RDBMS Hyperparameter Tuning Across MPP Search Partitions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for tuning control parameters in massively parallel database systems are inefficient and costly due to the overhead of moving data in and out of the system, and they fail to leverage the parallelism of the MPP architecture effectively.

Innovation Solution

Implementing hyperparameter tuning as an operation within the RDBMS that partitions the search space across compute units in the MPP system, allowing concurrent execution of function evaluations on different hyperparameter combinations, thereby optimizing the process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional partition-the-data paradigm is used for hyperparameter tuning, then data processing can be distributed across compute units, but data movement overhead increases and processing time increases

Engineering Contradiction:

Improvehyperparameter tuning efficiencyVSAvoiddata movement overhead

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The invention segments the hyperparameter search space into multiple partitions, with each compute unit responsible for evaluating a specific partition. This allows parallel evaluation of different hyperparameter combinations without requiring data movement between compute units, resolving the contradiction by maintaining productivity through segmentation while eliminating time loss from data transfer

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The invention introduces a coordinate system as an intermediary structure that maps hyperparameter combinations to compute units. This intermediary enables efficient distribution of hyperparameter evaluation tasks across compute units without moving data, as the coordinate system acts as a virtual mediator that routes computations rather than physical data transfers

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If external analytics libraries are used for hyperparameter tuning, then tuning capabilities are provided, but data must be moved in and out of the database system increasing cost and time

Engineering Contradiction:

Improvehyperparameter tuning capabilityVSAvoiddata movement cost

Core Design Contradiction:

Adaptability or versatilityVSLoss of energy

Solution Approach 1:

The invention makes the database system itself multi-functional by enabling it to perform both data storage and hyperparameter tuning operations internally. The database system universally handles both traditional data management tasks and machine learning model tuning, eliminating the need to export data to external libraries and import results back, thus resolving the contradiction by maintaining adaptability while reducing energy loss from data movement

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The invention enables the database system to self-service hyperparameter tuning operations by providing built-in functionality to evaluate multiple hyperparameter combinations directly on stored data. The system serves its own data without requiring external tools, eliminating costly data export/import operations while maintaining versatile tuning capabilities through integrated evaluation functions

Inventive Principle:

Principle #25Self-service

3Reliability

If hyperparameter tuning is implemented as an RDBMS operation, then data governance is maintained and parallelism is leveraged, but system complexity increases

Engineering Contradiction:

Improvedata governance complianceVSAvoidRDBMS operation complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The invention changes the parameters of the RDBMS by extending it with hyperparameter tuning capabilities that operate on existing data structures. Rather than fundamentally redesigning the system, it modifies operational parameters by adding evaluation functions that work with the existing data model, thus maintaining data governance while managing complexity through parameter-based extensions rather than structural overhauls

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12524681B2Performing hyperparameter tuning of models in a massively parallel database system

Publication Date: 2026.01.13 TERADATA US INC

AI summary

Hyperparameter tuning for a machine learning model is performed in a massively parallel database system. A computer system comprised of a plurality of compute units executes a relational database management system (RDBMS), wherein the RDBMS manages a relational database comprised of one or more tables storing data. One or more of the compute units perform the hyperparameter tuning for the machine learning model, wherein the hyperparameters are control parameters used in construction of the model, and the tuning of the hyperparameters is implemented as an operation in the RDBMS that accepts training and scoring data for the model, constructs the model using the hyperparameters and the training data, and generates goodness metrics for the model using the scoring data.