Reconfigurable Bayesian Inference Architecture Using Unary Fixed-Point Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Bayesian online change point detection algorithms face high computational requirements due to the need for high-precision floating point computations, making them inefficient for real-time processing in dynamic environments.
Innovation Solution
A reconfigurable computing architecture using unary fixed-point representation and pulse density or random pulse density modulation for Bayesian Online Change Point Detection, incorporating online machine learning for continuous update of processing parameters, allowing for efficient computation with reduced power dissipation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-precision floating point computations are used for Bayesian online change point detection, then measurement precision is improved, but use of energy and computational complexity increase
Solution Approach 1:
The patent transforms the numerical representation parameters from high-precision floating point to low-precision fixed point arithmetic. This parameter change maintains sufficient detection accuracy while dramatically reducing computational complexity and power consumption, enabling deployment on resource-constrained devices.
Solution Approach 2:
The patent replaces complex floating point computational mechanisms with simpler fixed point arithmetic operations. This substitution eliminates the need for complex division and multiplication units, reducing hardware complexity and energy consumption while maintaining functional equivalence for change point detection.
2Measurement precision
If floating point computations are used for Bayesian inference, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent changes the numerical precision parameter from floating point to fixed point representation. This simplifies the computational architecture by eliminating the need for complex floating point units, while fixed point arithmetic can be implemented with simple adders and shifters, reducing device complexity.
Solution Approach 2:
The patent substitutes complex floating point computational mechanisms with simpler fixed point arithmetic operations. This replacement reduces the hardware complexity of processing units while maintaining the mathematical correctness and accuracy of Bayesian inference algorithms.
3Productivity
If real-time processing is implemented for online change point detection, then productivity is improved, but use of energy increases
Solution Approach 1:
The patent adopts fixed point arithmetic parameters instead of floating point, which enables faster computation with simpler hardware. This parameter change allows real-time processing to be achieved with lower power consumption, as fixed point operations require fewer computational resources and can be executed more efficiently.
Solution Approach 2:
The patent segments the computational process into discrete fixed point operations that can be executed in parallel or pipelined. This segmentation enables real-time processing by breaking down complex Bayesian inference into manageable steps that can be completed within strict timing constraints while minimizing energy usage.
Data Source
AI summary
A reconfigurable computing architecture for Bayesian Online ChangePoint Detection (BOCPD) is provided. In an exemplary embodiment, the architecture may be employed for use in video processing, and more specifically to computing whether a pixel in a video sequence belongs to the background or to an object (foreground). Each pixel may be processed with only information from its intensity and its time history. The computing architecture employs unary fixed point representation for numbers in time, using pulse density or random pulse density modulation—i.e., a stream of zeros and ones, where the mean of that stream represents the encoded value.