Transfer learning based few-shot classification using optimal transport mapping from preprocessed latent space of backbone neural network.
Tomás ChobolaDaniel VasataPavel KordíkPublished in: MetaDL@AAAI (2021)
Keyphrases
- transfer learning
- latent space
- neural network
- text classification
- machine learning
- learning tasks
- labeled data
- machine learning algorithms
- pattern recognition
- feature space
- knowledge transfer
- reinforcement learning
- cross domain
- collaborative filtering
- multi task
- supervised learning
- semi supervised learning
- active learning
- parameter space
- latent variables
- model selection
- data mining
- domain specific
- decision trees
- dimensionality reduction
- text mining
- prior knowledge
- learning algorithm
- support vector
- feature selection
- feature extraction
- manifold learning
- matrix factorization
- information retrieval
- probabilistic latent semantic analysis
- target domain
- gaussian processes
- gaussian process
- training data
- pairwise
- feature vectors
- knn
- semi supervised
- image classification