Chunbin Lin

Chunbin Lin

Auto-generated profile, based on publicly available information

Chunbin Lin

Amazon AWS

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 June 2023
Verified
Conference Paper
Open access
GSGaurav Saxena
Gaurav Saxena
MRMohammad Rahman
Mohammad Rahman
NCNaresh Chainani
Naresh Chainani
CLChunbin Lin
Chunbin Lin
GCGeorge Caragea
George Caragea
BNBalakrishnan (Murali) Narayanaswamy
Balakrishnan (Murali) Narayanaswamy
There has been a lot of excitement around using machine learning to improve the performance and usability of database systems. However, few of these techniques have actually been used in the critical path of customer-facing database services. In this paper, we describe Auto-WLM, a machine learning based automatic workload manager currently used in production in Amazon Redshift. Auto-WLM is an example of how machine learning can improve the performance of large data-warehouses in practice and at scale. Auto-WLM intelligently schedules workloads to maximize throughput and horizontally scales clusters in response to workload spikes. While traditional heuristic-based workload management requires a lot of manual tuning (e.g. of the concurrency level, memory allocated to queries etc.) for each specific workload, Auto-WLM does this tuning automatically and as a result is able to quickly adapt and react to workload changes and demand spikes. At its core, Auto-WLM uses locally-trained query performance models to predict the query execution time and memory needs for each query, and uses this to make intelligent scheduling decisions. Currently, Auto-WLM makes millions of decisions every day, and constantly optimizes the performance for each individual Amazon Redshift cluster. In this paper, we will describe the advantages and challenges of implementing and deploying Auto-WLM, as well as outline areas of research that may be of interest to those in the "ML for systems'' community with an eye for practicality.

1 January 2017
Verified
Conference Paper
Open access
CLChunbin Lin
Chunbin Lin
JWJianguo Wang
Jianguo Wang
JLJiaheng Lu
Jiaheng Lu
This paper studies the location-sensitive auto-completion problem. We propose an efficient algorithm SQA running on a native index combining both IR-tree and Trie index. The experiments on real-life datasets demonstrate that SQA outperforms baseline methods by one order of magnitude.

1 August 2020 in Information Sciences
Verified
Journal Article
Closed access
JWJin Wang
Jin Wang
CLChunbin Lin
Chunbin Lin
MLMingda Li
Mingda Li
CZCarlo Zaniolo
Carlo Zaniolo

26 October 2017 in The VLDB Journal
Verified
Journal Article
Open access
PDF file
CLChunbin Lin
Chunbin Lin
JLJiaheng Lu
Jiaheng Lu
ZWZhewei Wei
Zhewei Wei
JWJianguo Wang
Jianguo Wang
XXXiaokui Xiao
Xiaokui Xiao
Traditional top-k algorithms, e.g., TA and NRA, have been successfully applied in many areas such as information retrieval, data mining and databases. They are designed to discover k objects, e.g., top-k restaurants, with highest overall scores aggregated from different attributes, e.g., price and location. However, new emerging applications like query recommendation require providing the best combinations of attributes, instead of objects. The straightforward extension based on the existing top-k algorithms is prohibitively expensive to answer top-k combinations because they need to enumerate all the possible combinations, which is exponential to the number of attributes. In this article, we formalize a novel type of top-k query, called top-k, m, which aims to find top-k combinations of attributes based on the overall scores of the top-m objects within each combination, where m is the number of objects forming a combination. We propose a family of efficient top-k, m algorithms with different data access methods, i.e., sorted accesses and random accesses and different query certainties, i.e., exact query processing and approximate query processing. Theoretically, we prove that our algorithms are instance optimal and analyze the bound of the depth of accesses. We further develop optimizations for efficient query evaluation to reduce the computational and the memory costs and the number of accesses. We provide a case study on the real applications of top-k, m queries for an online biomedical search engine. Finally, we perform comprehensive experiments to demonstrate the scalability and efficiency of top-k, m algorithms on multiple real-life datasets.

1 April 2020
Verified
Conference Paper
Closed access
JWJin Wang
Jin Wang
CLChunbin Lin
Chunbin Lin
Location-based services have become ubiquitous in smart life, but typing queries in mobile devices is tedious and error-prone. Therefore, query autocompletion is needed to instantly provide users with query suggestions based on the incomplete user input. A recent trend is to support error-tolerant autocompletion, which could improve the usability by allowing a small number of errors between the query input and prefixes of strings in database. In addition, the query autocompletion should be location-aware for location-based services since it makes more sense to provide query suggestions for nearby objects. Unfortunately, existing query autocompletion algorithms cannot efficiently support both error-tolerate and location-aware features at the same time.In this paper, we propose a novel framework AutoEL to support error-tolerant location-aware query autocompletion. The error-tolerate feature is enabled by applying edit distance to evaluate the textual similarity between given query and the underlying data, while the location-aware feature is guaranteed by choosing the k-nearest neighbors. To improve the efficiency, we construct a hybrid data structure to jointly index spatial and textural information. We also propose several optimizations on data partition as well as search algorithm. Extensive experiments on real datasets demonstrate that AutoEL outperforms the baseline methods by up to an order of magnitude.

20 April 2016 in arXiv
Verified
Preprint
Open access
PDF file
RHRuining He
Ruining He
CLChunbin Lin
Chunbin Lin
JWJianguo Wang
Jianguo Wang
JMJulian J. McAuley
Julian J. McAuley
Building successful recommender systems requires uncovering the underlying dimensions that describe the properties of items as well as users' preferences toward them. In domains like clothing recommendation, explaining users' preferences requires modeling the visual appearance of the items in question. This makes recommendation especially challenging, due to both the complexity and subtlety of people's 'visual preferences,' as well as the scale and dimensionality of the data and features involved. Ultimately, a successful model should be capable of capturing considerable variance across different categories and styles, while still modeling the commonalities explained by `global' structures in order to combat the sparsity (e.g. cold-start), variability, and scale of real-world datasets. Here, we address these challenges by building such structures to model the visual dimensions across different product categories. With a novel hierarchical embedding architecture, our method accounts for both high-level (colorfulness, darkness, etc.) and subtle (e.g. casualness) visual characteristics simultaneously.