画像認識システムにおける機械学習の事例として, 適切なものはどれか。
令和3年度 ITパスポート試験 問20
This question's explanation and your answers so far.
問 20
Incorrectア
A mechanism that unlocks a door via iris image authentication is merely an application of “biometric authentication” that matches against a pre-registered iris pattern. The computer is not learning features from the data on its own, so it is not appropriate as an example of machine learning.
- イ
正解 Machine learning is a technology in which a large amount of data is given to a computer, and the computer itself automatically discovers the regularities and features hidden in that data. Choice イ states that “by feeding a large number of fruit photos into a computer, the computer itself became able to automatically extract the features of the fruit,” meaning it learns the features from the image data on its own. This is a typical example of machine learning in an image recognition system (in particular, as in deep learning, features are extracted automatically from the data without people designing them), so the correct answer is イ.
Incorrectウ
Unlocking a smartphone via fingerprint image recognition is likewise a use case of “biometric authentication” that matches against a registered fingerprint. It does not involve any process of acquiring new features through learning, so it cannot be called an example of machine learning.
Incorrectエ
Typing an inquiry in text and receiving a response in a conversational form is an example of natural language processing (processing of text and words), such as a chatbot, and is not an “image recognition system” in the first place. It falls outside the scope of the question, so it is not appropriate.
- Machine learning is a technology in which the computer itself automatically discovers regularities and features from a large amount of data.
- Extracting features automatically from data without people designing them is a characteristic of deep learning, and in image recognition the key point is learning features from a large number of images.