Multilayer Computational Graphs for Document Classification Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computational models, such as Dirichlet-multinomial classification models, face limitations in accuracy due to their representation of training data, which affects their ability to classify documents effectively.
Innovation Solution
The use of multilayer computational graphs that operate in both forward and reverse computational orders to determine topic values and layer-specific deviation values, allowing for model adjustments and improved parameter modification, thereby enhancing classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional computational models (e.g., Dirichlet-multinomial classification models) are used for document classification, then the classification process is simple and fast, but the classification accuracy is limited due to inadequate representation of training data
Solution Approach 1:
The patent segments the computational model into multiple layers (input layer, hidden layers, output layer) where each layer processes specific aspects of the data. This multilayer structure enables more sophisticated representation of training data while maintaining manageable complexity through modular organization of computational operations.
Solution Approach 2:
The patent transitions from conventional single-layer models to multilayer computational graphs, adding the dimension of depth to the model architecture. This dimensional expansion allows the model to capture complex patterns and relationships in the data that single-layer models cannot represent, thereby improving classification accuracy.
2Measurement precision
If multilayer computational graphs are used to improve classification accuracy, then the model can better represent training data, but the training and computation process becomes more complex
Solution Approach 1:
The patent performs preliminary actions by pre-defining the multilayer computational graph structure and parameters before training begins. The forward computational order is established in advance, allowing systematic propagation of inputs through layers. This preliminary setup simplifies the training process by providing a structured framework that guides the learning procedure.
Solution Approach 2:
The patent implements feedback mechanisms through the reverse computational order, where deviation values are propagated backward from the output layer to the input layer. This feedback enables the model to learn from errors and adjust parameters systematically, making the training process more manageable despite the increased model complexity.
3Productivity
If forward computational order is used to determine topic values, then the computation flows naturally from input to output, but reverse computational order is needed to determine deviation values for model adjustment
Solution Approach 1:
The patent employs periodic action by alternating between forward computational order (for determining topic values) and reverse computational order (for determining deviation values). This periodic switching between computation directions enables the model to both process data efficiently and learn from errors systematically, balancing productivity with learning effectiveness.
Data Source
AI summary
A processing unit can successively operate layers of a multilayer computational graph (MCG) according to a forward computational order to determine a topic value associated with a document based at least in part on content values associated with the document. The processing unit can successively determine, according to a reverse computational order, layer-specific deviation values associated with the layers based at least in part on the topic value, the content values, and a characteristic value associated with the document. The processing unit can determine a model adjustment value based at least in part on the layer-specific deviation values. The processing unit can modify at least one parameter associated with the MCG based at least in part on the model adjustment value. The MCG can be operated to provide a result characteristic value associated with test content values of a test document.