Time-Shift Correction for Machine-Learning Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inaccurate timestamps in event data used for training machine-learning models, particularly in fields like medicine, compromise the quality and reliability of these models, posing a risk to subject safety.
Innovation Solution
A method to correct machine-learning model training data by modifying timestamps using a time-shift function based on a predetermined uncertainty range, which can be random or probabilistic, to account for errors in timestamp recording.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If timestamps are recorded manually in event data, then the data collection process is simple and flexible, but the timestamp accuracy deteriorates due to human error and clock inaccuracy
Solution Approach 1:
The system performs preliminary actions by automatically generating accurate timestamps at the moment events occur, before any manual recording can introduce errors. Event detection triggers automatic timestamp generation, ensuring precision is established upfront rather than relying on subsequent manual entry
Solution Approach 2:
The patent replaces the manual mechanical recording system with an automated electronic system. Event detection mechanisms automatically trigger timestamp generation and data recording, eliminating human operators from the timing process and thereby removing sources of human error and clock synchronization issues
2Productivity
If inaccurate timestamp data is used to train machine-learning models, then the training process is faster and requires less data processing, but the model reliability and safety deteriorate
Solution Approach 1:
Data correction is performed as a preliminary step before model training begins. The system identifies and corrects timestamp inaccuracies in the training dataset upfront, so that the model receives corrected data without requiring slower reprocessing during training. This maintains training speed while improving reliability
Solution Approach 2:
The system incorporates feedback mechanisms where timestamp accuracy is continuously monitored and corrected. Correction algorithms analyze timestamp patterns and apply adjustments based on detected errors, creating a feedback loop that improves data quality before it reaches the model training stage, thereby enhancing reliability without sacrificing productivity
Data Source
AI summary
Proposed concepts thus aim to provide schemes, solutions, concepts, designs, methods and systems pertaining to improving (i.e. increasing accuracy and/or reliability) machine-learning models by correcting data used to train such models. In particular, a timestamp of training data describing an event is modified according to a time-shift function and a predetermined time uncertainty range. In this way, an uncertainty/inaccuracy of the recording of the timestamp may be compensated for, such that a quality of the training data may be improved.