Natural Images Allow Universal Adversarial Attacks on Medical Image Classification Using Deep Neural Networks with Transfer Learning.
Akinori MinagiHokuto HiranoKazuhiro TakemotoPublished in: J. Imaging (2022)
Keyphrases
- natural images
- transfer learning
- image classification
- neural network
- algorithmic information theory
- learning tasks
- sparse coding
- higher order
- natural scenes
- knowledge transfer
- multiscale
- image patches
- object recognition
- denoising
- input image
- cross domain
- active learning
- image representation
- reinforcement learning
- feature extraction
- labeled data
- machine learning
- semi supervised learning
- statistics of natural images
- machine learning algorithms
- image features
- collaborative filtering
- visual features
- transfer knowledge
- natural image statistics
- computer vision
- text classification
- multilayer perceptron
- feed forward
- information extraction
- image processing
- target domain
- sparse representation
- text mining
- semi supervised
- natural language processing
- high dimensional
- pairwise
- image statistics
- back propagation
- unlabeled data