View source for 18060987. CLASS-INCREMENTAL LEARNING OF A CLASSIFIER simplified abstract (International Business Machines Corporation)

=CLASS-INCREMENTAL LEARNING OF A CLASSIFIER=


==Organization Name==

[[:Category:International Business Machines Corporation|International Business Machines Corporation]]


[[Category:International Business Machines Corporation]]

==Inventor(s)==

[[:Category:Kumudu Geethan Karunaratne of Gattikon (CH)|Kumudu Geethan Karunaratne of Gattikon (CH)]][[Category:Kumudu Geethan Karunaratne of Gattikon (CH)]]

[[:Category:Michael Andreas Hersche of Zurich (CH)|Michael Andreas Hersche of Zurich (CH)]][[Category:Michael Andreas Hersche of Zurich (CH)]]

[[:Category:Giovanni Cherubini of Rueschlikon (CH)|Giovanni Cherubini of Rueschlikon (CH)]][[Category:Giovanni Cherubini of Rueschlikon (CH)]]

[[:Category:Abu Sebastian of Adliswil (CH)|Abu Sebastian of Adliswil (CH)]][[Category:Abu Sebastian of Adliswil (CH)]]

[[:Category:Abbas Rahimi of RUESCHLIKON (CH)|Abbas Rahimi of RUESCHLIKON (CH)]][[Category:Abbas Rahimi of RUESCHLIKON (CH)]]

==CLASS-INCREMENTAL LEARNING OF A CLASSIFIER - A simplified explanation of the abstract==


This abstract first appeared for US patent application 18060987 titled 'CLASS-INCREMENTAL LEARNING OF A CLASSIFIER

The present disclosure discusses a method for training a classifier, which involves iteratively receiving training datasets, adding output vectors to memory, retraining the classifier, and updating prototype vectors.

* The classifier includes a controller and an explicit memory.
* Training involves receiving multiple training datasets with data samples of novel classes, adding output vectors to memory, retraining the classifier, minimizing distance between output vectors and prototype vectors, updating prototype vectors, and updating the explicit memory.
* The method aims to improve the classifier's ability to recognize novel classes and adapt to new data.
* By updating prototype vectors based on new training datasets, the classifier can better generalize and make accurate predictions.
* This approach enhances the classifier's performance and adaptability in various applications.

Potential Applications:
- Image recognition systems
- Speech recognition software
- Fraud detection algorithms
- Medical diagnosis tools

Problems Solved:
- Improving classifier accuracy
- Enhancing adaptability to new data
- Increasing generalization capabilities

Benefits:
- Improved classification accuracy
- Better adaptability to new classes
- Enhanced performance in various applications

Commercial Applications:
Title: Enhanced Classifier Training Method for Improved Performance in Image Recognition Systems
This technology can be utilized in industries such as:
- Healthcare for medical image analysis
- Finance for fraud detection systems
- Marketing for customer segmentation algorithms

Questions about the technology:
1. How does this method improve the classifier's ability to recognize novel classes?
- The method updates prototype vectors based on new training datasets, allowing the classifier to better generalize and adapt to new data.
2. What are the potential applications of this technology in real-world scenarios?
- This technology can be applied in image recognition systems, speech recognition software, fraud detection algorithms, and medical diagnosis tools.


==Original Abstract Submitted==

The present disclosure relates to training a classifier. The classifier includes a controller and an explicit memory. The training may include iteratively receiving one or more second training datasets, each comprising second data samples of a set of one or more associated novel classes, adding to the explicit memory one or more second output vectors indicative of the set of one or more associated novel classes, in response to providing the one or more second training datasets to the classifier, retraining the classifier using the one or more second training datasets and the first training dataset by minimizing a distance between the one or more second output vectors and the one or more prototype vectors, determining a set of updated prototype vectors indicative of first training dataset and the one or more second training datasets, and updating the explicit memory with the set of updated prototype vectors.



[[Category:G06N3/08]]
[[Category:G06N3/0464]]
[[Category:CPC_G06N3/08]]