Marchenko imaging has the particular ability to generate the subsurface image free of spurious artifacts related to internal multiples. However, conventional Marchenko imaging (C-MI) is performed based on the assumption that the sources and receivers of the recorded seismic data are placed on a flat surface, which is restrictive and hard to satisfy in mountainous areas with rugged topography. The elevation-static correction (or time shift) is a usual solution to ensure this assumption holds. But it only works well if the surface consistency is satisfied. To alleviate the limitations of C-MI in processing seismic data acquired from mountainous areas with rugged topography, we present a technique for conducting Marchenko imaging from a floating datum. The proposed Topography-Marchenko imaging (T-MI) is achieved by estimating an initial down-going focusing function between a floating datum and a focal point in the subsurface. In this work, we use seismic data corrected to a floating datum rather than a final datum as input to the iterative Marchenko scheme to retrieve Green’s functions. The retrieved Marchenko Green’s functions are further used to generate the subsurface image. The T-MI method can effectively avoid imaging distortions caused by the elevation-static correction. The proposed T-MI method is validated through applications to a synthetic model with rugged topography and a land dataset acquired from a mountainous area in Northwest Sichuan, China.

This paper describes novel models tailored for a new application, that of extracting the symptoms mentioned in clinical conversations along with their status. Lack of any publicly available corpus in this privacy-sensitive domain led us to develop our own corpus, consisting of about 3K conversations annotated by professional medical scribes. We propose two novel deep learning approaches to infer the symptom names and their status: (1) a new hierarchical span-attribute tagging (SA-T) model, trained using curriculum learning, and (2) a variant of sequence-to-sequence model which decodes the symptoms and their status from a few speaker turns within a sliding window over the conversation. This task stems from a realistic application of assisting medical providers in capturing symptoms mentioned by patients from their clinical conversations. To reflect this application, we define multiple metrics. From inter-rater agreement, we find that the task is inherently difficult. We conduct comprehensive evaluations on several contrasting conditions and observe that the performance of the models range from an F-score of 0.5 to 0.8 depending on the condition. Our analysis not only reveals the inherent challenges of the task, but also provides useful directions to improve the models.

The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several improvements that make the Skip-gram model more expressive and enable it to learn higher quality vectors more rapidly. We show that by subsampling frequent words we obtain significant speedup, and also learn higher quality representations as measured by our tasks. We also introduce Negative Sampling, a simplified variant of Noise Contrastive Estimation (NCE) that learns more accurate vectors for frequent words compared to the hierarchical softmax. An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of Canada'' and "Air'' cannot be easily combined to obtain "Air Canada''. Motivated by this example, we present a simple and efficient method for finding phrases, and show that their vector representations can be accurately learned by the Skip-gram model. "

Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with billions of parameters using tens of thousands of CPU cores. We have developed a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models. Within this framework, we have developed two algorithms for large-scale distributed training: (i) Downpour SGD, an asynchronous stochastic gradient descent procedure supporting a large number of model replicas, and (ii) Sandblaster, a framework that supports for a variety of distributed batch optimization procedures, including a distributed implementation of L-BFGS. Downpour SGD and Sandblaster L-BFGS both increase the scale and speed of deep network training. We have successfully used our system to train a deep network 100x larger than previously reported in the literature, and achieves state-of-the-art performance on ImageNet, a visual object recognition task with 16 million images and 21k categories. We show that these same techniques dramatically accelerate the training of a more modestly sized deep network for a commercial speech recognition service. Although we focus on and report performance of these methods as applied to training large neural networks, the underlying algorithms are applicable to any gradient-based machine learning algorithm.

Embedded systems are a key enabling technology for the recent vast increase in functionality of a huge list of critical infrastructures. Hybrid automata can be used to model system-level behaviors for the large category of systems that exhibit strong couplings between discrete and continuous dynamics. Many software tools have been developed for hybrid automata to enable model-based design of embedded systems and these software tools are constructed by using their own modeling languages. Model-based design frameworks, such as Model-Integrated Computing (MIC), Model Driven Architecture (MDA), and Model Driven Design (MDD), have been advocated to raise the level of abstraction in software tool design by placing stronger emphasis on the use of software models in the software tool design process. In particular, MIC places strong emphasis on the use of Domain Specific Modeling Languages(DSMLs) and model transformations in design flows. Practical and effective development of formal specifications for DSML semantics within model-based tools can be challenging, but could positively impact adoption and reuse of these tools. The semantic anchoring methodology was developed to address this challenge by formally tying DSMLs to a "semantic unit", which is a formal specification that captures the operational semantics of a specific model of computation. Leveraging our prior work with semantic units, we develop a semantic unit for hybrid automata. In this paper, we explicitly specify the operational semantics of hybrid automata, and develop the corresponding semantic unit and model transformation rules. We demonstrate the effectiveness of the infrastructure in a practical case study involving the hybrid automata DSMLs, HyVisual and ReachLab.

An emerging common trend in model-based design of embedded software and systems is the adoption of domain-specific modeling languages (DSMLs). While abstract syntax metamodeling enables the rapid and inexpensive development of DSMLs, the specification of DSML semantics is still a hard problem. In previous work, we have developed methods and tools for the semantic anchoring of DSMLs. Semantic anchoring introduces a set of reusable "semantic units" that provide reference semantics for basic behavioral categories using the abstract state machine (ASM) framework. In this paper, we extend the semantic anchoring framework to heterogeneous behaviors by developing a method for the composition of semantic units. Semantic unit composition reduces the required effort from DSML designers and improves the quality of the specification. The proposed method is demonstrated through a case study

In previous work we have discussed a semantic anchoring framework that enables the semantic specification of Domain-Specific Modeling languages by specifying semantic anchoring rules to predefined semantic units. This framework is further extended to support heterogeneous systems by developing a method for the composition of semantic units. In this paper, we explain the semantic unit composition through a case study.

Metamodeling facilitates the rapid, inexpensive development of domain-specific modeling languages (DSML-s). However, there are still challenges hindering the wide-scale industrial application of model-based design. One of these unsolved problems is the lack of a practical, effective method for the formal specification of DSML semantics. This problem has negative impact on reusability of DSML-s and analysis tools in domain specific tool chains. To address these issues, we propose a formal well founded methodology with supporting tools to anchor the semantics of DSML-s to precisely defined and validated "semantic units". In our methodology, each of the syntactic and semantic DSML components is defined precisely and completely. The main contribution of our approach is that it moves toward an infrastructure for DSML design that integrates formal methods with practical engineering tools. In this paper we use a mathematical model, Abstract State Machines, a common semantic framework to define the semantic domains of DSML-s.

Data coupling between modules, especially common coupling, has long been considered a source of concern in software design, but the issue is somewhat more complicated for products that are comprised of kernel modules together with optional nonkernel modules. This paper presents a refined categorization of common coupling based on definitions and uses between kernel and nonkernel modules and applies the categorization to a case study. Common coupling is usually avoided when possible because of the potential for introducing risky dependencies among software modules. The relative risk of these dependencies is strongly related to the specific definition-use relationships. In a previous paper, we presented results from a longitudinal analysis of multiple versions of the open-source operating system Linux. This paper applies the new common coupling categorization to version 2.4.20 of Linux, counting the number of instances of common coupling between each of the 26 kernel modules and all the other nonkernel modules. We also categorize each coupling in terms of the definition-use relationships. Results show that the Linux kernel contains a large number of common couplings of all types, raising a concern about the long-term maintainability of Linux.