Hypervector Sequence Encoding With In-Memory Shift and Rotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current high-dimensional computing models face limitations in energy efficiency and classification accuracy due to inefficient encoding processes, particularly in von Neumann architectures, which are memory-intensive and struggle with tasks requiring thousands to millions of memory entries.
Innovation Solution
A device and method for encoding an ordered group of symbols using a seed hypervector with circular shift and rotation operations, optimizing the encoding process to reduce energy consumption and improve area efficiency, enabling the generation of holographic hypervectors that represent sequences of symbols effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If von Neumann architecture is used for high-dimensional computing, then memory capacity can be increased, but energy consumption increases and computation efficiency decreases
Solution Approach 1:
The patent merges storage and computation functions into a single unified structure by implementing hypervector encoding directly within the memory array. The encoding unit performs circular shift and rotation operations on hypervectors stored in the memory array without requiring data to be moved to separate processing units, thereby eliminating the memory wall and reducing energy consumption associated with data transfer between storage and computation units.
Solution Approach 2:
The patent replaces the traditional von Neumann architecture's separate storage and computation units with an in-memory computing approach. The encoding operations (circular shifts and rotations) are performed directly within the memory array using bitwise operations, substituting the mechanical data movement and processing paradigm with an in-situ computational paradigm that operates on stored hypervectors.
2Ease of manufacture
If traditional encoding methods are used for ordered groups of symbols, then implementation is simpler, but area efficiency decreases and separate storage elements are required
Solution Approach 1:
The patent combines the encoding logic and storage functions into a single integrated unit. The encoding unit performs sequential processing of symbols using circular shift and rotation operations on hypervectors that are stored within the same memory array, eliminating the need for separate storage elements and reducing overall device area while maintaining implementation feasibility through systematic encoding procedures.
3Ease of operation
If channel-bound basis hypervectors are used, then encoding can be performed, but device complexity increases
Solution Approach 1:
The patent extracts the essential encoding operations (circular shift and rotation) from complex channel-bound basis hypervector systems and implements them as standalone encoding unit operations. This extraction simplifies the device architecture by removing the need for complex channel-bound basis hypervector management while retaining the core encoding functionality through sequential symbol processing and hypervector transformations.
Data Source
AI summary
The present disclosure relates to a method for representing an ordered group of symbols with a hypervector. The method comprises sequentially applying on at least part of the input hypervector associated with a current symbol a predefined number of circular shift operations associated with the current symbol, resulting in a shifted hypervector. A rotate operation may be applied on the shifted hypervector, resulting in an output hypervector. If the current symbol is not the last symbol of the ordered group of symbols the output hypervector may be provided as the input hypervector associated with a subsequent symbol of the current symbol; otherwise, the output hypervector of the last symbol of the ordered group of symbols may be provided as a hypervector that represents the ordered group of symbols.