Wavelet Tensor Conversion for Neural Network Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for risk detection in financial markets, such as rule-based determinations, decision trees, and artificial neural networks, suffer from high error rates, resource inefficiencies, and increased complexity, particularly when handling discrete data stored in relational or graph databases.
Innovation Solution
The use of wavelet databases and deep field networks that convert transactions into tensors using exponential smoothing averages, perform weighted summations, and apply models to detect anomalies and authorize transactions efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If artificial neural networks are used to detect anomalies, then measurement precision improves, but device complexity and resource consumption increase exponentially with the number of hidden neuron layers
Solution Approach 1:
The patent segments the anomaly detection task into multiple specialized neural networks, each trained on a specific group of wavelet coefficients. Instead of using one large complex network, the system divides the coefficient groups and creates specialized networks for each group, then integrates their outputs. This segmentation reduces the complexity of individual networks while maintaining overall detection accuracy through the coordinated work of multiple specialized networks.
2Measurement precision
If the number of hidden neuron layers increases to improve pattern recognition, then measurement precision improves, but resource consumption increases exponentially
Solution Approach 1:
The patent segments the computational workload by dividing wavelet coefficients into multiple groups and training separate neural networks on each group. This segmentation allows each individual network to have fewer layers and lower computational requirements, while the collective ensemble of networks achieves high pattern recognition accuracy. The resource consumption is distributed across multiple simpler networks rather than concentrated in one deep network.
Solution Approach 2:
The patent applies partial action by training multiple neural networks on different subsets of coefficient groups rather than training one network on all data. Each network processes only a portion of the total coefficient groups, reducing the computational burden per network. The integration of multiple partial results achieves comprehensive anomaly detection without requiring any single network to process the entire dataset with excessive complexity.
3Ease of operation
If discrete data is stored in relational databases, then ease of operation improves, but resource consumption increases due to costly searches and concatenations
Solution Approach 1:
The patent introduces wavelet transforms as an intermediary layer between the raw transaction data and the neural network processing. By transforming transaction data into wavelet coefficients before storage and analysis, the system creates a more efficient intermediate representation that reduces the computational cost of subsequent operations. The wavelet coefficients serve as an intermediary form that is more suitable for the pattern recognition tasks than raw discrete data, reducing resource consumption during retrieval and analysis.
4Adaptability or versatility
If multiple extant methods (rules, trees, networks) are implemented separately, then adaptability improves, but device complexity and processing efficiency worsen due to handling disparate results
Solution Approach 1:
The patent merges multiple detection approaches by integrating multiple specialized neural networks into a unified deep field network architecture. Each specialized network processes different coefficient groups, and their results are combined through the deep field network framework. This merging maintains the adaptability benefits of having multiple specialized components while reducing the overall system complexity by providing a unified integration framework that automatically combines results, eliminating the need for manual handling of disparate outputs from separate rule-based, decision tree, and neural network systems.
Data Source
AI summary
The present disclosure relates to systems and methods for detecting and identifying anomalies within a discrete wavelet database. In one implementation, the system may include one or more memories storing instructions and one or more processors configured to execute the instructions. The instructions may include instructions to receive a new wavelet, convert the net transaction to a wavelet, convert the wavelet to a tensor using an exponential smoothing average, calculate a difference field between the tensor and a field having one or more previous transactions represented as tensors, perform a weighted summation of the difference field to produce a difference vector, apply one or more models to the difference vector to determine a likelihood of the new wavelet representing an anomaly, and add the new wavelet to the field when the likelihood is below a threshold.