2019
DOI: 10.14778/3317315.3317323
|View full text |Cite
|
Sign up to set email alerts
|

Declarative recursive computation on an RDBMS

Abstract: A number of popular systems, most notably Google's TensorFlow, have been implemented from the ground up to support machine learning tasks. We consider how to make a very small set of changes to a modern relational database management system (RDBMS) to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. I… Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
7
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
4
3
1

Relationship

0
8

Authors

Journals

citations
Cited by 33 publications
(7 citation statements)
references
References 32 publications
0
7
0
Order By: Relevance
“…We note that [29] has propose the use of provenance for incremental model updates for linear and logistic regression models. Another intriguing direction towards this goal is to leverage the line of works on machine learning algorithm representation using relational algebra [18,22], for which provenance models exists.…”
Section: Challenges and Opportunitiesmentioning
confidence: 99%
“…We note that [29] has propose the use of provenance for incremental model updates for linear and logistic regression models. Another intriguing direction towards this goal is to leverage the line of works on machine learning algorithm representation using relational algebra [18,22], for which provenance models exists.…”
Section: Challenges and Opportunitiesmentioning
confidence: 99%
“…Extensions of SQL for matrix manipulations are reported in [27]. Most relevant is [23] in which a recursion mechanism is added to SQL which resembles for-loops. The expressive power of this extension is unknown, however.…”
Section: Contribution and Outlinementioning
confidence: 99%
“…Linear algebra-based algorithms have become a key component in data analytic workflows. As such, there is a growing interest in the database community to integrate linear algebra functionalities into relational database management systems [5,23,[25][26][27]. In particular, from a query language perspective, several proposals have recently been put forward to unify relational algebra and linear algebra.…”
Section: Introductionmentioning
confidence: 99%
“…A significant limitation of SQL is the lack of iteration constructs. Although loops can be expressed via recursive queries, RDBMS support for recursion is typically limited to fixed-points over sets [89,90]. Iterative processes until a certain condition is satisfied, which typically is not supported by SQL recursion, are very common in machine learning though.…”
Section: On Db and ML Foundationsmentioning
confidence: 99%
“…Finally, recent work [107], [65] proposes the extension of SQL with matrices/vectors and a set of linear algebra operators. [90] combines this approach with optimizations on executing recursion and large query plans on an RDBMS, which can make it suitable for distributed machine learning. Sql4ml does not assume any changes to the relational database system, nor to the ML framework, enabling portability.…”
Section: Extending Sql With Linear Algebramentioning
confidence: 99%