Information Processing With Local Models for Latent-Distance Explanation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Transformation models generated through metric learning are black box models, making it difficult for users to understand the factors affecting the distance between two tuples of data, thus hindering user convenience.
Innovation Solution
An information processing apparatus and method that transforms feature vector data into latent vector data, extracts neighbor data with a shorter distance from query data, generates a local model to estimate latent distance, and calculates the element contribution degree of each feature's effect on this distance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a transformation model is generated through metric learning, then the distance between similar data tuples becomes short and dissimilar data tuples becomes long, but the model becomes a black box that cannot be understood by users
Solution Approach 1:
The patent segments the transformation model into multiple interpretable components: (1) the transformation model itself that maps feature vectors to latent vectors, (2) the distance calculation unit that computes distances in latent space, and (3) the explanation generation unit that provides human-readable interpretations. This segmentation allows the system to maintain accurate distance calculation while making each component transparent and understandable to users.
Solution Approach 2:
The patent introduces an explanation generation unit as an intermediary between the black box transformation model and the user. This intermediary translates the internal transformations and distance calculations into human-readable explanations, allowing users to understand what factors contribute to the calculated distances without compromising the model's computational accuracy.
2Ease of operation
If the transformation model is made interpretable, then user convenience is enhanced, but the complexity of the system increases
Solution Approach 1:
The patent applies local quality by providing explanations at the specific level of individual distance calculations rather than requiring the entire system to be fundamentally redesigned. The explanation generation unit creates localized interpretations for each query, explaining which feature vector elements contribute most to the distance between specific data tuples. This approach enhances user convenience for specific operations without requiring complex system-wide changes.
3Measurement precision
If feature vector elements are transformed to latent vector elements, then the distance metric becomes more accurate, but the factors affecting the distance become hidden
Solution Approach 1:
The patent implements feedback by having the explanation generation unit analyze the transformation results and provide information about which feature vector elements most influenced the latent vector transformation and subsequent distance calculation. This feedback loop allows users to understand the relationship between original features and transformed distances, making the influencing factors detectable and measurable despite the transformation.
Data Source
AI summary
An information processing apparatus 4 includes: a transformation unit 423 configured to transform a plurality of tuples of feature vector data XV in a presentation space into a plurality of tuples of latent vector data ZV in a latent space; an extraction unit 423 configured to extract, based on the plurality of tuples of latent vector data, feature vector data having a shorter distance from query data DQ in the latent space than the other feature vector data, as neighbor data DNi from among the plurality of tuples of feature vector data; a generation unit 424 configured to generate, based on the neighbor data, a local model LM that outputs an estimation value dpi of a latent distance di when difference information Vi is inputted, the latent distance being a distance between the query data and the neighbor data in the latent space, the difference information being related to a difference, for each element of the features, between the query data and the neighbor data in the presentation space; and a calculation unit 425 configured to calculate, based on the local model and the difference information, an element contribution degree ci,f representing a magnitude of an effect that each element of the features of the neighbor data exerts on the latent distance.