The U.S. military wants to expand its use of artificial intelligence in warfare, but says it will take care to deploy the technology in accordance with the nation's values. The Pentagon outlined its first AI strategy in a report released Tuesday. The plan calls for accelerating the use of AI systems throughout the military, from intelligence-gathering operations to predicting maintenance problems in planes or ships. It urges the U.S. to advance such technology swiftly before other countries chip away at its technological advantage. "Other nations, particularly China and Russia, are making significant investments in AI for military purposes, including in applications that raise questions regarding international norms and human rights," the report says.

Robotic arms prepare dishes at a hot pot restaurant in Beijing, capital of China, Dec. 5, 2018. Six of the 11 artificial intelligence (AI) startups that are considered to be unicorns – which means to have a value of one billion U.S. dollars or above – come from China, according to CB Insights, a research firm that tracks venture capital and startups. SenseTime took the top spot with a valuation of 4.5 billion U.S. dollars, followed by Yitu Technology at 2.3 billion U.S. dollars and smaller unicorns 4Paradigm, Horizon Robotics and Momenta. The annual report published by CB Insights compiles a list of 100 of the most promising private companies. The selection is based on several factors, including patent activity, investor profile and market potential.

Learning image representations to capture fine-grained semantics has been a challenging and important task enabling many applications such as image search and clustering. In this paper, we present Graph-Regularized Image Semantic Embedding (Graph-RISE), a large-scale neural graph learning framework that allows us to train embeddings to discriminate an unprecedented O(40M) ultra-fine-grained semantic labels. Graph-RISE outperforms state-of-the-art image embedding algorithms on several evaluation tasks, including image classification and triplet ranking. We provide case studies to demonstrate that, qualitatively, image retrieval based on Graph-RISE effectively captures semantics and, compared to the state-of-the-art, differentiates nuances at levels that are closer to human-perception.

Sequential data modeling and analysis have become indispensable tools for analyzing sequential data, such as time-series data, because larger amounts of sensed event data have become available. These methods capture the sequential structure of data of interest, such as input-output relations and correlation among datasets. However, because most studies in this area are specialized or limited to their respective applications, rigorous requirement analysis of such models has not been undertaken from a general perspective. Therefore, we particularly examine the structure of sequential data, and extract the necessity of `state duration' and `state interval' of events for efficient and rich representation of sequential data. Specifically addressing the hidden semi-Markov model (HSMM) that represents such state duration inside a model, we attempt to add representational capability of a state interval of events onto HSMM. To this end, we propose two extended models: an interval state hidden semi-Markov model (IS-HSMM) to express the length of a state interval with a special state node designated as "interval state node"; and an interval length probability hidden semi-Markov model (ILP-HSMM) which represents the length of the state interval with a new probabilistic parameter "interval length probability." Exhaustive simulations have revealed superior performance of the proposed models in comparison with HSMM. These proposed models are the first reported extensions of HMM to support state interval representation as well as state duration representation.

We study the trade-offs between convergence rate and robustness to gradient errors in designing a first-order algorithm. We focus on gradient descent (GD) and accelerated gradient (AG) methods for minimizing strongly convex functions when the gradient has random errors in the form of additive white noise. With gradient errors, the function values of the iterates need not converge to the optimal value; hence, we define the robustness of an algorithm to noise as the asymptotic expected suboptimality of the iterate sequence to input noise power. For this robustness measure, we provide exact expressions for the quadratic case using tools from robust control theory and tight upper bounds for the smooth strongly convex case using Lyapunov functions certified through matrix inequalities. We use these characterizations within an optimization problem which selects parameters of each algorithm to achieve a particular trade-off between rate and robustness. Our results show that AG can achieve acceleration while being more robust to random gradient errors. This behavior is quite different than previously reported in the deterministic gradient noise setting. We also establish some connections between the robustness of an algorithm and how quickly it can converge back to the optimal solution if it is perturbed from the optimal point with deterministic noise. Our framework also leads to practical algorithms that can perform better than other state-of-the-art methods in the presence of random gradient noise.

We propose a new policy iteration theory as an important extension of soft policy iteration and Soft Actor-Critic (SAC), one of the most efficient model free algorithms for deep reinforcement learning. Supported by the new theory, arbitrary entropy measures that generalize Shannon entropy, such as Tsallis entropy and Renyi entropy, can be utilized to properly randomize action selection while fulfilling the goal of maximizing expected long-term rewards. Our theory gives birth to two new algorithms, i.e., Tsallis entropy Actor-Critic (TAC) and Renyi entropy Actor-Critic (RAC). Theoretical analysis shows that these algorithms can be more effective than SAC. Moreover, they pave the way for us to develop a new Ensemble Actor-Critic (EAC) algorithm in this paper that features the use of a bootstrap mechanism for deep environment exploration as well as a new value-function based mechanism for high-level action selection. Empirically we show that TAC, RAC and EAC can achieve state-of-the-art performance on a range of benchmark control tasks, outperforming SAC and several cutting-edge learning algorithms in terms of both sample efficiency and effectiveness.

The concordance correlation coefficient (CCC) is one of the most widely used reproducibility indices, introduced by Lin in 1989. In addition to its extensive use in assay validation, CCC serves various different purposes in other multivariate population-related tasks. For example, it is often used as a metric to quantify an inter-rater agreement. It is also often used as a performance metric for prediction problems. In terms of the cost function, however, there has been hardly any attempt to design one to train the predictive deep learning models. In this paper, we present a family of lightweight cost functions that aim to also maximise CCC, when minimising the prediction errors. To this end, we first reformulate CCC in terms of the errors in the prediction; and then as a logical next step, in terms of the sequence of the fixed set of errors. To elucidate our motivation and the results we obtain through these error rearrangements, the data we use is the set of gold standard annotations from a well-known database called `Automatic Sentiment Analysis in the Wild' (SEWA), popular thanks to its use in the latest Audio/Visual Emotion Challenges (\textsc{AVEC'17} and \textsc{AVEC'18}). We also present some new and interesting mathematical paradoxes we have discovered through this CCC reformulation endeavour.

Bayesian optimization (BO) is a powerful paradigm for derivative-free global optimization of a black-box objective function (BOF) that is expensive to evaluate. However, the overhead of BO can still be prohibitive if the maximum number of allowed function evaluations is less than required. In this paper, we investigate how to reduce the required number of function evaluations for BO without compromise in solution quality. We explore the idea of posterior regularization for harnessing low fidelity (LF) data within the Gaussian process upper confidence bound (GP-UCB) framework. The LF data are assumed to arise from previous evaluations of an LF approximation of the BOF. An extra GP expert called LF-GP is trained to fit the LF data. We develop a dynamic weighted product of experts (DW-POE) fusion operator. The regularization is induced from this operator on the posterior of the BOF. The impact of the LF-GP expert on the resulting regularized posterior is adaptively adjusted via Bayesian formalism. Extensive experimental results on benchmark BOF optimization tasks demonstrate the superior performance of the proposed algorithm over state-of-the-art.

Cooperation information sharing is important to theories of human learning and has potential implications for machine learning. Prior work derived conditions for achieving optimal Cooperative Inference given strong, relatively restrictive assumptions. We relax these assumptions by demonstrating convergence for any discrete joint distribution, robustness through equivalence classes and stability under perturbation, and effectiveness by deriving bounds from structural properties of the original joint distribution. We provide geometric interpretations, connections to and implications for optimal transport, and connections to importance sampling, and conclude by outlining open questions and challenges to realizing the promise of Cooperative Inference.

We consider the question of what functions can be captured by ReLU networks with an unbounded number of units (infinite width), but where the overall network Euclidean norm (sum of squares of all weights in the system, except for an unregularized bias term for each unit) is bounded; or equivalently what is the minimal norm required to approximate a given function. For functions $f : \mathbb R \rightarrow \mathbb R$ and a single hidden layer, we show that the minimal network norm for representing $f$ is $\max(\int |f''(x)| dx, |f'(-\infty) + f'(+\infty)|)$, and hence the minimal norm fit for a sample is given by a linear spline interpolation.