A general-purpose framework for learning by observation

Floyd, Michael S.

doi:10.22215/etd/2013-08584

Cited by 1 publication

(7 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this thesis we will look into two principal existing approaches to learning nonreactive behaviour from observation. One approach relies on graphical models such as Dynamic Bayesian Networks (DBN) [1] while the other uses a technique called Temporal Backtracking (TB) [7]. Both approaches are very recent and have not been studied or validated throughly yet.…”

Section: Ob Jectivesmentioning

confidence: 99%

“…• A comparative analysis of a case-based solution (temporal backtracking) [7] against the stochastic solution using the benchmark provided by Ontañón et al [1]. This resulted in the conclusion that for highly complex behaviour, the stochastic solution is better (Section 4.2).…”

Section: Contributionsmentioning

confidence: 99%

“…(c) The observation model is calculated based on the probability that the output is Y given the state was S. observation. In the next section, we shift to the general framework of the approach used by Michael Floyd [7], which is Case-based reasoning.…”

Section: Expectation Maximizationmentioning

confidence: 99%

“…Because CBR solves problems using cases, CBR is highly dependent on the quality of the knowledge being represented. In the application of CBR to LFO, Michael Floyd [7] proposes a simplied representation of a case by mapping the sensory input S to the action A at time t. Cases are stored in a collection that forms the knowledge base, and the collection is termed a case base. The case base represents the past experiences learned.…”

Section: Case-based Reasoningmentioning

confidence: 99%

“…The two solutions reviewed in this thesis are the Dynamic Bayesian Network [1] and the Temporal Backtracking method [7].…”

mentioning

confidence: 99%

See 4 more Smart Citations

Contributions to Techniques for Learning Non-Reactive Behaviour from Observation

Chan¹

View full text Add to dashboard Cite

Learning from observation allows an expert to train a software agent or robot without explicitly programming the behaviour. Behaviour learned can be broken down into categories: reactive and non-reactive. Actions in reactive behaviour are based on the current environment state whereas actions in non-reactive uses both current state, and any past action or states. We will analyze and compare using a common benchmark two recent and partially studied approaches to learning non-reactive behaviour from observation: Dynamic Bayesian Networks (DBN) and Temporal Backtracking (TB). The goal is to characterize situations where one approach should be preferred over the other. We will attempt to provide a more general case-based reasoning framework for non-reactive behaviour learning. Using the framework and the benchmark, we will analyze and compare three dierent metrics for comparing cases: run similarity, edit distance and Jaccard distance. This will allow characterization of situations where one metric should outperform the other.

show abstract

Section: Ob Jectivesmentioning

confidence: 99%

Section: Contributionsmentioning

confidence: 99%

Section: Expectation Maximizationmentioning

confidence: 99%

Section: Case-based Reasoningmentioning

confidence: 99%

“…The two solutions reviewed in this thesis are the Dynamic Bayesian Network [1] and the Temporal Backtracking method [7].…”

mentioning

confidence: 99%