機械学習のための方法、システムおよびコンピュータプログラム製品(埋め込みを用いるインクリメンタル機械学習)
JP7891884B2Active Publication Date: 2026-07-17INTERNATIONAL BUSINESS MACHINE CORPORATION
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- INTERNATIONAL BUSINESS MACHINE CORPORATION
- Filing Date
- 2022-09-16
- Publication Date
- 2026-07-17
AI Technical Summary
Benefits of technology
【0013】 また、本発明の実施形態は、機械学習モデルを再訓練することなく、結果としての候補出力を予測するためのグループに新しい候補出力を追加してもよい。これにより、コンピューティング資源およびメモリ資源を節約し、機械学習の速度を大幅に向上させることができる。
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure 0007891884000001 
Figure 0007891884000002 
Figure 0007891884000003
Abstract
To provide a method for incremental machine learning, a system, and a computer program product using embedding.SOLUTION: Machine learning is for generating a result including a new output, and a machine learning model is trained to determine a candidate output among a plurality of candidate outputs. A first embedding related to the plurality of candidate outputs is generated from a first set of training data by an intermediate layer of the trained machine learning model. A second embedding related to a new candidate output is generated from a second set of the training data by the intermediate layer of the trained machine learning model. A third embedding is determined for input data by the intermediate layer of the trained machine learning model. A candidate output as a result for the input data is predicted from a group of the plurality of candidate outputs, and a new candidate output based on a distance of the third embedding to the first embedding and second embedding.SELECTED DRAWING: Figure 6
Need to check novelty before this filing date? Find Prior Art