Multi-Label File Ranking With Siamese Masking for New Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document management systems struggle with efficiently ranking documents based on multiple labels, failing to account for correlations between labels and requiring separate models for each class, which complicates the addition of new labels and increases computational resources.
Innovation Solution
A hybrid approach using a Siamese network and masking layer within a single model to accommodate multiple labels, allowing for simultaneous ranking and selective back-propagation of gradients based on user preferences, enabling efficient addition of new labels and reducing computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If separate models are used for each class label, then each label can be ranked independently, but the system complexity increases and storage requirements multiply
Solution Approach 1:
The patent merges multiple separate ranking models into a single multi-label ranking model that processes multiple labels simultaneously. This unified model shares common parameters and computation pathways, reducing overall system complexity while maintaining the ability to handle multiple labels. The loss function combines gradients from all labels to update shared model parameters efficiently.
Solution Approach 2:
The patent creates a universal ranking model that handles multiple different labels through a single system. This multi-functional model uses a unified architecture with label-specific processing through masking mechanisms, allowing one model to perform the work of multiple specialized models while reducing redundancy.
2Reliability
If separate models are used for each class label, then each label ranking can be optimized independently, but storage space and computing power requirements increase
Solution Approach 1:
The patent combines multiple label-specific ranking tasks into a single multi-label model, sharing computational resources and model parameters across all labels. This reduces the total storage required for model weights and decreases computing power needs by eliminating redundant computations that would occur if separate models were maintained for each label.
3Productivity
If a single model is used for multiple labels, then resource efficiency improves, but the model must process all labels simultaneously which may reduce precision for individual labels
Solution Approach 1:
The patent applies local quality by introducing label-specific masking mechanisms within the unified model. The masking layer selectively activates or deactivates specific output neurons based on the target label, allowing the model to focus computational attention on relevant labels while maintaining overall efficiency. This enables precise label-specific ranking within the efficient single-model framework.
4Adaptability or versatility
If new labels are added to separate models, then each model remains specialized, but retraining is required for all models
Solution Approach 1:
The patent builds a universal multi-label model that can accommodate new labels through its flexible architecture. When new labels are introduced, the model can be extended by adding corresponding output neurons and updating the masking mechanism, without requiring retraining of the entire model structure. This significantly reduces adaptation time compared to retraining multiple separate specialized models.
Data Source
AI summary
A multi-label ranking method includes receiving, at a processor and from a first set of artificial neural networks (ANNs), multiple signals representing a first set of ANN output pairs for a first label. A signal representing a second set of ANN output pairs for a second label different from the first label is received at the processor from a second set of ANNs different from the first set of ANNs, substantially concurrently with the first set of ANN output pairs. A first activation function is solved based on the first set of ANN output pairs, and a second activation function is solved based on the second set of ANN output pairs. Loss values are calculated based on the solved activations, and a mask is generated based on at least one ground truth label. A signal, including a representation of the mask, is sent from the processor to each of the sets of ANNs.