18360229. APPARATUS AND METHOD WITH OUT-OF-DISTRIBUTION DATA DETECTION simplified abstract (Samsung Electronics Co., Ltd.)
Contents
APPARATUS AND METHOD WITH OUT-OF-DISTRIBUTION DATA DETECTION
Organization Name
Inventor(s)
APPARATUS AND METHOD WITH OUT-OF-DISTRIBUTION DATA DETECTION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18360229 titled 'APPARATUS AND METHOD WITH OUT-OF-DISTRIBUTION DATA DETECTION
The patent application describes an apparatus and method for detecting out-of-distribution data using a neural network model.
- The apparatus includes processors that generate output data based on input data and determine if the input data is out-of-distribution.
- The method involves calculating loss values for neighboring data within a reference distance from the input data and comparing them to a threshold to detect out-of-distribution data.
- This technology aims to improve the accuracy and reliability of neural network models by identifying data that is different from the training data.
Potential Applications: - Enhancing the performance of machine learning models in various industries such as finance, healthcare, and autonomous vehicles. - Improving the security of systems by detecting anomalies in data that could indicate malicious activity.
Problems Solved: - Addressing the challenge of identifying out-of-distribution data that can lead to inaccurate predictions and decisions. - Enhancing the trustworthiness of neural network models by detecting data that deviates from the training set.
Benefits: - Increased accuracy and robustness of machine learning models. - Enhanced security and reliability in data-driven applications. - Improved decision-making based on more reliable data analysis.
Commercial Applications: Title: "Enhanced Data Detection Technology for Machine Learning Applications" This technology can be utilized in industries such as finance for fraud detection, healthcare for patient diagnosis, and autonomous vehicles for object recognition. The market implications include improved efficiency, reduced risks, and enhanced performance in data-driven systems.
Questions about Data Detection Technology: 1. How does this technology contribute to the advancement of machine learning models? - This technology enhances the reliability and accuracy of machine learning models by detecting out-of-distribution data that can lead to errors in predictions. 2. What are the potential implications of using this technology in autonomous vehicles? - By detecting anomalies in data, this technology can improve the safety and performance of autonomous vehicles by ensuring accurate object recognition and decision-making processes.
Original Abstract Submitted
An apparatus and method with out-of-distribution data detection is provided. The apparatus includes one or more processors configured to execute instructions; and one or more memories storing the instructions, wherein, the execution of the instructions by the one or more processors configures the one or more processors to generate output data using a neural network model provided input data; determine, based on the output data, a maximum loss value among calculated loss values that correspond to neighboring data within a reference distance from the input data; and detect, based on the maximum loss value and a threshold, whether the input data is out-of-distribution (OOD) data that is different from in-distribution data corresponding to training data used in a training of the neural network model.
