Parameterized Event Mapping for Neural Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems face challenges in achieving versatile mappings of data processing system elements with modest storage requirements, particularly in handling multidimensional data like image data, where fixed mappings are insufficient and lookup tables consume substantial storage space.
Innovation Solution
A neural processing system with a mapping facility that computes system element identifiers as a function of event message source data, allowing for one-to-one, many-to-one, or one-to-many mappings, and includes a region selection facility to dynamically determine and map regions of interest, reducing the need for extensive storage by using parameterized functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a lookup table is used to map each source address to a respective destination, then versatile mappings can be achieved, but substantial storage space is required
Solution Approach 1:
The patent applies parameter changes by using a parameterized mapping function that takes input parameters (source address, mapping type, destination constraints) and computes the destination address dynamically. This replaces the static lookup table with a computational function that adapts to different mapping scenarios, achieving versatile mappings without requiring storage for all possible mapping configurations.
Solution Approach 2:
The patent substitutes the mechanical/storage-based lookup table system with a computational/mathematical mapping function system. Instead of storing pre-computed mapping tables in memory, the system uses algorithms that compute mappings on-demand based on input parameters, thereby eliminating the need for substantial storage space while maintaining mapping versatility.
2Ease of manufacture
If a fixed mapping is used to provide data from the data array, then the mapping is simple, but it cannot adapt to different circumstances
Solution Approach 1:
The patent applies dynamics by transitioning from a fixed, static mapping to a dynamic, adaptive mapping function. The mapping function can change its behavior based on input parameters such as source address, mapping type (one-to-one, many-to-one, one-to-many), and destination constraints. This allows the system to adapt to different circumstances while maintaining a relatively simple computational structure.
Solution Approach 2:
The patent implements universality through a single parameterized mapping function that can handle multiple mapping scenarios (one-to-one, many-to-one, one-to-many mappings) and adapt to different data array configurations. This universal function replaces multiple fixed mapping schemes, providing both simplicity and adaptability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A data processing system (1) is provided that comprises a plurality (130) of data processing system elements having a respective storage location (10_j) in said processing system for storing a state and having a respective system element identifier (NID) that identifies the data processing system element within said data processing system. The data processing system further comprises a message exchange facility (21, 110, 120) shared by said plurality of data processing system elements, said plurality of data processing system elements being configured to transmit event messages via the message exchange facility and to receive event messages transmitted via the message exchange facility. The data processing system further comprises an update facility (134) to update the states of respective data processing system elements in accordance with event messages directed to said respective data processing system elements and to enable generation of event messages to be transmitted by respective data processing system elements in accordance with their state. The data processing system further comprises a mapping facility (60) being configured to selectively compute at least one system element identifier (NID) as a mapping function of event message source data, in accordance with one or more parameters specified for the mapping function, the event message being directed via the message exchange facility to the data processing system element specified by the computed system element identifier (NID).