Multi-label Shallow Neural Network for Tabular Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for tabular data are often restricted to a single label, requiring multiple models for multiple labels and are not optimized for efficiency with tabular data.
Innovation Solution
A multi-label shallow neural network is developed for tabular data, which normalizes the data and uses a shallow neural network with an input layer, a hidden layer, and an output layer to predict multiple classifications efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple separate machine learning models are used for multiple labels, then each label can be predicted accurately, but the device complexity and computational resources increase
Solution Approach 1:
The patent combines multiple single-label classification models into a single multi-label neural network model. The model includes an input layer that receives tabular data, one or more hidden layers for feature extraction, and an output layer with multiple output nodes corresponding to multiple labels. This merging approach maintains prediction accuracy for multiple labels while reducing device complexity by eliminating the need for multiple separate models.
Solution Approach 2:
The neural network model is designed to perform multiple classification functions simultaneously through a single unified structure. The output layer contains multiple output nodes, each responsible for predicting a different label, allowing the model to handle multiple classification tasks in one pass through the network, thereby improving efficiency and reducing computational overhead.
2Measurement precision
If deep neural networks with multiple hidden layers are used, then feature extraction capability improves, but processing time and computational resources increase
Solution Approach 1:
The patent employs a shallow neural network architecture with one or more hidden layers rather than deep networks with many layers. This partial approach to depth provides sufficient feature extraction capability for tabular data while avoiding the excessive computational overhead and processing time associated with deeper networks. The model achieves adequate feature extraction with fewer layers, optimizing the trade-off between capability and efficiency.
3Device complexity
If traditional machine learning models are used for tabular data, then simplicity is maintained, but multi-label prediction capability is lost
Solution Approach 1:
The patent modifies the output layer parameters of a neural network to enable multi-label prediction. The output layer includes multiple output nodes, each corresponding to a different label, with sigmoid activation functions applied independently to each node. This parameter change allows the model to predict multiple labels simultaneously while maintaining a relatively simple architecture suitable for tabular data, thus improving adaptability without significantly increasing complexity.
Data Source
AI summary
The disclosed computer-implemented method includes normalizing tabular data corresponding to a query target, and inputting the normalized tabular data into a shallow neural network corresponding to a fully-connected three-layer model comprising an input layer, a hidden layer, and an output layer. Normalizing the tabular data may replace feature detection for the shallow neural network. The method may further include predicting a plurality of classifications for the query target, wherein a plurality of nodes of the output layer respectively correspond to the plurality of classifications. Various other methods, systems, and computer-readable media are also disclosed.