arXiv subjects

Wentao Wu

Publications and source records attributed to Wentao Wu.

At least 55 records · Page 3Linked to original sources

arXiv2022

Stochastic Gradient Descent without Full Data Shuffle

Stochastic gradient descent (SGD) is the cornerstone of modern machine learning (ML) systems. Despite its computational efficiency, SGD requires random data access that is inherently inefficient when implemented in systems that rely on block-addressable secondary storage such as HDD and SSD, e.g., TensorFlow/PyTorch and in-DB ML systems over large files. To address this impedance mismatch, various data shuffling strategies have been proposed to balance the convergence rate of SGD (which favors randomness) and its I/O performance (which favors sequential access). In this paper, we first conduct a systematic empirical study on existing data shuffling strategies, which reveals that all existing strategies have room for improvement -- they all suffer in terms of I/O performance or convergence rate. With this in mind, we propose a simple but novel hierarchical data shuffling strategy, CorgiPile. Compared with existing strategies, CorgiPile avoids a full data shuffle while maintaining comparable convergence rate of SGD as if a full shuffle were performed. We provide a non-trivial theoretical analysis of CorgiPile on its convergence behavior. We further integrate CorgiPile into PyTorch by designing new parallel/distributed shuffle operators inside a new CorgiPileDataSet API. We also integrate CorgiPile into PostgreSQL by introducing three new physical operators with optimizations. Our experimental results show that CorgiPile can achieve comparable convergence rate with the full shuffle based SGD for both deep learning and generalized linear models. For deep learning models on ImageNet dataset, CorgiPile is 1.5X faster than PyTorch with full data shuffle. For in-DB ML with linear models, CorgiPile is 1.6X-12.8X faster than two state-of-the-art in-DB ML systems, Apache MADlib and Bismarck, on both HDD and SSD.

cs.LG↗

arXiv2022

Data Debugging with Shapley Importance over End-to-End Machine Learning Pipelines

Developing modern machine learning (ML) applications is data-centric, of which one fundamental challenge is to understand the influence of data quality to ML training -- "Which training examples are 'guilty' in making the trained ML model predictions inaccurate or unfair?" Modeling data influence for ML training has attracted intensive interest over the last decade, and one popular framework is to compute the Shapley value of each training example with respect to utilities such as validation accuracy and fairness of the trained ML model. Unfortunately, despite recent intensive interest and research, existing methods only consider a single ML model "in isolation" and do not consider an end-to-end ML pipeline that consists of data transformations, feature extractors, and ML training. We present DataScope (ease.ml/datascope), the first system that efficiently computes Shapley values of training examples over an end-to-end ML pipeline, and illustrate its applications in data debugging for ML training. To this end, we first develop a novel algorithmic framework that computes Shapley value over a specific family of ML pipelines that we call canonical pipelines: a positive relational algebra query followed by a K-nearest-neighbor (KNN) classifier. We show that, for many subfamilies of canonical pipelines, computing Shapley value is in PTIME, contrasting the exponential complexity of computing Shapley value in general. We then put this to practice -- given an sklearn pipeline, we approximate it with a canonical pipeline to use as a proxy. We conduct extensive experiments illustrating different use cases and utilities. Our results show that DataScope is up to four orders of magnitude faster over state-of-the-art Monte Carlo-based methods, while being comparably, and often even more, effective in data debugging.

cs.LG↗

arXiv2022

Factor Windows: Cost-based Query Rewriting for Optimizing Correlated Window Aggregates

Window aggregates are ubiquitous in stream processing. In Azure Stream Analytics (ASA), a stream processing service hosted by Microsoft's Azure cloud, we see many customer queries that contain aggregate functions (such as MIN and MAX) over multiple correlated windows (e.g., tumbling windows of length five minutes and ten minutes) defined on the same event stream. In this paper, we present a cost-based optimization framework for optimizing such queries by sharing computation among multiple windows. In particular, we introduce the notion of factor windows, which are auxiliary windows that are not in the input query but may nevertheless help reduce the overall computation cost, and our cost-based optimizer can produce rewritten query plans that have lower costs than the original query plan by utilizing factor windows. Since our optimization techniques are at the level of query (plan) rewriting, they can be implemented on any stream processing system that supports a declarative, SQL-like query language without changing the underlying query execution engine. We formalize the shared computation problem, present the optimization techniques in detail, and report evaluation results over both synthetic and real datasets. Our results show that, compared to the original query plans, the rewritten plans output by our cost-based optimizer can yield significantly higher (up to 16.8x) throughput.

cs.DB↗

arXiv2021

OpenBox: A Generalized Black-box Optimization Service

Black-box optimization (BBO) has a broad range of applications, including automatic machine learning, engineering, physics, and experimental design. However, it remains a challenge for users to apply BBO methods to their problems at hand with existing software packages, in terms of applicability, performance, and efficiency. In this paper, we build OpenBox, an open-source and general-purpose BBO service with improved usability. The modular design behind OpenBox also facilitates flexible abstraction and optimization of basic BBO components that are common in other existing systems. OpenBox is distributed, fault-tolerant, and scalable. To improve efficiency, OpenBox further utilizes "algorithm agnostic" parallelization and transfer learning. Our experimental results demonstrate the effectiveness and efficiency of OpenBox compared to existing systems.

cs.LG↗

arXiv2021

VolcanoML: Speeding up End-to-End AutoML via Scalable Search Space Decomposition

End-to-end AutoML has attracted intensive interests from both academia and industry, which automatically searches for ML pipelines in a space induced by feature engineering, algorithm/model selection, and hyper-parameter tuning. Existing AutoML systems, however, suffer from scalability issues when applying to application domains with large, high-dimensional search spaces. We present VolcanoML, a scalable and extensible framework that facilitates systematic exploration of large AutoML search spaces. VolcanoML introduces and implements basic building blocks that decompose a large search space into smaller ones, and allows users to utilize these building blocks to compose an execution plan for the AutoML problem at hand. VolcanoML further supports a Volcano-style execution model - akin to the one supported by modern database systems - to execute the plan constructed. Our evaluation demonstrates that, not only does VolcanoML raise the level of expressiveness for search space decomposition in AutoML, it also leads to actual findings of decomposition strategies that are significantly more efficient than the ones employed by state-of-the-art AutoML systems such as auto-sklearn.

cs.LG↗

arXiv2021

Towards Demystifying Serverless Machine Learning Training

The appeal of serverless (FaaS) has triggered a growing interest on how to use it in data-intensive applications such as ETL, query processing, or machine learning (ML). Several systems exist for training large-scale ML models on top of serverless infrastructures (e.g., AWS Lambda) but with inconclusive results in terms of their performance and relative advantage over "serverful" infrastructures (IaaS). In this paper we present a systematic, comparative study of distributed ML training over FaaS and IaaS. We present a design space covering design choices such as optimization algorithms and synchronization protocols, and implement a platform, LambdaML, that enables a fair comparison between FaaS and IaaS. We present experimental results using LambdaML, and further develop an analytic model to capture cost/performance tradeoffs that must be considered when opting for a serverless infrastructure. Our results indicate that ML training pays off in serverless only for models with efficient (i.e., reduced) communication and that quickly converge. In general, FaaS can be much faster but it is never significantly cheaper than IaaS.

cs.DC↗

arXiv2021

A Data Quality-Driven View of MLOps

Developing machine learning models can be seen as a process similar to the one established for traditional software development. A key difference between the two lies in the strong dependency between the quality of a machine learning model and the quality of the data used to train or perform evaluations. In this work, we demonstrate how different aspects of data quality propagate through various stages of machine learning development. By performing a joint analysis of the impact of well-known data quality dimensions and the downstream machine learning process, we show that different components of a typical MLOps pipeline can be efficiently designed, providing both a technical and theoretical perspective.

cs.LG↗

arXiv2020

Nearest Neighbor Classifiers over Incomplete Information: From Certain Answers to Certain Predictions

Machine learning (ML) applications have been thriving recently, largely attributed to the increasing availability of data. However, inconsistency and incomplete information are ubiquitous in real-world datasets, and their impact on ML applications remains elusive. In this paper, we present a formal study of this impact by extending the notion of Certain Answers for Codd tables, which has been explored by the database research community for decades, into the field of machine learning. Specifically, we focus on classification problems and propose the notion of "Certain Predictions" (CP) -- a test data example can be certainly predicted (CP'ed) if all possible classifiers trained on top of all possible worlds induced by the incompleteness of data would yield the same prediction. We study two fundamental CP queries: (Q1) checking query that determines whether a data example can be CP'ed; and (Q2) counting query that computes the number of classifiers that support a particular prediction (i.e., label). Given that general solutions to CP queries are, not surprisingly, hard without assumption over the type of classifier, we further present a case study in the context of nearest neighbor (NN) classifiers, where efficient solutions to CP queries can be developed -- we show that it is possible to answer both queries in linear or polynomial time over exponentially many possible worlds. We demonstrate one example use case of CP in the important application of "data cleaning for machine learning (DC for ML)." We show that our proposed CPClean approach built based on CP can often significantly outperform existing techniques in terms of classification accuracy with mild manual cleaning effort.

cs.LG↗

arXiv2020

A Note On Operator-Level Query Execution Cost Modeling

External query execution cost modeling using query execution feedback has found its way in various database applications such as admission control and query scheduling. Existing techniques in general fall into two categories, plan-level cost modeling and operator-level cost modeling. It has been shown in the literature that operator-level cost modeling can often significantly outperform plan-level cost modeling. In this paper, we study operator-level cost modeling from a robustness perspective. We address two main challenges in practice regarding limited execution feedback (for certain operators) and mixed cost estimates due to the use of multiple cost modeling techniques. We propose a framework that deals with these issues and present a comprehensive analysis of this framework. We further provide a case study to demonstrate the efficacy of our framework in the context of index tuning, which is itself a new application of external cost modeling techniques.

cs.DB↗

arXiv2019

Data Science through the looking glass and what we found there

The recent success of machine learning (ML) has led to an explosive growth both in terms of new systems and algorithms built in industry and academia, and new applications built by an ever-growing community of data science (DS) practitioners. This quickly shifting panorama of technologies and applications is challenging for builders and practitioners alike to follow. In this paper, we set out to capture this panorama through a wide-angle lens, by performing the largest analysis of DS projects to date, focusing on questions that can help determine investments on either side. Specifically, we download and analyze: (a) over 6M Python notebooks publicly available on GITHUB, (b) over 2M enterprise DS pipelines developed within COMPANYX, and (c) the source code and metadata of over 900 releases from 12 important DS libraries. The analysis we perform ranges from coarse-grained statistical characterizations to analysis of library imports, pipelines, and comparative studies across datasets and time. We report a large number of measurements for our readers to interpret, and dare to draw a few (actionable, yet subjective) conclusions on (a) what systems builders should focus on to better serve practitioners, and (b) what technologies should practitioners bet on given current trends. We plan to automate this analysis and release associated tools and results periodically.

cs.LG↗

arXiv2019

Quantitative Overfitting Management for Human-in-the-loop ML Application Development with ease.ml/meter

Simplifying machine learning (ML) application development, including distributed computation, programming interface, resource management, model selection, etc, has attracted intensive interests recently. These research efforts have significantly improved the efficiency and the degree of automation of developing ML models. In this paper, we take a first step in an orthogonal direction towards automated quality management for human-in-the-loop ML application development. We build ease. ml/meter, a system that can automatically detect and measure the degree of overfitting during the whole lifecycle of ML application development. ease. ml/meter returns overfitting signals with strong probabilistic guarantees, based on which developers can take appropriate actions. In particular, ease. ml/meter provides principled guidelines to simple yet nontrivial questions regarding desired validation and test data sizes, which are among commonest questions raised by developers. The fact that ML application development is typically a continuous procedure further worsens the situation: The validation and test data sets can lose their statistical power quickly due to multiple accesses, especially in the presence of adaptive analysis. ease. ml/meter addresses these challenges by leveraging a collection of novel techniques and optimizations, resulting in practically tractable data sizes without compromising the probabilistic guarantees. We present the design and implementation details of ease. ml/meter, as well as detailed theoretical analysis and empirical evaluation of its effectiveness.

cs.LG↗

arXiv2019

Continuous Integration of Machine Learning Models with ease.ml/ci: Towards a Rigorous Yet Practical Treatment

Continuous integration is an indispensable step of modern software engineering practices to systematically manage the life cycles of system development. Developing a machine learning model is no difference - it is an engineering process with a life cycle, including design, implementation, tuning, testing, and deployment. However, most, if not all, existing continuous integration engines do not support machine learning as first-class citizens. In this paper, we present ease.ml/ci, to our best knowledge, the first continuous integration system for machine learning. The challenge of building ease.ml/ci is to provide rigorous guarantees, e.g., single accuracy point error tolerance with 0.999 reliability, with a practical amount of labeling effort, e.g., 2K labels per test. We design a domain specific language that allows users to specify integration conditions with reliability constraints, and develop simple novel optimizations that can lower the number of labels required by up to two orders of magnitude for test conditions popularly used in real production systems.

cs.LG↗

arXiv2018

Stochastic simulation of collisions using high-order weak convergence algorithms

Electron collisions, described by stochastic differential equations (SDEs), were simulated using a second-order weak convergence algorithm. Using stochastic analysis, we constructed an SDE for energetic electrons in Lorentz plasma to incorporate the collisions with heavy static ions as well as background thermal electrons. Instead of focusing on the errors of each trajectory, we initiated from the weak convergence condition to directly improve the numerical accuracy of the distribution and its moments. A second-order weak convergence algorithm was developed, estimated the order of the algorithm in a collision simulation case, and compared it with the results obtained using the Euler-Maruyama method and the Cadjan-Ivanov method. For a fixed error level, the second order method largely reduce the computation time compared to the Euler-Maruyama method and the Cadjan-Ivnaov method. Finally, the backward runaway phenomenon was numerically studied and the probability of this runaway was calculated using the new method.

physics.plasm-ph↗

arXiv2018

Reducing Noise for PIC Simulations Using Kernel Density Estimation Algorithm

Noise is a major concern for Particle-In-Cell (PIC) simulations. We propose a new theoretical and algorithmic framework to evaluate and reduce the noise level for PIC simulations based on the Kernel Density Estimation (KDE) theory, which has been widely adopted in machine learning and big data science. According to this framework, the error on particle density estimation for PIC simulations can be characterized by the Mean Integrated Square Error (MISE), which consists of two parts, systematic error and noise. A careful analysis shows that in the standard PIC methods noise is the dominate error, and the noise level can be reduced if we select different shape functions that are capable of balancing the systematic error and the noise. To improve performance, we use the von Mises distribution as the shape function and seek an optimal particle width that minimizes the MISE, represented by a Cross-Validation (CV) function. This procedure significantly reduces both the noise and the MISE for PIC simulations. A particle-wise width adjustment algorithm and a width update algorithm are further developed to reduce the MISE. Simulations using the examples of Langmuir wave and Landau Damping demonstrate that the KDE algorithm developed in the present study reduces the noise level on density estimation by 98%, and gives a much more accurate result on the linear damping rate compared to the standard PIC methods. Meanwhile, it is computational efficient that can save 40% time to achieve the same accuracy.

physics.comp-ph↗

arXiv2017

MLBench: How Good Are Machine Learning Clouds for Binary Classification Tasks on Structured Data?

We conduct an empirical study of machine learning functionalities provided by major cloud service providers, which we call machine learning clouds. Machine learning clouds hold the promise of hiding all the sophistication of running large-scale machine learning: Instead of specifying how to run a machine learning task, users only specify what machine learning task to run and the cloud figures out the rest. Raising the level of abstraction, however, rarely comes free - a performance penalty is possible. How good, then, are current machine learning clouds on real-world machine learning workloads? We study this question with a focus on binary classication problems. We present mlbench, a novel benchmark constructed by harvesting datasets from Kaggle competitions. We then compare the performance of the top winning code available from Kaggle with that of running machine learning clouds from both Azure and Amazon on mlbench. Our comparative study reveals the strength and weakness of existing machine learning clouds and points out potential future directions for improvement.

cs.DC↗

arXiv2017

Ease.ml: Towards Multi-tenant Resource Sharing for Machine Learning Workloads

We present ease.ml, a declarative machine learning service platform we built to support more than ten research groups outside the computer science departments at ETH Zurich for their machine learning needs. With ease.ml, a user defines the high-level schema of a machine learning application and submits the task via a Web interface. The system automatically deals with the rest, such as model selection and data movement. In this paper, we describe the ease.ml architecture and focus on a novel technical problem introduced by ease.ml regarding resource allocation. We ask, as a "service provider" that manages a shared cluster of machines among all our users running machine learning workloads, what is the resource allocation strategy that maximizes the global satisfaction of all our users? Resource allocation is a critical yet subtle issue in this multi-tenant scenario, as we have to balance between efficiency and fairness. We first formalize the problem that we call multi-tenant model selection, aiming for minimizing the total regret of all users running automatic model selection tasks. We then develop a novel algorithm that combines multi-armed bandits with Bayesian optimization and prove a regret bound under the multi-tenant setting. Finally, we report our evaluation of ease.ml on synthetic data and on one service we are providing to our users, namely, image classification with deep neural networks. Our experimental evaluation results show that our proposed solution can be up to 9.8x faster in achieving the same global quality for all users as the two popular heuristics used by our users before ease.ml.

cs.DB↗

arXiv2016

Sampling-Based Query Re-Optimization

Despite of decades of work, query optimizers still make mistakes on "difficult" queries because of bad cardinality estimates, often due to the interaction of multiple predicates and correlations in the data. In this paper, we propose a low-cost post-processing step that can take a plan produced by the optimizer, detect when it is likely to have made such a mistake, and take steps to fix it. Specifically, our solution is a sampling-based iterative procedure that requires almost no changes to the original query optimizer or query evaluation mechanism of the system. We show that this indeed imposes low overhead and catches cases where three widely used optimizers (PostgreSQL and two commercial systems) make large errors.

cs.DB↗

arXiv2015

Revisiting Differentially Private Regression: Lessons From Learning Theory and their Consequences

Private regression has received attention from both database and security communities. Recent work by Fredrikson et al. (USENIX Security 2014) analyzed the functional mechanism (Zhang et al. VLDB 2012) for training linear regression models over medical data. Unfortunately, they found that model accuracy is already unacceptable with differential privacy when $\varepsilon = 5$. We address this issue, presenting an explicit connection between differential privacy and stable learning theory through which a substantially better privacy/utility tradeoff can be obtained. Perhaps more importantly, our theory reveals that the most basic mechanism in differential privacy, output perturbation, can be used to obtain a better tradeoff for all convex-Lipschitz-bounded learning tasks. Since output perturbation is simple to implement, it means that our approach is potentially widely applicable in practice. We go on to apply it on the same medical data as used by Fredrikson et al. Encouragingly, we achieve accurate models even for $\varepsilon = 0.1$. In the last part of this paper, we study the impact of our improved differentially private mechanisms on model inversion attacks, a privacy attack introduced by Fredrikson et al. We observe that the improved tradeoff makes the resulting differentially private model more susceptible to inversion attacks. We analyze this phenomenon formally.

cs.CR↗