Database Function Creation for Deep Learning Model Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of training machine learning and deep learning models presents a barrier for both technical and non-technical users, making it difficult to utilize these methods in computing due to computational expense and accessibility issues.
Innovation Solution
A system that enables the creation and execution of database functions using deep learning models, allowing users to interact with models as black boxes, with pre-trained models packaged in a way that can be understood and executed by non-technical users, utilizing a processor and interface to define inputs, outputs, and cluster processing information, and incorporating transfer learning for related tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models are used for data-driven predictions, then prediction accuracy and capability are improved, but computational expense and technical complexity increase
Solution Approach 1:
The patent introduces database functions as an intermediary layer between deep learning models and users. These functions package complex model inference operations into simple, executable database queries that can be called without understanding the underlying neural network complexity, thus maintaining high prediction accuracy while reducing technical complexity for end users
Solution Approach 2:
The patent creates simplified copies or wrappers of deep learning model interfaces through database functions. Instead of requiring users to interact with complex model APIs, the system provides simplified database function calls that replicate the essential prediction capability while abstracting away the computational and technical complexity
2Measurement precision
If deep learning models are trained from scratch, then model performance is improved, but training time and computational resources increase
Solution Approach 1:
The patent implements pre-trained deep learning models that have already been trained on large datasets before deployment. Users can directly apply these pre-trained models to their data through database functions without undergoing the lengthy training process, thus achieving good model performance while avoiding the time loss associated with training from scratch
Solution Approach 2:
The patent allows users to modify parameters such as input data characteristics and task-specific configurations of pre-trained models through database function parameters. This enables adaptation of general-purpose pre-trained models to specific applications without requiring retraining, maintaining performance while minimizing time investment
3Adaptability or versatility
If complex machine learning operations are exposed to users, then functionality and capability are improved, but ease of operation deteriorates
Solution Approach 1:
The patent creates universal database functions that can handle multiple deep learning operations (classification, regression, clustering, etc.) through a single unified interface. This allows diverse functionality to be accessed through consistent, simple database query syntax, improving ease of operation while maintaining comprehensive functionality
Solution Approach 2:
The database function serves as a mediator that translates simple user queries into complex deep learning operations. Users interact with familiar database query language rather than complex ML APIs, making the system accessible to non-technical users while still providing advanced functionality through the intermediary translation layer
4Productivity
If computational resources are allocated for model execution, then processing capability is improved, but resource consumption increases
Solution Approach 1:
The patent merges multiple model inference operations into single database function calls that can process batches of data together. By combining multiple small operations into larger batch operations, the system improves overall processing capability while reducing the relative overhead and resource consumption per data point
Data Source
AI summary
A function creation method is disclosed. The method comprises defining one or more database function inputs, defining cluster processing information, defining a deep learning model, and defining one or more database function outputs. A database function is created based at least in part on the one or more database function inputs, the cluster set-up information, the deep learning model, and the one or more database function outputs. In some embodiments, the database function enables a non-technical user to utilize deep learning models.