Explainable Neural Classifier Layers for User-Readable AI Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles lack explainability, as current data-driven learning models are 'black boxes' that fail to provide clear explanations for their actions, leading to user discomfort due to the opaque nature of their decision-making processes.
Innovation Solution
An explainable learning system is developed that extends Convolutional Neural Networks (CNNs) to capture semantic content learned by original classifiers, enabling the generation of user-understandable explanations for predicted outputs by designating optimal semantic content layers and training new connections to produce accurate and interpretable explanations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If connectionist learning models (black boxes) are used to achieve high classification accuracy, then prediction accuracy is improved, but interpretability deteriorates
Solution Approach 1:
The patent introduces an intermediary component called an 'explanation generator' that sits between the black box classifier and the user. This explanation generator takes the internal representations and decision pathways from the classifier and transforms them into human-understandable explanations, thereby mediating between the high-accuracy black box model and the need for interpretability without sacrificing either
2Loss of information
If extended architecture with semantic content layers is added to the neural network, then explainability is improved, but device complexity increases
Solution Approach 1:
The patent segments the neural network architecture into distinct functional modules: the original classifier network for accurate prediction, and an added explanation generation network for interpretability. By dividing the system into these separate but interconnected segments, the patent achieves explainability while managing complexity through modular design, where each segment has a specific function
Solution Approach 2:
The patent adds a new dimension to the traditional classifier architecture by incorporating semantic content layers and explanation generation pathways. This dimensional extension transforms the system from a single-function classifier into a multi-functional system that simultaneously performs classification and explanation generation, addressing the explainability need without fundamentally redesigning the core classifier
Data Source
AI summary
Methods, systems, and apparatuses to build an explainable user output to receive input feature data by a neural network of multiple layers of an original classifier; determine a semantic function to label data samples with semantic categories; determine a semantic accuracy for each layer of the original classifier within the neural network; compare each layer based on results from the comparison of the semantic accuracy; designate a layer based on an amount of computed semantic accuracy; extend the designated layer by a category branch to the neural network to extract semantic data samples from the semantic content to train a set of new connections of an explainable classifier to compute a set of output explanations with an accuracy measure associated each output explanation for each semantic category of the plurality of semantic categories, and compare the accuracy measure for each output explanation to generate the output explanation in a user understandable format.