Over the last fifty years, the world has seen an impressive growth in science - in terms of publications, number of highly-trained scientists and impact of results. This is not only good for science, but also for everyone who benefits from the outcomes and applications of scientific research. An area that has seen much growth lately is artificial intelligence (AI). A full four thousand scientific papers were submitted to the International Joint Conference on Artificial Intelligence (IJCAI), the top international conference in AI, which was held in Stockholm earlier this summer. While growth in science and technology has happened in most countries and regions in the world, recent developments in Asia are particularly impressive.

But more and more innovation is now coming from other regions, and there are reasons to believe the AI revolution might first take off elsewhere. Due to its size, capital, and cultural landscape, China, in particular, seems poised to lead AI evolution. China's economic growth since the 1980s has been staggering, and government investment has been key toward creating new technologies in the country. In July of 2017, the Chinese government resolved to make the nation the world's leader in AI by 2030. Although any government can make similarly bold claims, China's track record in recent years has been remarkable, with the nation now home to some of the world's most extensive and modern infrastructure.

AI is currently making its way into society in various ways, such as seen in the defeat of a top-class professional go player against AI in a game of go, and the start of initiatives in autonomous cars using AI. Of course, the same can be said of the world of business, and one of the things that is attracting much attention in the use of AI is IBM Watson ("Watson") developed by IBM Corporation. Watson is provided over the cloud, and it presents answers grounded in reason based on an interpretation of complex questions, conversations, etc. posed to it in natural language by analyzing an extremely vast amount of data. In addition, by gaining new experience, it begins to rapidly improve the accuracy of the answers. Currently, based on the Japanese version of Watson, Softbank has started offering various services. The ease of implementing these services and their affordability may become a powerful ally to small and medium-sized companies in the business arena.

SK Telecom is definitely a forward looking companies and will be one of the first companies to trial and start 5G. It is not wasting time when it comes to speech recognition and AI as SK Telecom has announced the first commercial adoption of FPGA accelerators from Xilinx in the AI domain for large scale data centers in South Korea. These two companies made some big claims over the efficiency of the FPGA based solution. SK Telecom has chosen the Xilinx Kintex UltraScale FPGAs as its artificial intelligence (AI) accelerators in its data center. Xilinx FPGA will take on the responsible task of running SKT's automatic speech-recognition (ASR) application to accelerate Nugu, SKT's voice-activated assistant.

U.S. technology companies are converging on Taiwan to build regional research and development centers, drawn by the island's relatively low wages and the government's strategy of forging closer ties with Washington. IBM and Oath, the parent company of Yahoo, have all announced plans this year to build their research and development hubs on the island and to initiate large recruitment projects. "The approach of President Tsai Ing-wen's government to shift away from China and forge closer ties with the U.S. is a strong push behind many U.S. companies' investments over the past one year," said Gordon Sun, director of the Economic Forecasting Center at the Taiwan Institute of Economic Research. Attracting investments from U.S. tech companies is part of the government's plans to build an artificial intelligence industry. Taiwan has pledged to pour 10 billion New Taiwan dollars ($326 million) each year into AI-related investments over the next three years.

The Jacobian-based Saliency Map Attack is a family of adversarial attack methods for fooling classification models, such as deep neural networks for image classification tasks. By saturating a few pixels in a given image to their maximum or minimum values, JSMA can cause the model to misclassify the resulting adversarial image as a specified erroneous target class. We propose two variants of JSMA, one which removes the requirement to specify a target class, and another that additionally does not need to specify whether to only increase or decrease pixel intensities. Our experiments highlight the competitive speeds and qualities of these variants when applied to datasets of hand-written digits and natural scenes.

Abstract-- Extremely preterm infants commonly require intubation and invasive mechanical ventilation after birth. While the duration of mechanical ventilation should be minimized in order to avoid complications, extubation failure is associated with increases in morbidities and mortality. As part of a prospective observational study aimed at developing an accurate predictor of extubation readiness, Markov and semi-Markov chain models were applied to gain insight into the respiratory patterns of these infants, with more robust timeseries modeling using semi-Markov models. This model revealed interesting similarities and differences between newborns who succeeded extubation and those who failed. The parameters of the model were further applied to predict extubation readiness via generative (joint likelihood) and discriminative (support vector machine) approaches. Results showed that up to 84% of infants who failed extubation could have been accurately identified prior to extubation.

We focus on solving the clustered lasso problem, which is a least squares problem with the $\ell_1$-type penalties imposed on both the coefficients and their pairwise differences to learn the group structure of the regression parameters. Here we first reformulate the clustered lasso regularizer as a weighted ordered-lasso regularizer, which is essential in reducing the computational cost from $O(n^2)$ to $O(n\log (n))$. We then propose an inexact semismooth Newton augmented Lagrangian (SSNAL) algorithm to solve the clustered lasso problem or its dual via this equivalent formulation, depending on whether the sample size is larger than the dimension of the features. An essential component of the SSNAL algorithm is the computation of the generalized Jacobian of the proximal mapping of the clustered lasso regularizer. Based on the new formulation, we derive an efficient procedure for its computation. Comprehensive results on the global convergence and local linear convergence of the SSNAL algorithm are established. For the purpose of exposition and comparison, we also summarize/design several first-order methods that can be used to solve the problem under consideration, but with the key improvement from the new formulation of the clustered lasso regularizer. As a demonstration of the applicability of our algorithms, numerical experiments on the clustered lasso problem are performed. The experiments show that the SSNAL algorithm substantially outperforms the best alternative algorithm for the clustered lasso problem.

In sequence generation task, many works use policy gradient for model optimization to tackle the intractable backpropagation issue when maximizing the non-differentiable evaluation metrics or fooling the discriminator in adversarial learning. In this paper, we replace policy gradient with proximal policy optimization (PPO), which is a proved more efficient reinforcement learning algorithm, and propose a dynamic approach for PPO (PPO-dynamic). We demonstrate the efficacy of PPO and PPO-dynamic on conditional sequence generation tasks including synthetic experiment and chit-chat chatbot. The results show that PPO and PPO-dynamic can beat policy gradient by stability and performance.

There are many models, often called unnormalized models, whose normalizing constants are not calculated in closed form. Maximum likelihood estimation is not directly applicable to unnormalized models. Score matching, contrastive divergence method, pseudo-likelihood, Monte Carlo maximum likelihood, and noise contrastive estimation (NCE) are popular methods for estimating parameters of such models. In this paper, we focus on NCE. The estimator derived from NCE is consistent and asymptotically normal because it is an M-estimator. NCE characteristically uses an auxiliary distribution to calculate the normalizing constant in the same spirit of the importance sampling. In addition, there are several candidates as objective functions of NCE. We focus on how to reduce asymptotic variance. First, we propose a method for reducing asymptotic variance by estimating the parameters of the auxiliary distribution. Then, we determine the form of the objective functions, where the asymptotic variance takes the smallest values in the original estimator class and the proposed estimator classes. We further analyze the robustness of the estimator.