Searcharxiv⌕ Search

arXiv subjects

Yuan Tang

Publications and source records attributed to Yuan Tang.

48 records · Page 3Linked to original sources
arXiv2020

A Scalable and Cloud-Native Hyperparameter Tuning System

In this paper, we introduce Katib: a scalable, cloud-native, and production-ready hyperparameter tuning system that is agnostic of the underlying machine learning framework. Though there are multiple hyperparameter tuning systems available, this is the first one that caters to the needs of both users and administrators of the system. We present the motivation and design of the system and contrast it with existing hyperparameter tuning systems, especially in terms of multi-tenancy, scalability, fault-tolerance, and extensibility. It can be deployed on local machines, or hosted as a service in on-premise data centers, or in private/public clouds. We demonstrate the advantage of our system using experimental results as well as real-world, production use cases. Katib has active contributors from multiple companies and is open-sourced at \emph{https://github.com/kubeflow/katib} under the Apache 2.0 license.

cs.DC↗
arXiv2020

An exploration of Homogeneous Isotropic Turbulence in He II Using Particle Tracking Velocimetry

Despite being a quantum two-fluid system, superfluid helium-4 (He II) is observed to behave similarly to classical fluids when a flow is generated by mechanical forcing. This similarity has brought up the feasibility of utilizing He II for high Reynolds number classical turbulence research, considering the small kinematic viscosity of He II. However, it has been suggested that the non-classical dissipation mechanism in He II at small scales may alter its turbulent statistics and intermittency. In this work, we report our study of a nearly homogeneous and isotropic turbulence (HIT) generated by a towed grid in He II. We measure the velocity field using particle tracking velocimetry with solidified deuterium particles as the tracers. By correlating the velocities measured simultaneously on different particle trajectories or at different times along the same particle trajectory, we are able to conduct both Eulerian and Lagrangian flow analyses. Spatial velocity structure functions obtained through the Eulerian analysis show scaling behaviors in the inertial subrange similar to that for classical HIT but with enhanced intermittency. The Lagrangian analysis allows us to examine the flow statistics down to below the dissipation length scale. Interestingly, strong deviations from the classical scaling behaviors are observed in this regime. We discuss how these deviations may relate to the motion of quantized vortices in the superfluid component in He II.

cond-mat.other↗
arXiv2020

SQLFlow: A Bridge between SQL and Machine Learning

Industrial AI systems are mostly end-to-end machine learning (ML) workflows. A typical recommendation or business intelligence system includes many online micro-services and offline jobs. We describe SQLFlow for developing such workflows efficiently in SQL. SQL enables developers to write short programs focusing on the purpose (what) and ignoring the procedure (how). Previous database systems extended their SQL dialect to support ML. SQLFlow (https://sqlflow.org/sqlflow ) takes another strategy to work as a bridge over various database systems, including MySQL, Apache Hive, and Alibaba MaxCompute, and ML engines like TensorFlow, XGBoost, and scikit-learn. We extended SQL syntax carefully to make the extension working with various SQL dialects. We implement the extension by inventing a collaborative parsing algorithm. SQLFlow is efficient and expressive to a wide variety of ML techniques -- supervised and unsupervised learning; deep networks and tree models; visual model explanation in addition to training and prediction; data processing and feature extraction in addition to ML. SQLFlow compiles a SQL program into a Kubernetes-native workflow for fault-tolerable execution and on-cloud deployment. Current industrial users include Ant Financial, DiDi, and Alibaba Group.

cs.DB↗
arXiv2019

Improving the Space-Time Efficiency of Processor-Oblivious Matrix Multiplication Algorithms

Classic cache-oblivious parallel matrix multiplication algorithms achieve optimality either in time or space, but not both, which promotes lots of research on the best possible balance or tradeoff of such algorithms. We study modern processor-oblivious runtime systems and figure out several ways to improve algorithm's time bound while still bounding space and cache requirements to be asymptotically optimal. By our study, we give out sublinear time, optimal work, space and cache algorithms for both general matrix multiplication on a semiring and Strassen-like fast algorithm. Our experiments also show such algorithms have empirical advantages over classic counterparts. Our study provides new insights and research angles on how to optimize cache-oblivious parallel algorithms from both theoretical and empirical perspectives.

cs.DC↗
arXiv2019

Nested Dataflow Algorithms for Dynamic Programming Recurrences with more than O(1) Dependency

Dynamic programming problems have wide applications in real world and have been studied extensively in both serial and parallel settings. In 1994, Galil and Park developed work-efficient and sublinear-time algorithms for several important dynamic programming problems based on the closure method and matrix product method. However, in the same paper, they raised an open question whether such an algorithm exists for the general GAP problem. % In this paper, we answer their question by developing the first work-efficient and sublinear-time GAP algorithm based on the closure method and Nested Dataflow method. % We also improve the time bounds of classic work-efficient, cache-oblivious and cache-efficient algorithms for the 1D problem and GAP problem, respectively.

cs.DS↗
arXiv2017

TensorFlow Estimators: Managing Simplicity vs. Flexibility in High-Level Machine Learning Frameworks

We present a framework for specifying, training, evaluating, and deploying machine learning models. Our focus is on simplifying cutting edge machine learning for practitioners in order to bring such technologies into production. Recognizing the fast evolution of the field of deep learning, we make no attempt to capture the design space of all possible model architectures in a domain- specific language (DSL) or similar configuration language. We allow users to write code to define their models, but provide abstractions that guide develop- ers to write models in ways conducive to productionization. We also provide a unifying Estimator interface, making it possible to write downstream infrastructure (e.g. distributed training, hyperparameter tuning) independent of the model implementation. We balance the competing demands for flexibility and simplicity by offering APIs at different levels of abstraction, making common model architectures available out of the box, while providing a library of utilities designed to speed up experimentation with model architectures. To make out of the box models flexible and usable across a wide range of problems, these canned Estimators are parameterized not only over traditional hyperparameters, but also using feature columns, a declarative specification describing how to interpret input data. We discuss our experience in using this framework in re- search and production environments, and show the impact on code health, maintainability, and development speed.

cs.DC↗
arXiv2016

TF.Learn: TensorFlow's High-level Module for Distributed Machine Learning

TF.Learn is a high-level Python module for distributed machine learning inside TensorFlow. It provides an easy-to-use Scikit-learn style interface to simplify the process of creating, configuring, training, evaluating, and experimenting a machine learning model. TF.Learn integrates a wide range of state-of-art machine learning algorithms built on top of TensorFlow's low level APIs for small to large-scale supervised and unsupervised problems. This module focuses on bringing machine learning to non-specialists using a general-purpose high-level language as well as researchers who want to implement, benchmark, and compare their new methods in a structured environment. Emphasis is put on ease of use, performance, documentation, and API consistency.

cs.DC↗
arXiv2016

lfda: An R Package for Local Fisher Discriminant Analysis and Visualization

Local Fisher discriminant analysis is a localized variant of Fisher discriminant analysis and it is popular for supervised dimensionality reduction method. lfda is an R package for performing local Fisher discriminant analysis, including its variants such as kernel local Fisher discriminant analysis and semi-supervised local Fisher discriminant analysis. It also provides visualization functions to easily visualize the dimension reduction results by using either rgl for 3D visualization or ggfortify for 2D visualization in ggplot2 style.

stat.AP↗
arXiv2016

Incorporating Hierarchical Structure Into Dynamic Systems: An Application Of Estimating HIV Epidemics At Sub-National And Sub-Population Level

Dynamic models have been successfully used in producing estimates of HIV epidemics at national level, due to their epidemiological nature and their ability to simultaneously estimate prevalence, incidence, and mortality rates. Recently, HIV interventions and policies have required more information at sub-national and sub-population levels to support local planning, decision making and resource allocation. Unfortunately, many areas and high-risk groups lack sufficient data for deriving stable and reliable results, and this is a critical technical barrier to more stratified estimates. One solution is to borrow information from other areas and groups within the same country. However, directly assuming hierarchical structures within the HIV dynamic models is complicated and computationally time consuming. In this paper, we propose a simple and innovative way to incorporate the hierarchical information into the dynamic systems by using auxiliary data. The proposed method efficiently uses information from multiple areas and risk groups within each country without increasing the computational burden. As a result, the new model improves predictive ability in general with especially significant improvement in areas and risk groups with sparse data.

stat.ME↗
arXiv2016

Extending the Nested Parallel Model to the Nested Dataflow Model with Provably Efficient Schedulers

The nested parallel (a.k.a. fork-join) model is widely used for writing parallel programs. However, the two composition constructs, i.e. "$\parallel$" (parallel) and "$;$" (serial), are insufficient in expressing "partial dependencies" or "partial parallelism" in a program. We propose a new dataflow composition construct "$\leadsto$" to express partial dependencies in algorithms in a processor- and cache-oblivious way, thus extending the Nested Parallel (NP) model to the \emph{Nested Dataflow} (ND) model. We redesign several divide-and-conquer algorithms ranging from dense linear algebra to dynamic-programming in the ND model and prove that they all have optimal span while retaining optimal cache complexity. We propose the design of runtime schedulers that map ND programs to multicore processors with multiple levels of possibly shared caches (i.e, Parallel Memory Hierarchies) and provide theoretical guarantees on their ability to preserve locality and load balance. For this, we adapt space-bounded (SB) schedulers for the ND model. We show that our algorithms have increased "parallelizability" in the ND model, and that SB schedulers can use the extra parallelizability to achieve asymptotically optimal bounds on cache misses and running time on a greater number of processors than in the NP model. The running time for the algorithms in this paper is $O\left(\frac{\sum_{i=0}^{h-1} Q^{*}({\mathsf t};σ\cdot M_i)\cdot C_i}{p}\right)$, where $Q^{*}$ is the cache complexity of task ${\mathsf t}$, $C_i$ is the cost of cache miss at level-$i$ cache which is of size $M_i$, $σ\in(0,1)$ is a constant, and $p$ is the number of processors in an $h$-level cache hierarchy.

cs.DC↗
arXiv2002

A Sample of Ultra Steep Spectrum Sources Selected from the Westerbork In the Southern Hemisphere (WISH) survey

The 352 MHz Westerbork In the Southern Hemisphere (WISH) survey is the southern extension of the WENSS, covering 1.60 sr between -9 < DEC < -26 to a limiting flux density of ~18 mJy (5sigma). Due to the very low elevation of the observations, the survey has a much lower resolution in declination than in right ascension (54" x 54"cosec(DEC)). A correlation with the 1.4 GHz NVSS shows that the positional accuracy is less constrained in declination than in right ascension, but there is no significant systematic error. We present a source list containing 73570 sources. We correlate this WISH catalogue with the NVSS to construct a sample of faint Ultra Steep Spectrum (USS) sources, which is accessible for follow-up studies with large optical telescopes in the southern hemisphere. This sample is aimed at increasing the number of known high redshift radio galaxies to allow detailed follow-up studies of these massive galaxies and their environments in the early Universe.

astro-ph↗