On the Difficulty of Defending Self-Supervised Learning against Model Extraction.
Adam DziedzicNikita DhawanMuhammad Ahmad KaleemJonas GuanNicolas PapernotPublished in: ICML (2022)
Keyphrases
- prior knowledge
- statistical model
- learning models
- probabilistic model
- learning tasks
- learning phase
- knowledge acquisition
- learning mechanism
- learning systems
- neural nets
- classification models
- neural network model
- conceptual model
- learning algorithm
- active learning
- learned models
- hidden variables
- structured prediction
- privileged information
- experimental data
- network structure
- mobile learning
- em algorithm
- state space
- hidden markov models
- high level
- decision trees