Neural Network Node Weight Update via Input-Output Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI systems face limitations due to rigid representations in Artificial Neural Networks (ANNs), leading to inaccuracies in predictions and a lack of rules for autonomous network formation, which hampers the generation of high-capacity functioning networks.
Innovation Solution
A data processing system is configured with a network of nodes, each with multiple inputs and weights, where updating units adjust weights based on input correlations during a learning mode. The system includes nodes grouped to either excite or inhibit other nodes, with probability values for weight updates and set points to regulate network activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If rigid representations are used in ANNs, then the network can focus on limited features for identification, but this leads to inaccuracy in predictions
Solution Approach 1:
The patent changes the representation parameters from rigid to widespread across all nodes. Each node contributes to all representations through weight updates based on input-output correlations, transforming the network from focusing on limited features to distributing representations widely, thereby improving prediction accuracy while maintaining operational capability
2Measurement precision
If dense coding networks are implemented, then more accurate predictions can be achieved, but a lack of rules for autonomous network formation makes it difficult to generate functioning networks with high capacity
Solution Approach 1:
The patent implements self-service through autonomous network formation. Nodes automatically adjust their weights based on input-output correlations without external intervention. The updating units enable nodes to self-organize and form functional networks with high capacity, eliminating the need for manual network design while achieving accurate predictions
Solution Approach 2:
The patent uses feedback mechanisms where output signals are fed back to update input weights through correlation calculations. This feedback loop enables autonomous learning and network formation, allowing the system to generate functioning high-capacity networks automatically while maintaining prediction accuracy
3Adaptability or versatility
If nodes are configured to excite or inhibit other nodes, then network capacity and representation distribution are improved, but the system complexity increases
Solution Approach 1:
The patent segments nodes into functional groups (excitatory and inhibitory) with distinct roles. This segmentation simplifies the complex interactions by organizing nodes into manageable categories, each following specific connection rules, thereby reducing overall system complexity while maintaining high network capacity and adaptability
Data Source
AI summary
The disclosure relates to a data processing system (100), configured to have one or more system input(s) (110a, 110b, . . . , 110z) comprising data to be processed and a system output (120), comprising: a network, NW, (130) comprising a plurality of nodes (130a, 130b, . . . , 130x), each node configured to have a plurality of inputs (132a, 132b, . . . , 132y), each node (130a, 130b, . . . , 130x) comprising a weight (Wa, . . . , Wy) for each input (132a, 132b, . . . , 132y), and each node configured to produce an output (134a, 134b, . . . , 134x); and one or more updating units (150) configured to update the weights (Wa, . . . , Wy) of each node based on correlation of each respective input (132a, . . . , 132c) of the node (130a) with the corresponding output (134a) during a learning mode; one or more processing units (140x) configured to receive a processing unit input and configured to produce a processing unit output by changing the sign of the received processing unit input; and wherein the system output (120) comprises the outputs (134a, 134b, . . . , 134x) of each node (130a, 130b, . . . , 130x), wherein nodes (130a, 130b) of a first group (160) of the plurality of nodes are configured to excite one or more other nodes ( . . . , 130x) of the plurality of nodes (130a, 130b, . . . , 130x) by providing the output (134a, 134b) of each of the nodes (130a, 130b) of the first group (160) of nodes as input (132d, . . . , 132y) to the one or more other nodes ( . . . , 130x), wherein nodes (130x) of a second group (162) of the plurality of nodes are configured to inhibit one or more other nodes (130a, 130b, . . . ) of the plurality of nodes (130a, 130b, . . . , 130x) by providing the output (134x) of each of the nodes (130x) of the second group (162) as a processing unit input to a respective processing unit (140x), each respective processing unit (140x) being configured to provide the processing unit output as input (132b, 132e, . . . ) to the one or more other nodes (130a, 130b, . . . ) and wherein each node of the plurality of nodes (130a, 130b, . . . , 130x) belongs to one of the first and second groups (160, 162) of nodes. The disclosure further relates to a method, and a computer program product.