Feature extraction — это процесс получения информативных и неизбыточных признаков из сырых данных или промежуточных представлений модели. Он может означать ручное или автоматизированное преобразование входов, включая использование pretrained models для получения представлений для downstream tasks.
In machine learning, pattern recognition, and image processing, feature extraction starts from an initial set of measured data and builds derived values (features) intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and in some cases leading to better human interpretations.
Overloaded term having either of the following definitions: - Retrieving intermediate feature representations calculated by an unsupervised or pre-trained model (for example, hidden layer values in a neural network) for use in another model as input. - Synonym for feature engineering.
In the context of generative AI, feature extraction refers to the process of identifying and selecting relevant features from input data to be used in model training. These features are then used to generate new data that resembles the original input. For example, in image generation, feature extraction might involve identifying edges, textures, and colors. In natural language processing, it could involve extracting keywords, phrases, and grammatical structures. The extracted features are then used by the generative model to create new content.