The-state-of-the-art neural quality estimation (QE) of machine translation model consists of two sub-networks that are tuned separately, a bidirectional recurrent neural network (RNN) encoder-decoder trained for neural machine translation, called the predictor, and an RNN trained for sentence-level QE tasks, called the estimator. We propose to combine the two sub-networks into a whole neural network, called the unified neural network. When training, the bidirectional RNN encoder-decoder are initialized and pre-trained with the bilingual parallel corpus, and then, the networks are trained jointly to minimize the mean absolute error over the QE training samples. Compared with the predictor and estimator approach, the use of a unified neural network helps to train the parameters of the neural networks that are more suitable for the QE task. Experimental results on the benchmark data set of the WMT17 sentence-level QE shared task show that the proposed unified neural network approach consistently outperforms the predictor and estimator approach and significantly outperforms the other baseline QE approaches.

It is well known that both shape and motion can be factorized directly from the measurement matrix constructed from feature points trajectories under orthographic camera model. In practical applications, the measurement matrix might be contaminated by noises and contains outliers. A direct SVD (Singular Value Decomposition) to the measurement matrix with outliers would yield erroneous result. This paper presents a novel algorithm for computing SVD with outliers. We decompose the SVD computation as a set of alternate linear regression subproblems. The linear regression subproblems are solved robustly by applying the RANSAC strategy. The proposed robust factorization method with outliers can improve the reconstruction result remarkably. Quantitative and qualitative experiments illustrate the good performance of the proposed method.

The problem of multi-body motion segmentation is important in many computer vision applications. In this paper, we propose a novel algorithm called fuzzy k-subspace clustering for robust segmentation. The proposed method exploits the property that under orthographic camera model the tracked feature points of moving objects reside in multiple subspaces. We compute a partition of feature points into corresponding subspace clusters. First, we find a "soft partition" of feature points based on fuzzy k-subspace algorithm. The proposed fuzzy k-subspace algorithm iteratively minimizes the objective function using Weighted Singular Value Decomposition. Then the points with high partition confidence are gathered to form the subspace bases and the remaining points are classified using their distance to the bases. The proposed method can handle the case of missing data naturally, meaning that the feature points do not have to be visible throughout the sequence. The method is robust to noise and insensitive to initialization. Extensive experiments on synthetic and real data show the effectiveness of the proposed fuzzy k-subspace clustering algorithm.

In this paper, an adaptive expansion strategy (AES) is proposed for multiple-input/multiple-output (MIMO) detection in the presence of circular signals. By exploiting channel properties, the AES classifies MIMO channels into three types: excellent, average and deep fading. To avoid unnecessary branch-searching, the AES adopts single expansion (SE), partial expansion (PE) and full expansion (FE) for excellent channels, average channels and deep fading channels, respectively. In the PE, the non-circularity of signal is exploited, and the widely linear processing is extended from non-circular signals to circular signals by I (or Q) component cancellation. An analytical performance analysis is given to quantify the performance improvement. Simulation results show that the proposed algorithm can achieve quasi-optimal performance with much less complexity (hundreds of flops/symbol are saved) compared with the fixed-complexity sphere decoder (FSD) and the sphere decoder (SD).

This paper addresses the problem of view invariant action recognition using 2D trajectories of landmark points on human body. It is a challenging task since for a specific action category, the 2D observations of different instances might be extremely different due to varying viewpoint and changes in speed. By assuming that the execution of an action can be approximated by dynamic linear combination of a set of basis shapes, a novel view invariant human action recognition method is proposed based on non-rigid matrix factorization and Hidden Markov Models (HMMs). We show that the low dimensional weight coefficients of basis shapes by measurement matrix non-rigid factorization contain the key information for action recognition regardless of the viewpoint changing. Based on the extracted discriminative features, the HMMs is used for temporal dynamic modeling and robust action classification. The proposed method is tested using real life sequences and promising performance is achieved.

This paper presents an approach for cross-pose face recognition by virtual view generation using an appearance clustering based local view transition model. Previously, the traditional global pattern based view transition model (VTM) method was extended to its local version called LVTM, which learns the linear transformation of pixel values between frontal and non-frontal image pairs from training images using partial image in a small region for each location, instead of transforming the entire image pattern. In this paper, we show that the accuracy of the appearance transition model and the recognition rate can be further improved by better exploiting the inherent linear relationship between frontal-nonfrontal face image patch pairs. This is achieved based on the observation that variations in appearance caused by pose are closely related to the corresponding 3D structure and intuitively frontal-nonfrontal patch pairs from more similar local 3D face structures should have a stronger linear relationship. Thus for each specific location, instead of learning a common transformation as in the LVTM, the corresponding local patches are first clustered based on an appearance similarity distance metric and then the transition models are learned separately for each cluster. In the testing stage, each local patch for the input non-frontal probe image is transformed using the learned local view transition model corresponding to the most visually similar cluster. The experimental results on a real-world face dataset demonstrated the superiority of the proposed method in terms of recognition rate.

Considering that different people are different in their linguistic preference and in order to determine the consensus state when using Computing with Words (CWW) for supporting consensus decision making, this paper first proposes an interval composite scale based 2-tuple linguistic model, which realizes the process of translation from word to interval numerical and the process of retranslation from interval numerical to word. Second, this paper proposes an interval composite scale based personalized individual semantics model (ICS-PISM), which can provide different linguistic representation models for different decision-makers. Finally, this paper proposes a consensus decision making model with ICS-PISM, which includes a semantic translation and retranslation phase during decision process and determines the consensus state of the whole decision process. These models proposed take into full consideration that human language contains vague expressions and usually real-world preferences are uncertain, and provide efficient computation models to support consensus decision making.

Aiming at the contradiction between exploration and exploitation in deep reinforcement learning, this paper proposes “reward-based exploration strategy combined with Softmax action selection” (RBE-Softmax) as a dynamic exploration strategy to guide the agent to learn. The superiority of the proposed method is that the characteristic of agent's learning process is utilized to adapt exploration parameters online, and the agent is able to select potential optimal action more effectively. The proposed method is evaluated in discrete and continuous control tasks on OpenAI Gym, and the empirical evaluation results show that RBE-Softmax method leads to statistically-significant improvement in the performance of deep reinforcement learning algorithms.

As an important method to solve sequential decision-making problems, reinforcement learning learns the policy of tasks through the interaction with environment. But it has difficulties scaling to large-scale problems. One of the reasons is the exploration and exploitation dilemma which may lead to inefficient learning. We present an approach that addresses this shortcoming by introducing qualitative knowledge into reinforcement learning using cloud control systems to represent ‘if-then’ rules. We use it as the heuristics exploration strategy to guide the action selection in deep reinforcement learning. Empirical evaluation results show that our approach can make significant improvement in the learning process.

The current approaches to semantic role classification usually first define a representation vector for a candidate role and feed the vector into a deep neural network to perform classification. The representation vector contains some lexicalization features like word embeddings, lemmar embeddings. From linguistics, the semantic role frame of a sentence is a joint structure with strong dependencies between arguments which is not considered in current deep SRL systems. Therefore, this paper proposes a global deep reranking model to exploit these strong dependencies. The evaluation experiments on the CoNLL 2009 shared tasks show that our system can outperforms a strong local system significantly that does not consider role dependency relations.

In this paper, we study domain adaptation of semantic role classification. Most systems utilize the supervised method for semantic role classification. But, these methods often suffer severe performance drops on out-of-domain test data. The reason for the performance drops is that there are giant feature differences between source and target domain. This paper proposes a framework called Adversarial Domain Adaption Network (ADAN) to relieve domain adaption of semantic role classification. The idea behind our method is that the proposed framework can derive domain-invariant features via adversarial learning and narrow down the gap between source and target feature space. To evaluate our method, we conduct experiments on English portion in the CoNLL 2009 shared task. Experimental results show that our method can largely reduce the performance drop on out-of-domain test data.

We propose and implement a cloud architecture of virtual Trusted Platform Modules (TPMs) to improve the usability of TPMs. In this architecture, virtual TPMs can be obtained from the TPM cloud on demand. Hence, the TPM functionality is available for applications that do not have physical TPMs in their local platforms. Moreover, the TPM cloud allows users to access their keys and data in the same virtual TPM even if they move to untrusted platforms. The TPM cloud is easy to access for applications in different languages since cloud computing delivers services in standard protocols. The functionality of the TPM cloud is demonstrated by applying it to implement the Needham-Schroeder public-key protocol for web authentications, such that the strong security provided by TPMs is integrated into high level applications. The chain of trust based on the TPM cloud is discussed and the security properties of the virtual TPMs in the cloud is analyzed.

The commonly used Deep Q Networks is known to overestimate action values under certain conditions. It's also proved that overestimations do harm to performance, which might cause instability and divergence of learning. In this paper, we present the Deep Sarsa and Q Networks (DSQN) algorithm, which can considered as an enhancement to the Deep Q Networks algorithm. First, DSQN algorithm takes advantage of the experience replay and target network techniques in Deep Q Networks to improve the stability of neural networks. Second, double estimator is utilized for Q-learning to reduce overestimations. Especially, we introduce Sarsa learning to Deep Q Networks for removing overestimations further. Finally, DSQN algorithm is evaluated on cart-pole balancing, mountain car and lunarlander control task from the OpenAI Gym. The empirical evaluation results show that the proposed method leads to reduced overestimations, more stable learning process and improved performance.