Neural Network Arrangement With Programmable Threshold Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The selection and execution of machine learning algorithms are challenging due to the numerous alternatives available, each with varying suitability, and existing algorithms are tightly coupled to specific tasks, making it difficult to effectively address complex data analysis requirements without programmer intervention.
Innovation Solution
A computer-implemented method using a network of processing nodes with adjustable weights and threshold functions, allowing iterative training through backpropagation and evolutionary algorithms to model target functions, with a bypass function for selective input mapping, enabling programmable node emphasis and dynamic network adaptation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional machine learning algorithms are used, then specific tasks can be addressed, but the algorithms are tightly coupled to tasks and lack flexibility for other applications
Solution Approach 1:
The patent implements a universal neural network architecture where processing nodes can be dynamically configured through programmable threshold functions and bypass mechanisms. This allows the same physical hardware to perform multiple different machine learning tasks by adjusting weights and thresholds, rather than requiring separate specialized algorithms for each task.
Solution Approach 2:
The patent introduces dynamic programmability into neural network nodes through threshold functions that can be adjusted during operation and bypass functions that selectively activate different computational paths. This dynamic configuration enables the network to adapt its structure and behavior based on the specific task at hand, resolving the rigidity of traditional fixed algorithms.
2Measurement precision
If machine learning algorithms are tailored specifically for tasks, then task performance is optimized, but programmer intervention is required for selection and configuration
Solution Approach 1:
The patent enables the neural network to automatically select and configure appropriate computational patterns through its programmable threshold functions and bypass mechanisms. The system can autonomously determine which nodes to activate and how to route information based on the input data characteristics, eliminating the need for manual algorithm selection and configuration by programmers.
Solution Approach 2:
The patent pre-configures the neural network with multiple possible computational paths and threshold functions that can be activated based on the specific task requirements. This preliminary preparation of multiple configurations allows the system to quickly adapt to different tasks without requiring real-time programming intervention, as the appropriate configuration is already prepared and can be selected automatically.
3Measurement precision
If complex data sets are analyzed using traditional methods, then analysis can be performed, but the process is slow and less accurate compared to machine learning
Solution Approach 1:
The patent divides the neural network into multiple processing nodes arranged in layers, where each node performs a specific computational function. This segmentation allows parallel processing of complex data sets, significantly improving analysis speed compared to sequential traditional methods while maintaining high accuracy through the distributed nature of the computation across many nodes.
Solution Approach 2:
The patent implements continuous training and adjustment of the neural network through iterative optimization processes. The network continuously learns from data and refines its weights and thresholds, maintaining high accuracy while the parallel architecture ensures continuous high-speed processing without interruption, resolving the trade-off between accuracy and speed.
Data Source
AI summary
A computer implemented method of a machine learning algorithm modelling a target function mapping inputs in an input domain to outputs in an output range, the machine learning algorithm including an array of processing nodes arranged in a network of layers of nodes including an input layer for receiving an input value, an output layer for providing an output value, and one or more intermediate layers between the input and output layers, each node in the processing set being outside the input layer receiving input from at least some adjacent nodes logically closer to the input layer via weighted connections between nodes, and each node being outside the output layer generating output to at least some adjacent nodes logically closer to the output layer via weighted connections between nodes, wherein each node includes: an adjustable weight for application to each input to the node, the adjustment weight being responsive to a threshold function applied to a value of the node input; a combination function for combining outputs of the threshold function; and a node bypass function for selectively mapping one or more of the inputs to the node to the output of the node, the method comprising iteratively training the machine learning algorithm to model the target function by adjustment, at each iteration, of at least weights of connections between at least a subset of the nodes, such that the nodes of the network are programmable during operation of the algorithm by adjustment of the threshold function and the bypass function so as to selectively emphasise subsets of nodes in the network.