Smoothing Discretized Values via Transition Matrix for Authentication

Solution Overview

Problem

Machine learning models face inefficiencies due to 'discretization shock' when users transition between mapped values, leading to sudden changes in risk scores and increased false positives in adaptive authentication systems.

Innovation Solution

A method to smooth discretized values by computing a smoothed feature score based on transition probabilities between discretized values, using a transition matrix to weight scores and reduce the impact of abrupt changes, thereby minimizing 'discretization shock'.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If discretization procedure is used to map ranges of values to smaller subset of values, then learning efficiency is improved and quality of learning increases, but discretization shock occurs causing sudden changes in risk scores

Engineering Contradiction:

Improvelearning efficiencyVSAvoidrisk score stability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system pre-calculates transition probabilities between adjacent discretized values and stores them in a transition matrix before scoring. When a user transitions between buckets, the pre-computed probabilities are immediately applied to smooth the risk score change, preventing discretization shock without requiring real-time calculations.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The transition matrix acts as an intermediary between the discretization process and the risk scoring mechanism. It mediates the abrupt score changes by introducing probability-based weighting that gradually transitions scores between adjacent buckets, thereby eliminating the shock effect while maintaining the benefits of discretization.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If discretization procedure is used to organize values into buckets, then the range of values to deal with is reduced, but false positives increase due to sudden score changes

Engineering Contradiction:

Improvevalue range complexityVSAvoidauthentication accuracy

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The system pre-calculates transition probabilities between adjacent discretized values and stores them in a transition matrix before scoring. When a user transitions between buckets, the pre-computed probabilities are immediately applied to smooth the risk score change, preventing discretization shock without requiring real-time calculations.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system changes the parameter representation by introducing transition probabilities as an additional dimension. Instead of directly using discrete bucket values, the system transforms them into probability-weighted scores that continuously reflect the likelihood of transitioning between states, thereby reducing false positives while maintaining simplified value ranges.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If discretization procedure is used to map user-age to buckets, then learning effectiveness is improved, but user experience is impaired by additional authentication challenges

Engineering Contradiction:

Improvelearning effectivenessVSAvoiduser experience

Core Design Contradiction:

ProductivityVSEase of operation

Solution Approach 1:

The system pre-calculates transition probabilities between adjacent discretized values and stores them in a transition matrix before scoring. When a user transitions between buckets, the pre-computed probabilities are immediately applied to smooth the risk score change, preventing discretization shock without requiring real-time calculations.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The transition matrix acts as an intermediary between the discretization process and the risk scoring mechanism. It mediates the abrupt score changes by introducing probability-based weighting that gradually transitions scores between adjacent buckets, thereby eliminating the shock effect while maintaining the benefits of discretization.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS10511585B1Smoothing of discretized values using a transition matrix

Publication Date: 2019.12.17 EMC IP HLDG CO LLC

AI summary

Techniques are provided for smoothing discretized values used, for example, for authentication or identity assurance. An illustrative method comprises obtaining at least one probability of transitioning between at least two discretized values of a given feature; computing a smoothed feature score for the given feature for a transition from a first one of the discretized values to a second one of the discretized values based on the probability of the transition from the first discrete value to the second discrete value; and performing one or more of authenticating a user and verifying an identity of a user based at least in part on the smoothed feature score. The probabilities of transitioning between the discretized values are optionally stored in a transition matrix. Feature scores for first and second discretized values are optionally weighted based on the probability of the transition from the first discrete value to the second discrete value.