Online matching markets face increasing needs to accurately learn the matching qualities between demand and supply for effective design of matching policies. However, the growing diversity of participants introduces a high-dimensional challenge in practice, as there are a substantial number of unknown matching rewards and learning all rewards requires a large amount of data. We leverage a natural low-rank matrix structure of the matching rewards in these two-sided markets, and propose to utilize matrix completion (specifically the nuclear norm regularization approach) to accelerate the reward learning process with only a small amount of offline data. A key challenge in our setting is that the matrix entries are observed with matching interference, distinct from the independent sampling assumed in existing matrix completion literature. We propose a new proof technique and prove a near-optimal average accuracy guarantee with improved dependence on the matrix dimensions. Furthermore, to guide matching decisions, we develop a novel "double-enhancement" procedure that refines the nuclear norm regularized estimates and further provides near-optimal entry-wise estimations. Our paper makes the first investigation into adopting matrix completion techniques for matching problems. We also extend our approach to online learning settings for optimal matching and stable matching by incorporating matrix completion in multi-armed bandit algorithms. We present improved regret bounds in matrix dimensions through reduced costs during the exploration phase. Finally, we demonstrate the practical value of our methods using both synthetic data and real data of labor markets.

We study the stochastic bandit problem with ReLU neural network structure. We show that a $\tilde{O}(\sqrt{T})$ regret guarantee is achievable by considering bandits with one-layer ReLU neural networks; to the best of our knowledge, our work is the first to achieve such a guarantee. In this specific setting, we propose an OFU-ReLU algorithm that can achieve this upper bound. The algorithm first explores randomly until it reaches a linear regime, and then implements a UCB-type linear bandit algorithm to balance exploration and exploitation. Our key insight is that we can exploit the piecewise linear structure of ReLU activations and convert the problem into a linear bandit in a transformed feature space, once we learn the parameters of ReLU relatively accurately during the exploration stage. To remove dependence on model parameters, we design an OFU-ReLU+ algorithm based on a batching strategy, which can provide the same theoretical guarantee.

A key challenge to deploying reinforcement learning in practice is avoiding excessive (harmful) exploration in individual episodes. We propose a natural constraint on exploration—uniformly outperforming a conservative policy (adaptively estimated from all data observed thus far), up to a per-episode exploration budget. We design a novel algorithm that uses a UCB reinforcement learning policy for exploration, but overrides it as needed to satisfy our exploration constraint with high probability. Importantly, to ensure unbiased exploration across the state space, our algorithm adaptively determines when to explore. We prove that our approach remains conservative while minimizing regret in the tabular setting. We experimentally validate our results on a sepsis treatment task and an HIV treatment task, demonstrating that our algorithm can learn while ensuring good performance compared to the baseline policy for every patient; the latter task also demonstrates that our approach extends to continuous state spaces via deep reinforcement learning.

The rise of mobile technologies and cloud computing has increased the importance of energy consumption. On-package voltage stacking, where current is recycled between multiple cores, is a potentially effective solution to this growing issue. Two converters, a load-to-load ladder buck converter and a bus-to-load isolated resonant converter, are particularly appropriate for on-package voltage stacking. A four core system composed of either converter topology is evaluated under several current mismatch scenarios in terms of transient and DC voltage drops, voltage ripple, settling time, and power efficiency. The load-to-load buck converter exhibits higher power efficiency and smaller transient voltage drops as compared to the bus-to-load resonant converter. The resonant converter is lower cost, smaller in area, and provides isolation.

Sparse regression has recently been applied to enable transfer learning from very limited data. We study an extension of this approach to unsupervised learning—in particular, learning word embeddings from unstructured text corpora using low-rank matrix factorization. Intuitively, when transferring word embeddings to a new domain, we expect that the embeddings change for only a small number of words—e.g., the ones with novel meanings in that domain. We propose a novel group-sparse penalty that exploits this sparsity to perform transfer learning when there is very little text data available in the target domain—e.g., a single article of text. We prove generalization bounds for our algorithm. Furthermore, we empirically evaluate its effectiveness, both in terms of prediction accuracy in downstream tasks as well as in terms of interpretability of the results.

Recommendation systems can help users process large amounts of information, and generative adversarial networks (GANs) show great potential in recommendation systems. In this paper, we propose a new GAN model to enhance the information flow within the generator based on the information flow between the original generator and discriminator. Our experimental results indicate that our model reduces the discrepancy between the generator and the discriminator. Both the generator and discriminator yield considerable performance improvements compared to other strong baselines. The improvements by [email protected] and MRR are significant, which can reach 30.98% and 30.17%, respectively.

Due to the increasing throughput of high performance integrated circuits, the power consumption of recent high performance computing systems has grown significantly, leading to high on-chip current demand. The large current flowing within the power delivery network leads to challenging issues such as electromigration, low power efficiency, and thermal hotspots. As a technique to reduce on-chip current demand, voltage stacking has become a topic of growing interest within the industrial and academic communities. The challenges of on-chip voltage stacking are however significant. The limitations of relying on on-chip decoupling capacitors when load imbalances occur within a high current system are reviewed. To manage these load imbalances, a ladder topology switched capacitor converter is proposed to regulate the voltages between layers within a voltage stacked system. A 20X improvement in voltage drop is demonstrated on a case study. The current path within a voltage stacked system is quite different from a standard system. A horizontal current path is formed due to the serial connection between layers, producing large parasitic impedances within the power network. The on-chip power network within a voltage stacked system therefore requires careful consideration and specialized design techniques.

The stringent requirements of power noise on complex multi-domain power delivery networks (PDN), and the complicated relationship between signal integrity and power integrity (PI) have led to an ever challenging PI sign-off process. A lumped PDN model is widely used, where the power network is treated as a two-port network with the impedances extracted by an electromagnetic solver. A distributed model of the power network is however preferred during a PI sign-off flow, providing a more accurate circuit model for time domain simulations. Hundreds or even thousands of ports need to be properly evaluated during the PDN extraction process, which can be computationally expensive and error prone. A Python tool is described here to enable a fast and configurable process for distributed port assignment during the PDN extraction process. An enhanced automation flow, integrated with the Python tool, has also been developed to support early power network exploration. In one case study, a 360X speedup in the port assignment process is achieved while revealing a high risk power network within the package. The proposed automation flow is versatile and highly adaptive for different power network topologies.

The conventional power network design process requires iterative modifications to the existing power network to eliminate hot spots and to converge to target impedance parameters. At later stages in the IC design process, this procedure may require significant time and human resources due to the limited flexibility to accommodate necessary changes. Power delivery exploration during early stages of the design process may bring considerable savings to the system development effort. The number of iterations may be greatly reduced by choosing the initial parameters sufficiently close to the optimum. This paper presents a power delivery exploration framework based on constrained global optimization. The power network parameters are estimated at early stages of the development process, while considering both electrical and nonelectrical factors, such as area and cost. A Laplace transform-based circuit simulator is described that is well suited for optimization purposes due to the high computational efficiency when a large number of iterations is required. The proposed framework has been applied to the distribution of voltage domains in a large scale complex integrated system, while minimizing the cost of the decoupling capacitor placement. The optimal number of voltage rails are determined, demonstrating an approximately 40% lower on-chip area than alternative solutions.

Heterographic pun plays a critical role in human writing and literature, which usually has a similar sounding or spelling structure. It is important and difficult research to recognize the heterographic pun because of the ambiguity. However, most existing methods for this task only focus on designing features with rule-based or machine learning methods. In this paper, we propose an end-to-end computational approach - Pronunciation Spelling Understanding Gated Attention (PSUGA) network. For pronunciation, we exploit the hierarchical attention model with phoneme embedding. While for spelling, we consider the character-level, word-level, tag-level, position-level and contextual-level embedding with attention model. To deal with the two parts, we present a gated attention mechanism to control the information integration. We have conducted extensive experiments on SemEval2017 task7 and Pun of the Day datasets. Experimental results show that our approach significantly outperforms state-of-the-art methods.

BACKGROUND: The number of biomedical research articles have increased exponentially with the advancement of biomedicine in recent years. These articles have thus brought a great difficulty in obtaining the needed information of researchers. Information retrieval technologies seek to tackle the problem. However, information needs cannot be completely satisfied by directly introducing the existing information retrieval techniques. Therefore, biomedical information retrieval not only focuses on the relevance of search results, but also aims to promote the completeness of the results, which is referred as the diversity-oriented retrieval. RESULTS: We address the diversity-oriented biomedical retrieval task using a supervised term ranking model. The model is learned through a supervised query expansion process for term refinement. Based on the model, the most relevant and diversified terms are selected to enrich the original query. The expanded query is then fed into a second retrieval to improve the relevance and diversity of search results. To this end, we propose three diversity-oriented optimization strategies in our model, including the diversified term labeling strategy, the biomedical resource-based term features and a diversity-oriented group sampling learning method. Experimental results on TREC Genomics collections demonstrate the effectiveness of the proposed model in improving the relevance and the diversity of search results. CONCLUSIONS: The proposed three strategies jointly contribute to the improvement of biomedical retrieval performance. Our model yields more relevant and diversified results than the state-of-the-art baseline models. Moreover, our method provides a general framework for improving biomedical retrieval performance, and can be used as the basis for future work.

Yufeng Diao, Hongfei Lin, Di Wu, Liang Yang, Kan Xu, Zhihao Yang, Jian Wang, Shaowu Zhang, Bo Xu, Dongyu Zhang. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018.

Over the past decades, aggressive voltage scaling combined with increased power demands has placed stringent requirements on on-chip power quality. Unwanted voltage fluctuations and droops may cause a variety of issues, ranging from glitch power to device malfunction. If revealed at the later stages of the design process, mitigation techniques may become unbearably costly in both time and money. A framework for exploratory power delivery optimization is described to enhance the power delivery network during early stages of the design process in accordance with design specifications. The power delivery design process is converted into a constrained minimization problem, consisting of design metrics combined into objective and constraint functions. The framework supports the optimization of the power network characteristics while considering external, non-electrical design specifications, such as cost and area, providing a comprehensive network analysis capability. In one case study, a 15% reduction in decoupling capacitor placement along with a 38.6% reduction in power consumption is achieved while satisfying performance and power quality constraints.

Learning to rank utilizes machine learning methods to solve ranking problems by constructing ranking models in a supervised way, which needs fixed-length feature vectors of documents as inputs, and outputs the ranking models learned by iteratively reducing the pre-defined ranking loss. The document features are always extracted based on classic textual statistics, and different features contribute differently to ranking performance. Given that well-defined features would contribute more to the retrieval performance, we investigate the usage of autoencoders to enrich the feature representations of documents. Autoencoders, as basic building blocks of deep neural networks, have been successfully used in many text mining tasks for generating effective features. To enrich the feature space for learning to rank, we introduce supervision into the loss functions of autoencoders. Specifically, we first train a linear ranking model on the training data, and then incorporate the learned weights into the reconstruction costs of an autoencoder. Meanwhile, we accumulate the costs of documents for a given query with query-level constraints for producing more useful features. We evaluate the effectiveness of our model on three LETOR datasets, and show that our model can generate effective document features to improve the retrieval performance.

The two main problems of biomedical event extraction are trigger identification and argument detection which can both be considered as classification problems. In this paper, we propose a distributed representation method, which combines context, consisted by dependency-based word embedding, and task-based features represented in a distributed way on deep learning models to realize biomedical event extraction. The experimental results on Multi-Level Event Extraction (MLEE) corpus show higher F-scores compared to the state-of-the-art SVM method. This demonstrates that our proposed method is effective for biomedical event extraction.

The physical dimensions of standard cells constrain the dimensions of power networks, affecting the on-chip power noise. An exploratory modeling methodology is presented for estimating power noise in advanced technology nodes. The models are evaluated for 14, 10, and 7 nm technologies to assess the impact on performance. Scaled technologies are shown to be more sensitive to power noise, resulting in potential loss of performance enhancements achieved by scaling. Stripes between local track rails is evaluated as a means to reduce power noise, exhibiting up to 56.5% improvement in power noise for the 7 nm technology node. A strong dependence on the width of a stripe is observed, indicating that fewer wide stripes are more favorable then many thin stripes. As a promising alternative material for power network interconnects, graphene is shown to exhibit good potential in reducing power noise. The effects of different scaling scenarios of local power rails on power noise are also discussed.

The effects of inductive coupling among TSVs within large TSV arrays are investigated in this paper. A comparison of the equivalent inductance of a paired TSV model and arrayed TSV macromodel is presented for three TSV distribution topologies, grouped, lined, and uniform, within the power network. Modified closed-form expressions are proposed to determine the equivalent inductance of a TSV in these large TSV arrays. Simulation results show that this method achieves hundred times speed improvement as compared to an electromagnetic field solver while maintaining accuracy within 7%.

An efficient monitoring network is very important in accessing the marine environmental quality and its protection and management. In an estuary, there are fronts that separate distinctly different water masses and affect material transport, nutrient distribution, pollutant aggregation, and diffusion. This stratified heterogeneous surface neither satisfies the stationary requirements of kriging, nor can be handled adequately by removing a spatially continuous trend. This article presents a stratified optimization method for a multivariate monitoring network. In this method, principal component analysis (PCA) was used to reduce the dimensionality of the correlated targets, and the mean of surface with nonhomogeneity (MSN) method was adopted to produce the best linear unbiased estimator for a spatially stratified heterogeneous surface that failed to satisfy the requirements for a kriging estimate. The existing monitoring network in the Yangtze River estuary and its adjacent sea, which was designed by purposive sampling year ago was optimized as an illustration. The optimization consisted of two steps: reduce the redundant monitoring sites and then optimally add new sites to the remaining sites. After optimization, the inclusion of 51 sites in the monitoring network was found to produce a smaller total estimated error than that of the current network, which has 70 sites; moreover, the use of 55 sites can produce a higher precision of estimation for all three principal components (PCs) than that of the current 70 sites. The results demonstrated that the proposed method is suitable for optimizing environmental monitoring sites that have dominant stratified nonhomogeneity and that involve multiple factors.

The central issue in language model estimation is smoothing, which is a technique for avoiding zero probability estimation problem and overcoming data sparsity. There are three representative smoothing methods: Jelinek‐Mercer (JM) method; Bayesian smoothing using Dirichlet priors (Dir) method; and absolute discounting (Dis) method, whose parameters are usually estimated empirically. Previous research in information retrieval (IR) on smoothing parameter estimation tends to select a single value from optional values for the collection, but it may not be appropriate for all the queries. The effectiveness of all the optional values should be considered to improve the ranking performance. Recently, learning to rank has become an effective approach to optimize the ranking accuracy by merging the existing retrieval methods. In this article, the smoothing methods for language modeling in information retrieval (LMIR) with different parameters are treated as different retrieval methods, then a learning to rank approach to learn a ranking model based on the features extracted by smoothing methods is presented. In the process of learning, the effectiveness of all the optional smoothing parameters is taken into account for all queries. The experimental results on the Learning to Rank for Information Retrieval (LETOR) LETOR3.0 and LETOR4.0 data sets show that our approach is effective in improving the performance of LMIR.

Recently, the combination of topic models and random fields has been frequently and successfully applied to image classification due to their complementary effect. However, the number of classes is usually needed to be assigned manually. This letter presents an efficient unsupervised semantic classification method for high-resolution satellite images. We add label cost, which can penalize a solution based on a set of labels that appear in it by optimization of energy, to the random fields of latent topics, and an iterative algorithm is thereby proposed to make the number of classes finally be converged to an appropriate level. Compared with other mentioned classification algorithms, our method not only can obtain accurate semantic segmentation results by larger scale structures but also can automatically assign the number of segments. The experimental results on several scenes have demonstrated its effectiveness and robustness.

Learning to rank algorithms are usually grouped into three types: the point wise approach, the pairwise approach, and the listwise approach, according to the input spaces. Much of the prior work is based on the three approaches to learn the ranking model to predict the relevance of a document to a query. In this paper, we focus on the problem of constructing new input space based on groups of documents with the same relevance judgment. A novel approach is proposed based on cross entropy to improve the existing ranking method. The experimental results show that our approach leads to significant improvements in retrieval effectiveness.

In this paper, we use query-level regression as the loss function. The regression loss function has been used in pointwise methods, however pointwise methods ignore the query boundaries and treat the data equally across queries, and thus the effectiveness is limited. We show that regression is an effective loss function for learning to rank when used in query-level. We use neural network to model the ranking function and gradient descent for optimization and refer our method as ListReg. Experimental results show that ListReg significantly outperforms pointwise Regression and the state-of-the-art listwise method in most cases.