Neural Network Sensor Fusion Using Sparse Stochastic Neurons
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks with stochastic neurons for sensor signal fusion require high energy and memory resources, making them less suitable for mobile devices and robots where energy efficiency is crucial.
Innovation Solution
The method reduces the number of computing operations by setting small preciseness values to zero, using sparse matrix operations and hardware acceleration, and employing machine learning for weight optimization, thereby improving energy efficiency and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If neural networks with stochastic neurons are used for sensor signal fusion, then fusion performance is improved, but energy consumption and memory requirements increase
Solution Approach 1:
The patent changes the parameter of precision values by setting them to zero when below a threshold, transforming the neural network operations from dense to sparse. This parameter change reduces the number of computing operations significantly, thereby lowering energy consumption while maintaining fusion performance through the stochastic neuron approach that handles uncertainty effectively.
Solution Approach 2:
The patent applies partial action by selectively processing only those precision values that exceed the threshold, rather than processing all precision values uniformly. This selective processing reduces the computational burden and energy consumption while maintaining the essential fusion capabilities where needed.
2Reliability
If neural networks with stochastic neurons are used for sensor signal fusion, then fusion performance is improved, but memory space requirements increase
Solution Approach 1:
The patent changes the storage parameter by setting precision values to zero when below threshold, reducing the amount of data that needs to be stored and processed. This parameter change transforms memory requirements from dense storage to sparse storage, significantly reducing memory space requirements while maintaining fusion performance.
Solution Approach 2:
The patent extracts and removes unnecessary precision values from the computation by setting them to zero, thereby reducing the memory footprint. This extraction of insignificant data maintains the essential information needed for fusion while reducing overall memory requirements.
3Use of energy by moving object
If the number of computing operations is reduced, then energy consumption decreases, but computational accuracy may be compromised
Solution Approach 1:
The patent changes the parameter of precision values to zero for values below threshold, which reduces computing operations and energy consumption. This parameter change is justified because the stochastic neuron approach can handle uncertainty and maintain computational accuracy through probabilistic reasoning even when some precision values are set to zero.
Solution Approach 2:
The patent treats low precision values as disposable by setting them to zero, rather than expending computational resources on them. This approach is valid because the stochastic neuron model can compensate for the loss of these values through its probabilistic nature, maintaining overall computational accuracy while reducing energy consumption.
Data Source
AI summary
A computer-implemented method for the fusion of a plurality of sensor signals using a neural network, a sensor signal including at least one first value that characterizes an expected value of a physical variable and including a second value that characterizes a scatter of the physical variable. In addition the neural network ascertains, based on the plurality of sensor signals, an output that characterizes a fusion of the plurality of sensor signals. The output is a function of a first intermediate output of the neural network. The first intermediate output is ascertained by at least one first neuron and including an ascertained first value that characterizes an expected value of a fusion of the plurality of sensor values, and including an ascertained second value that characterizes a scatter of the fusion, the ascertained second value of the first intermediate output being set to zero if a specifiable condition is fulfilled.