Metadata-Actuated Graph Execution in AI Processing Cores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine intelligence systems, particularly deep learning ANNs, are computationally and energy intensive due to large data structures and complex recursive relationships between layers, making them difficult to parallelize and resource-hungry.
Innovation Solution
Implement a processing core that conditionally executes a directed graph using metadata stored in association with data tiles, allowing for efficient execution by selectively performing arithmetic logic operations based on metadata, reducing computational complexity and memory bandwidth requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard execution of directed graph is performed, then computational accuracy is maintained, but computational resources and energy consumption increase significantly
Solution Approach 1:
The patent applies partial action by conditionally executing only necessary arithmetic logic operations based on metadata evaluation. Instead of performing all operations in the directed graph unconditionally, the system evaluates metadata first and executes only those operations that are required, thereby reducing energy consumption while maintaining computational accuracy for essential calculations.
Solution Approach 2:
The patent implements preliminary action by evaluating metadata before executing arithmetic logic operations. This preliminary evaluation determines which operations are necessary, allowing the system to prepare and execute only the required computations, thus reducing overall energy consumption while preserving accuracy where needed.
2Reliability
If standard execution of directed graph is performed, then complete computational results are obtained, but memory bandwidth requirements increase
Solution Approach 1:
The system performs partial action by loading and processing only the necessary data tiles from memory based on metadata evaluation. By determining which operations are required beforehand, the system retrieves only the corresponding data tiles from memory, significantly reducing memory bandwidth requirements while ensuring computational completeness for the necessary operations.
Solution Approach 2:
The patent extracts only the necessary data tiles from memory by evaluating metadata first. This extraction principle allows the system to identify and load only the specific data elements required for the conditional operations, rather than loading all possible data, thereby reducing memory bandwidth consumption while maintaining computational completeness.
3Productivity
If conditional execution with metadata is implemented, then computational resources are reduced, but system complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing metadata that indicates which arithmetic logic operations are necessary. This preliminary preparation allows the execution engine to make simple conditional decisions based on the metadata without complex real-time analysis, thereby improving computational efficiency while managing system complexity through pre-processing.
Solution Approach 2:
The patent introduces metadata as an intermediary between the directed graph definition and the execution engine. This intermediary layer provides the execution engine with pre-analyzed information about necessary operations, simplifying the execution decision-making process and improving efficiency without requiring complex real-time computational analysis.
4Measurement precision
If deep learning ANNs with recursive relationships are used, then model accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by conditionally executing operations in deep learning ANNs based on metadata evaluation. For recursive relationships and complex layer interactions, the system determines which specific operations are necessary and executes only those, reducing computational complexity while preserving the accuracy benefits of deep learning models with multiple layers and recursive elements.
Data Source
AI summary
A processing core and associated methods for the efficient execution of a directed graph are disclosed. A disclosed processing core includes a memory and a first data tile stored in the memory. The first data tile includes a first set of data elements and metadata stored in association with the first set of data elements. The processing core also includes a second data tile stored in the memory. The second data tile includes a second set of data elements. The processing core also includes an arithmetic logic unit configured to conduct an arithmetic logic operation using data from the first set of data elements and the second set of data elements. The processing core also includes a control unit configured to evaluate the metadata and control the arithmetic logic unit to conditionally execute the arithmetic logic operation based on the evaluation of the metadata.