We formulate the problem of finding robust speech features for supervised learning with small sample complexity as a problem of learning representations of the signal that are maximally invariant to intraclass transformations and deformations.
We formulate the problem of extracting robust speech features for supervised learning with small sample complexity as a problem of learning representations of the signal that are maximally invariant to intraclass transformations and deformations.
In turn, these bounds allow us to derive explicitly the sample complexity of these problems.
The first contribution is to derive the sample complexity of various constrained control problems.
In the second part of the thesis, we explore how data collected adaptively with a constantly updated estimation can lead to signifcant reduction in sample complexity for multiple hypothesis testing problems.
Our result resolves the long-standing open problem on the sample complexity of Markov decision process and provides new insights on how to use data efficiently in learning and optimization.
The chapter develops a comprehensive theory of the complexity of induction, which analyzes the inherent difficulty of various kinds of learning problems in terms of sample complexity and computational complexity.
In the first part, we present sample complexity results for two classes of RL problems -- deterministic systems with general forms of VFG and Markov decision processes (MDPs) with a finite hypothesis class.
What are the trade-offs between the parameters at play (e.g., dimensionality of the problem, size of the data sample, complexity)?
The algorithm and analysis can be extended to solve two-person stochastic games and feature-based Markov decision problems while achieving near-optimal sample complexity.
