3D Spike Generation for Low-Power Time-Series Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Transforming time-series data into a suitable representation for machine learning algorithms is computationally intensive, requiring large amounts of power and memory, making it unsuitable for low-power computing hardware.
Innovation Solution
A method is developed to generate three-dimensional spikes by mapping time-series data to a 2D grid, performing mathematical operations, and applying spatial filtering, which reduces computational requirements while maintaining discriminative capability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time-series data is transformed using conventional computational methods, then the data representation suitable for machine learning is achieved, but the power consumption and memory requirements become excessively high
Solution Approach 1:
The time-series data is divided into multiple segments that are mapped to different positions on a 2D grid. Each segment is processed independently to generate spike trains, which are then combined to form the final 3D spike representation. This segmentation allows the computational task to be distributed across multiple simple parallel operations rather than requiring intensive sequential processing.
Solution Approach 2:
The patent transforms 1D time-series data into a 3D spike representation by introducing spatial dimensions (2D grid positions) and temporal dimension (spike trains). This dimensional transformation enables the data to be processed using simple spatial mapping and combination operations instead of complex temporal analysis, significantly reducing computational requirements while preserving discriminative information.
2Measurement precision
If time-series data is transformed using conventional computational methods, then the data representation suitable for machine learning is achieved, but the memory requirements become excessively high
Solution Approach 1:
By segmenting the time-series data and mapping segments to different 2D grid positions, the patent processes data in distributed fashion. Each grid position maintains only local segment information and its corresponding spike train, rather than requiring the entire dataset to be held in memory simultaneously. This segmentation dramatically reduces peak memory requirements.
Solution Approach 2:
The patent extracts only the essential features from each time-series segment by converting them into spike trains based on threshold crossings. This extraction process discards redundant information while retaining discriminative patterns, enabling efficient storage and processing with minimal memory footprint.
3Use of energy by moving object
If simple mapping operations are used to reduce computational requirements, then power consumption is reduced, but the discriminative capability of the data representation may be compromised
Solution Approach 1:
The patent applies different processing operations to different regions of the 2D grid based on local data characteristics. Each grid position performs spike train generation based on its local segment, and spatial filtering is applied locally to enhance discriminative features specific to that region. This local quality approach preserves discriminative capability while using simple operations throughout.
Solution Approach 2:
The patent performs preliminary spatial filtering on the 2D grid before generating the final 3D spike representation. This preliminary action enhances discriminative features in the spatial domain, ensuring that the subsequent combination of spike trains preserves important patterns. By preparing the data in advance with simple filtering operations, the system maintains high discriminative capability without requiring complex processing during the final spike generation stage.
Data Source
AI summary
Aspects described herein include a method of generating three-dimensional (3D) spikes. The method comprises receiving a signal comprising time-series data and generating a first two-dimensional (2D) grid. Generating the first 2D grid comprises mapping segments of the time-series data to respective positions of the first 2D grid, and generating, for each position, a spike train corresponding to the respective mapped segment. The method further comprises generating a second 2D grid including performing, for each position, a mathematical operation on the spike train of the corresponding position of the first 2D grid. The method further comprises generating a third 2D grid including performing spatial filtering on the positions of the second 2D grid. The method further comprises generating a 3D grid based on a combination of the first 2D grid, the second 2D grid, and the third 2D grid. The 3D grid comprises one or more 3D spikes.