of machine learning are good reason to consider the limitations of learning algorithms and, in a broader sense, the limitations of AI. In the context of a particular competition (or'game'), a natural question to ask is whether an absolute winner AI might exist -- one that, given sufficient resources, will always achieve the best possible outcome.

Recently, the reciprocal recommendation, especially for online dating applications, has attracted more and more research attention. Different from conventional recommendation problems, the reciprocal recommendation aims to simultaneously best match users' mutual preferences. Intuitively, the mutual preferences might be affected by a few key attributes that users like or dislike. Meanwhile, the interactions between users' attributes and their key attributes are also important for key attributes selection. Motivated by these observations, in this paper we propose a novel reinforced random convolutional network (RRCN) approach for the reciprocal recommendation task. In particular, we technically propose a novel random CNN component that can randomly convolute non-adjacent features to capture their interaction information and learn feature embeddings of key attributes to make the final recommendation. Moreover, we design a reinforcement learning based strategy to integrate with the random CNN component to select salient attributes to form the candidate set of key attributes. We evaluate the proposed RRCN against a number of both baselines and the state-of-the-art approaches on two real-world datasets, and the promising results have demonstrated the superiority of RRCN against the compared approaches in terms of a number of evaluation criteria.

Mitigating the new user cold-start problem has been critical in the recommendation system for online service providers to influence user experience in decision making which can ultimately affect the intention of users to use a particular service. Previous studies leveraged various side information from users and items; however, it may be impractical due to privacy concerns. In this paper, we present ColdGAN, an end-to-end GAN based model with no use of side information to resolve this problem. The main idea of the proposed model is to train a network that learns the rating distributions of experienced users given their cold-start distributions. We further design a time-based function to restore the preferences of users to cold-start states. With extensive experiments on two real-world datasets, the results show that our proposed method achieves significantly improved performance compared with the state-of-the-art recommenders.

Signed link prediction in social networks aims to reveal the underlying relationships (i.e. links) among users (i.e. nodes) given their existing positive and negative interactions observed. Most of the prior efforts are devoted to learning node embeddings with graph neural networks (GNNs), which preserve the signed network topology by message-passing along edges to facilitate the downstream link prediction task. Nevertheless, the existing graph-based approaches could hardly provide human-intelligible explanations for the following three questions: (1) which neighbors to aggregate, (2) which path to propagate along, and (3) which social theory to follow in the learning process. To answer the aforementioned questions, in this paper, we investigate how to reconcile the \textit{balance} and \textit{status} social rules with information theory and develop a unified framework, termed as Signed Infomax Hyperbolic Graph (\textbf{SIHG}). By maximizing the mutual information between edge polarities and node embeddings, one can identify the most representative neighboring nodes that support the inference of edge sign. Different from existing GNNs that could only group features of friends in the subspace, the proposed SIHG incorporates the signed attention module, which is also capable of pushing hostile users far away from each other to preserve the geometry of antagonism. The polarity of the learned edge attention maps, in turn, provide interpretations of the social theories used in each aggregation. In order to model high-order user relations and complex hierarchies, the node embeddings are projected and measured in a hyperbolic space with a lower distortion. Extensive experiments on four signed network benchmarks demonstrate that the proposed SIHG framework significantly outperforms the state-of-the-arts in signed link prediction.

Bayesian optimization is a popular method for solving the problem of global optimization of an expensive-to-evaluate black-box function. It relies on a probabilistic surrogate model of the objective function, upon which an acquisition function is built to determine where next to evaluate the objective function. In general, Bayesian optimization with Gaussian process regression operates on a continuous space. When input variables are categorical or discrete, an extra care is needed. A common approach is to use one-hot encoded or Boolean representation for categorical variables which might yield a {\em combinatorial explosion} problem. In this paper we present a method for Bayesian optimization in a combinatorial space, which can operate well in a large combinatorial space. The main idea is to use a random mapping which embeds the combinatorial space into a convex polytope in a continuous space, on which all essential process is performed to determine a solution to the black-box optimization in the combinatorial space. We describe our {\em combinatorial Bayesian optimization} algorithm and present its regret analysis. Numerical experiments demonstrate that our method outperforms existing methods.

The Bayesian treatment of neural networks dictates that a prior distribution is specified over their weight and bias parameters. This poses a challenge because modern neural networks are characterized by a large number of parameters, and the choice of these priors has an uncontrolled effect on the induced functional prior, which is the distribution of the functions obtained by sampling the parameters from their prior distribution. We argue that this is a hugely limiting aspect of Bayesian deep learning, and this work tackles this limitation in a practical and effective way. Our proposal is to reason in terms of functional priors, which are easier to elicit, and to "tune" the priors of neural network parameters in a way that they reflect such functional priors. Gaussian processes offer a rigorous framework to define prior distributions over functions, and we propose a novel and robust framework to match their prior with the functional prior of neural networks based on the minimization of their Wasserstein distance. We provide vast experimental evidence that coupling these priors with scalable Markov chain Monte Carlo sampling offers systematically large performance improvements over alternative choices of priors and state-of-the-art approximate Bayesian deep learning approaches. We consider this work a considerable step in the direction of making the long-standing challenge of carrying out a fully Bayesian treatment of neural networks, including convolutional neural networks, a concrete possibility.

Multi-scene reinforcement learning involves training the RL agent across multiple scenes / levels from the same task, and has become essential for many generalization applications. However, the inclusion of multiple scenes leads to an increase in sample variance for policy gradient computations, often resulting in suboptimal performance with the direct application of traditional methods (e.g. PPO, A3C). One strategy for variance reduction is to consider each scene as a distinct Markov decision process (MDP) and learn a joint value function dependent on both state (s) and MDP (M). However, this is non-trivial as the agent is usually unaware of the underlying level at train / test times in multi-scene RL. Recently, Singh et al. [1] tried to address this by proposing a dynamic value estimation approach that models the true joint value function distribution as a Gaussian mixture model (GMM). In this paper, we argue that the error between the true scene-specific value function and the predicted dynamic estimate can be further reduced by progressively enforcing sparse cluster assignments once the agent has explored most of the state space. The resulting agents not only show significant improvements in the final reward score across a range of OpenAI ProcGen environments, but also exhibit increased navigation efficiency while completing a game level.

Salesforce is backing an AI project called SharkEye which aims to save the lives of beachgoers from one of the sea's deadliest predators. Shark attacks are, fortunately, quite rare. However, they do happen and most cases are either fatal or cause life-changing injuries. Just last week, a fatal shark attack in Australia marked the eighth of the year--an almost 100-year record for the highest annual death toll. Once rare sightings in Southern California beaches are now becoming increasingly common as sharks are preferring the warmer waters close to shore.

Corporate contact centers are embracing big data to offer an improved and more customized customer experience. Also, presently, contact centers are digitizing and collecting each customer interaction that happens by means of telephone, social media, email, text or even face to face. Following this move into big data, contact centers are utilizing speech analytics to take their products, processes and customer service efforts step ahead. Voice analytics is the process of digitally analyzing interactions between customers and agents. What's more, despite the fact that the innovation has been around for over 10 years, late headways in digitalization, machine learning and artificial intelligence (AI) have made it all the more remarkable and have empowered contact centers to change piles of data into real-time insights.

One strand of research has used probabilistic regularization for continual learning, with two of the main approaches in this vein being Online Elastic Weight Consolidation (Online EWC) and Variational Continual Learning (VCL). VCL employs variational inference, which in other settings has been improved empirically by applying likelihood-tempering. We show that applying this modification to VCL recovers Online EWC as a limiting case, allowing for interpolation between the two approaches. In order to mitigate the observed overpruning effect of VI, we take inspiration from a common multi-task architecture, neural networks with task-specific FiLM layers, and find that this addition leads to significant performance gains, specifically for variational methods. In the small-data regime, GVCL strongly outperforms existing baselines. In larger datasets, GVCL with FiLM layers outperforms or is competitive with existing baselines in terms of accuracy, whilst also providing significantly better calibration. Continual learning methods enable learning when a set of tasks changes over time. This topic is of practical interest as many real-world applications require models to be regularly updated as new data is collected or new tasks arise. Standard machine learning models and training procedures fail in these settings (French, 1999), so bespoke architectures and fitting procedures are required. This paper makes two main contributions to continual learning for neural networks. First, we develop a new regularization-based approach to continual learning. Regularization approaches adapt parameters to new tasks while keeping them close to settings that are appropriate for old tasks. Two popular approaches of this type are Variational Continual Learning (VCL) (Nguyen et al., 2018) and Online Elastic Weight Consolidation (Online EWC) (Kirkpatrick et al., 2017; Schwarz et al., 2018). The former is based on a variational approximation of a neural network's posterior distribution over weights, while the latter uses Laplace's approximation. In this paper, we propose Generalized Variational Continual Learning (GVCL) of which VCL and Online EWC are two special cases. Under this unified framework, we are able to combine the strengths of both approaches. GVCL is closely related to likelihood-tempered Variational Inference (VI), which has been found to improve performance in standard learning settings (Zhang et al., 2018; Osawa et al., 2019).