SearcharxivSearch

arXiv subjects

Anuj Kumar

Publications and source records attributed to Anuj Kumar.

At least 73 records · Page 4Linked to original sources

On local well-posedness of logarithmic inviscid regularizations of generalized SQG equations in borderline Sobolev spaces

This paper studies a family of generalized surface quasi-geostrophic (SQG) equations for an active scalar $θ$ on the whole plane whose velocities have been mildly regularized, for instance, logarithmically. The well-posedness of these regularized models in borderline Sobolev regularity have previously been studied by D. Chae and J. Wu when the velocity $u$ is of lower singularity, i.e., $u=-\nabla^{\perp}Λ^{β-2}p(Λ)θ$, where $p$ is a logarithmic smoothing operator and $β\in [0,1]$. We complete this study by considering the more singular regime $β\in(1,2)$. The main tool is the identification of a suitable linearized system that preserves the underlying commutator structure for the original equation. We observe that this structure is ultimately crucial for obtaining continuity of the flow map. In particular, straightforward applications of previous methods for active transport equations fail to capture the more nuanced commutator structure of the equation in this more singular regime. The proposed linearized system nontrivially modifies the flux of the original system in such a way that it coincides with the original flux when evaluated along solutions of the original system. The requisite estimates are developed for this modified linear system to ensure its well-posedness.

math.AP

El Volumen Louder Por Favor: Code-switching in Task-oriented Semantic Parsing

Being able to parse code-switched (CS) utterances, such as Spanish+English or Hindi+English, is essential to democratize task-oriented semantic parsing systems for certain locales. In this work, we focus on Spanglish (Spanish+English) and release a dataset, CSTOP, containing 5800 CS utterances alongside their semantic parses. We examine the CS generalizability of various Cross-lingual (XL) models and exhibit the advantage of pre-trained XL language models when data for only one language is present. As such, we focus on improving the pre-trained models for the case when only English corpus alongside either zero or a few CS training instances are available. We propose two data augmentation methods for the zero-shot and the few-shot settings: fine-tune using translate-and-align and augment using a generation model followed by match-and-filter. Combining the few-shot setting with the above improvements decreases the initial 30-point accuracy gap between the zero-shot and the full-data settings by two thirds.

cs.CL

On the existence, uniqueness, and smoothing of solutions to the generalized SQG equations in critical Sobolev spaces

This paper studies the dissipative generalized surface quasi-geostrophic equations in a supercritical regime where the order of the dissipation is small relative to order of the velocity, and the velocities are less regular than the advected scalar by up to one order of derivative. We also consider a non-degenerate modification of the endpoint case in which the velocity is less smooth than the advected scalar by slightly more than one order. The existence and uniqueness theory of these equations in the borderline Sobolev spaces is addressed, as well as the instantaneous smoothing effect of their corresponding solutions. In particular, it is shown that solutions emanating from initial data belonging to these Sobolev classes immediately enter a Gevrey class. Such results appear to be the first of its kind for a quasilinear parabolic equation whose coefficients are of higher order than its linear term; they rely on an approximation scheme which modifies the flux in such a way that preserves the underlying commutator structure lost by having to work in the critical space setting, as well as delicate adaptations of well-known commutator estimates to Gevrey classes.

math.AP

An asymptotic theory for the high-Reynolds-number flow past a shear-free circular cylinder

We present an asymptotic theory for analytical characterization of the high-Reynolds-number incompressible flow of a Newtonian fluid past a shear-free circular cylinder. The viscosity-induced modifications to this flow are localized and except in the neighborhood of the rear stagnation point, behave like a linear perturbation of the inviscid flow. Our theory gives a highly accurate description of these modifications by including the contribution from the most significant viscous term in a correctional perturbation expansion about an inviscid base state. We derive the boundary layer equation for the flow and deduce a similarity transformation that leads to a set of infinite, shear-free-condition-incompatible, self-similar solutions. By suitably combining members from this set, we construct an all-boundary-condition-compatible solution to the boundary layer equation. We derive the governing equation for vorticity transport through the narrow wake region and determine its closed-form solution. The near and far field forms of our wake solution are desirably consistent with the boundary layer solution and the well-known, self-similar planar wake solution, respectively. We analyze the flow in the rear stagnation region by formulating an elliptic partial integro-differential equation for the distortion streamfunction that specifically accounts for the fully nonlinear and inviscid dynamics of the viscous correctional terms. The drag force and its atypical logarithmic dependence on Reynolds number, deduced from our matched asymptotic analysis, are in remarkable agreement with the high-resolution simulation results. The logarithmic dependence gives rise to a critical Reynolds number below which the viscous correction term, counterintuitively, reduces the net dissipation in the flow field.

physics.flu-dyn

Best Practices for Data-Efficient Modeling in NLG:How to Train Production-Ready Neural Models with Less Data

Natural language generation (NLG) is a critical component in conversational systems, owing to its role of formulating a correct and natural text response. Traditionally, NLG components have been deployed using template-based solutions. Although neural network solutions recently developed in the research community have been shown to provide several benefits, deployment of such model-based solutions has been challenging due to high latency, correctness issues, and high data needs. In this paper, we present approaches that have helped us deploy data-efficient neural solutions for NLG in conversational systems to production. We describe a family of sampling and modeling techniques to attain production quality with light-weight neural network models using only a fraction of the data that would be necessary otherwise, and show a thorough comparison between each. Our results show that domain complexity dictates the appropriate approach to achieve high data efficiency. Finally, we distill the lessons from our experimental findings into a list of best practices for production-level NLG model development, and present them in a brief runbook. Importantly, the end products of all of the techniques are small sequence-to-sequence models (2Mb) that we can reliably deploy in production.

cs.CL

Conversational Semantic Parsing

The structured representation for semantic parsing in task-oriented assistant systems is geared towards simple understanding of one-turn queries. Due to the limitations of the representation, the session-based properties such as co-reference resolution and context carryover are processed downstream in a pipelined system. In this paper, we propose a semantic representation for such task-oriented conversational systems that can represent concepts such as co-reference and context carryover, enabling comprehensive understanding of queries in a session. We release a new session-based, compositional task-oriented parsing dataset of 20k sessions consisting of 60k utterances. Unlike Dialog State Tracking Challenges, the queries in the dataset have compositional forms. We propose a new family of Seq2Seq models for the session-based parsing above, which achieve better or comparable performance to the current state-of-the-art on ATIS, SNIPS, TOP and DSTC2. Notably, we improve the best known results on DSTC2 by up to 5 points for slot-carryover.

cs.CL

Pressure-driven flows in helical pipes: bounds on flow rate and friction factor

In this paper, we use the well-known background method to obtain a rigorous lower bound on the volume flow rate through a helical pipe driven by a pressure differential in the limit of large Reynolds number. As a consequence, we also obtain an equivalent upper bound on the friction factor. These bounds are also valid for toroidal and straight pipes as limiting cases. By considering a two-dimensional background flow with varying boundary layer thickness along the circumference of the pipe, we obtain these bounds as a function of the curvature and torsion of the pipe and therefore capture the geometrical aspects of the problem. In this paper, we also present a sufficient criterion to find out which pressure-driven flow and surface-velocity-driven flow problems can be tackled using the background method.

physics.flu-dyn

Information Extraction of Clinical Trial Eligibility Criteria

Clinical trials predicate subject eligibility on a diversity of criteria ranging from patient demographics to food allergies. Trials post their requirements as semantically complex, unstructured free-text. Formalizing trial criteria to a computer-interpretable syntax would facilitate eligibility determination. In this paper, we investigate an information extraction (IE) approach for grounding criteria from trials in ClinicalTrials(dot)gov to a shared knowledge base. We frame the problem as a novel knowledge base population task, and implement a solution combining machine learning and context free grammar. To our knowledge, this work is the first criteria extraction system to apply attention-based conditional random field architecture for named entity recognition (NER), and word2vec embedding clustering for named entity linking (NEL). We release the resources and core components of our system on GitHub at https://github.com/facebookresearch/Clinical-Trial-Parser. Finally, we report our per module and end to end performances; we conclude that our system is competitive with Criteria2Query, which we view as the current state-of-the-art in criteria extraction.

cs.CL

Federated User Representation Learning

Collaborative personalization, such as through learned user representations (embeddings), can improve the prediction accuracy of neural-network-based models significantly. We propose Federated User Representation Learning (FURL), a simple, scalable, privacy-preserving and resource-efficient way to utilize existing neural personalization techniques in the Federated Learning (FL) setting. FURL divides model parameters into federated and private parameters. Private parameters, such as private user embeddings, are trained locally, but unlike federated parameters, they are not transferred to or averaged on the server. We show theoretically that this parameter split does not affect training for most model personalization approaches. Storing user embeddings locally not only preserves user privacy, but also improves memory locality of personalization compared to on-server training. We evaluate FURL on two datasets, demonstrating a significant improvement in model quality with 8% and 51% performance increases, and approximately the same level of performance as centralized training with only 0% and 4% reductions. Furthermore, we show that user embeddings learned in FL and the centralized setting have a very similar structure, indicating that FURL can learn collaboratively through the shared parameters while preserving user privacy.

cs.LG

Active Federated Learning

Federated Learning allows for population level models to be trained without centralizing client data by transmitting the global model to clients, calculating gradients locally, then averaging the gradients. Downloading models and uploading gradients uses the client's bandwidth, so minimizing these transmission costs is important. The data on each client is highly variable, so the benefit of training on different clients may differ dramatically. To exploit this we propose Active Federated Learning, where in each round clients are selected not uniformly at random, but with a probability conditioned on the current model and the data on the client to maximize efficiency. We propose a cheap, simple and intuitive sampling scheme which reduces the number of required training iterations by 20-70% while maintaining the same model accuracy, and which mimics well known resampling techniques under certain conditions.

cs.LG

Explore-Exploit: A Framework for Interactive and Online Learning

Interactive user interfaces need to continuously evolve based on the interactions that a user has (or does not have) with the system. This may require constant exploration of various options that the system may have for the user and obtaining signals of user preferences on those. However, such an exploration, especially when the set of available options itself can change frequently, can lead to sub-optimal user experiences. We present Explore-Exploit: a framework designed to collect and utilize user feedback in an interactive and online setting that minimizes regressions in end-user experience. This framework provides a suite of online learning operators for various tasks such as personalization ranking, candidate selection and active learning. We demonstrate how to integrate this framework with run-time services to leverage online and interactive machine learning out-of-the-box. We also present results demonstrating the efficiencies that can be achieved using the Explore-Exploit framework.

cs.LG

Semantic Parsing for Task Oriented Dialog using Hierarchical Representations

Task oriented dialog systems typically first parse user utterances to semantic frames comprised of intents and slots. Previous work on task oriented intent and slot-filling work has been restricted to one intent per query and one slot label per token, and thus cannot model complex compositional requests. Alternative semantic parsing systems have represented queries as logical forms, but these are challenging to annotate and parse. We propose a hierarchical annotation scheme for semantic parsing that allows the representation of compositional queries, and can be efficiently and accurately parsed by standard constituency parsing models. We release a dataset of 44k annotated queries (fb.me/semanticparsingdialog), and show that parsing models outperform sequence-to-sequence approaches on this dataset.

cs.CL

Towards end-to-end spoken language understanding

Spoken language understanding system is traditionally designed as a pipeline of a number of components. First, the audio signal is processed by an automatic speech recognizer for transcription or n-best hypotheses. With the recognition results, a natural language understanding system classifies the text to structured data as domain, intent and slots for down-streaming consumers, such as dialog system, hands-free applications. These components are usually developed and optimized independently. In this paper, we present our study on an end-to-end learning system for spoken language understanding. With this unified approach, we can infer the semantic meaning directly from audio features without the intermediate text representation. This study showed that the trained model can achieve reasonable good result and demonstrated that the model can capture the semantic attention directly from the audio features.

cs.CL

Spin dynamics, short-range order and superparamagnetism in superconducting ferromagnet RuSr2Gd1.4Ce0.6Cu2O10-δ

We report structural, detailed DC and linear/non-linear AC, isothermal and thermoremanent magnetization study of the rutheno-cuprate superconducting ferromagnet RuSr2Gd1.4Ce0.6Cu2O10-δ (GdRu-1222). Structural analysis, by employing Rietveld refinement of X-ray diffraction pattern, reveals that GdRu-1222 crystallizes in tetragonal phase with I4/mmm space group. GdRu-1222 is a reported superconducting ferromagnet with Ru spins magnetic ordering at around 110 K and superconductivity below 40 K in Cu-O2 planes. Detailed linear/non-linear first and higher order harmonic of AC susceptibility studies unveiled the complex magnetism of GdRu-1222. A frequency dependent cusp is observed in AC susceptibility (χac) vs. T measurements. The change in cusp position with applied frequency followed the well known Vogel-Fulcher law, which is a feature to describe a spin-glass (SG) system with possibility of embedded homogeneous/non-homogeneous magnetically interacting/non-interacting ferromagnetic clusters. Such an interpretation is also supported by thermoremanent magnetization (TRM) study at T = 60 K. Detailed interpretation of AC magnetization results revealed the formation of magnetic (ferromagnetic) homogenous/non-homogenous clusters of different sizes embedded in spin-glass (SG) matrix. The magnetization vs. applied field loops do not saturate, even at high applied fields (50 kOe), resulting in the short-range magnetic ordering in the system, which causes the formation of clusters that freeze at low temperatures. Temperature variation of first- and third-order susceptibility harmonics show good agreement with Wohlfarth's model (WM), leading to the superparamagnetism (SPM) state. Detailed magnetization (DC and AC both) results and their analysis helped in explaining the temperature dependent magnetism of the GdRu-1222 system.

cond-mat.supr-con

Synthesis and superconductivity of new BiS2 based superconductor PrO0.5F0.5BiS2

We report synthesis and superconductivity at 3.7K in PrO0.5F0.5BiS2. The newly discovered material belongs to the layered sulfide based REO0.5F0.5BiS2 compounds having ZrCuSiAs type structure. The bulk polycrystalline compound is synthesized by vacuum encapsulation technique at 7800C in single step. Detailed structural analysis has shown that the as synthesized PrO0.5F0.5BiS2 is crystallized in tetragonal P4/nmm space group with lattice parameters a = 4.015(5) Å, c = 13.362(4) Å. Bulk superconductivity is observed in PrO0.5F0.5BiS2 below 4K from magnetic and transport measurements. Electrical transport measurements showed superconducting transition temperature (Tc) onset at 3.7K and Tc (ρ=0) at 3.1K. Hump at Tc related to superconducting transition is not observed in heat capacity measurement and rather a Schottky-type anomaly is observed at below ~6K. The compound is slightly semiconducting in normal state. Isothermal magnetization (MH) exhibited typical type II behavior with lower critical field (Hc1) of around 8Oe.

cond-mat.supr-con

Appearance of superconductivity in new BiS2 based layered LaO0.5F0.5BiS2

We report here synthesis, structural, DC magnetization, and transport studies of new BiS2 based layered LaO0.5F0.5BiS2 superconductor [1]. The sample was synthesized by conventional solid state route via vacuum encapsulation technique at 800oC for 12h. LaO0.5F0.5BiS2 crystallizes in tetragonal P4/nmm space group with lattice parameters a = 4.0703(5)Å, c = 13.3902(4)Å. Bulk superconductivity is confirmed with superconducting transition temperature (Tc) of 2.7K by DC magnetization measurements. The Isothermal magnetization (MH) measurements showed closed loops with clear signatures of flux pinning and irreversible behavior. The resistivity measurements confirmed an onset Tc of 2.7K. The magneto-transport ρ(T,H) measurements showed a resistive broadening and decrease in Tc (ρ=0) to lower temperatures with increasing magnetic field. The magnetic phase diagram involving upper critical and irreversibility fields as a function of temperature has been ascertained. Our DC magnetization and electrical transport measurements confirm the appearance of bulk superconductivity in LaO0.5F0.5BiS2 superconductor.

cond-mat.supr-con

Superconductivity at 5K in NdO0.5F0.5BiS2

We report appearance of superconductivity at 5K in NdO0.5F0.5BiS2 and supplement the discovery [1] of the same in layered sulfide based ZrCuSiAs type compounds. The bulk polycrystalline compound is synthesized by conventional solid state route via vacuum encapsulation technique. Detailed structural analysis showed that the studied compound is crystallized in tetragonal P4/nmm space group with lattice parameters a = 3.9911(3) Å, c = 13.3830(2) Å. Bulk superconductivity is established in NdO0.5F0.5BiS2 at 5K by both transport and magnetic measurements. Electrical transport measurements showed superconducting Tc onset at 5.2K and Tc (ρ=0) at 4.7K. Under applied magnetic field both Tc onset and Tc (ρ =0) decrease to lower temperatures and an upper critical field [Hc2(0)] of above 23kOe is estimated. Both AC and DC magnetic susceptibility measurements showed bulk superconductivity below 5K. Isothermal magnetization (MH) exhibited typical type II behavior with lower critical field (Hc1) of around 15Oe. Isothermal magnetization (MH) exhibited typical type-II behavior with lower critical field (Hc1) of around 15Oe. Specific heat [Cp(T)] is investigated in the temperature range of 1.9-50K in zero external magnetic field. A Schottky-type anomaly is observed at low temperature below 7K. This low temperature Schottky can be attributed to the change in the entropy of the system.

cond-mat.supr-con

Bulk Superconductivity in Bismuth-oxy-sulfide Bi4O4S3

Very recent report [1] on observation of superconductivity in Bi4O4S3 could potentially reignite the search for superconductivity in a broad range of layered sulphides. We report here synthesis of Bi4O4S3 at 5000C by vacuum encapsulation technique and basic characterizations. Detailed structural, magnetization, and electrical transport results are reported. Bi4O4S3 is contaminated by small amounts of Bi2S3 and Bi impurities. The majority phase is tetragonal I4/mmm space group with lattice parameters a = 3.9697(2)Å, c = 41.3520(1)Å. Both AC and DC magnetization measurements confirmed that Bi4O4S3 is a bulk superconductor with superconducting transition temperature (Tc) of 4.4K. Isothermal magnetization (MH) measurements indicated closed loops with clear signatures of flux pinning and irreversible behavior. The lower critical field (Hc1) at 2K, of the new superconductor is found to be ~39 Oe. The magneto-transport R(T, H) measurements showed a resistive broadening and decrease in Tc (R=0) to lower temperatures with increasing magnetic field. The extrapolated upper critical field Hc2(0) is ~ 310kOe with a corresponding Ginzburg-Landau coherence length of ~100Å . In the normal state the ρ ~ T2 is not indicated. Our magnetization and electrical transport measurements substantiate the appearance of bulk superconductivity in as synthesized Bi4O4S3. On the other hand same temperature heat treated Bi is not superconducting, thus excluding possibility of impurity driven superconductivity in the newly discovered Bi4O4S3 superconductor.

cond-mat.supr-con