Divergent representations of ethological visual inputs emerge from supervised, unsupervised, and reinforcement learning.
Grace W. LindsayJosh MerelTom Mrsic-FlogelManeesh SahaniPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- supervised learning
- unsupervised learning
- semi supervised
- supervised classification
- unsupervised methods
- learning algorithm
- machine learning
- visual information
- mid level
- function approximation
- visual cues
- dynamic programming
- word sense induction
- discriminant projection
- state space
- unsupervised feature selection
- weakly supervised
- supervised training
- low level
- visual features
- learning problems
- sensory inputs
- learning process
- visual perception
- high level
- deep belief networks
- data sets
- supervised methods
- visual patterns
- object recognition
- reinforcement learning algorithms
- data driven
- optimal control
- markov decision processes
- labeled data
- temporal difference
- model free
- symbolic representation
- visual representations
- training set
- hidden markov models
- multi agent
- learning tasks
- higher level