Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 27 for “"Regret analysis"”.
-
A methodology for the robustness-based evaluation of systems-of-systems alternatives using regret analysis
… is defined as the integral of the alternative s regret over the likelihood-weighted plausible scenario space. Surrogate modeling techniques were used to overcome shortcomings associated with conventional regret analysis, including the discrete nature of scenario cases and static results. The new …
-
Nonparametric Estimation and Model Combination in a Bandit Problem with Covariates
… asymptotic strong consistency and finite-time regret analysis. In addition, although many nonparametric and parametric estimation methods in supervised learning may be applied to the randomized allocation strategy in a convenient plug-in fashion, guidance on how to choose among these estimation …
-
Semi-Autonomous Networks: Effective Control of Networked Systems through Protocols, Design, and Modeling
… objective of this dissertation is to explore the analysis and design of controlled networked dynamic systems - dubbed semi-autonomous networks. This work approaches the problem of effective control of semi-autonomous networks from three fronts: protocols which are run on individual agents in the …
-
Learning in three classic bandit strategies
… the Gittins Indices. <br><br>We also provide an analysis of the Narendra Strategy in a quenched context, giving a characterisation of consistency and a study of the rates of convergence that can be observed. <br><br>Finally we perform a regret analysis of the Thompson Strategy, showing in …
-
Quantitative Measures of Regret and Trust in Human-Robot Collaboration Systems
… interaction (sHRI) factors and in particular regret and trust. Humans experience regret during decision-making under uncertainty when they feel that a better result could be obtained if chosen differently. A framework to quantitatively measure regret is proposed in this thesis. We embed …
-
Data-Efficient Decision-Making
… allow for achieving instance-specific fast regret rates? (ii) How can these problem structures be leveraged to design \textit{practical} algorithms that achieve fast theoretical rates? Part I of this thesis investigates the above questions from an online perspective. Chapter 2 studies the …
-
Pessimistic-Optimistic Bandit Learning with Applications to Communication Networks
… pessimistic-optimistic algorithm, derive novel regret and constraint violation analysis, and support these findings by extensive experiments, including trace-based simulations. We also apply these techniques to the case of learning in a bipartite queueing network, in which the feedback delay and …
-
Online learning algorithms for reinforcement learning with function approximation
… algorithms, for which I introduce a type of analysis which is novel (and useful) to reinforcement learning in its lack of restrictive assumptions on the behaviour policy. I will introduce three novel algorithms attempting to advance the areas of kernel, empirical and theoretical reinforcement …
-
Multi-Armed Bandits with Risk-Aware Performance Measures
… at Risk and Expected Shortfall and deriving regret guarantees for both light- and heavy-tailed reward distributions. Building on this analysis, I next propose the Range Value at Risk Lower Confidence Bound algo- rithm, which relies on a novel uniform concentration argument for order …
-
Neural accelerated marginalized particle filtering for one-way coupled distributed parameter systems
… as a multi-armed bandit problem. Through a regret analysis, we show that mobile sensors can locate the pollutant sources faster than state-of-the-art search algorithms. The distributed and online PDE-based state estimation and joint path planning for mobile sensors is the first major …
-
Fast algorithms for minimally supervised learning
… are equivalent and provide complexity analysis for a hybrid MPI/OpenMP implementation. The resulting method achieves provable scalability while preserving clustering structure in high dimensions. The second part addresses representative sample selection via optimal experimental design. …
-
Optimization and Information Problems in Operations
… admits some ideal properties that facilitate our regret analysis under our dynamic programming setting. Instead of studying the policy performance with a long horizon and large quantities of inventory, we study the asymptotic policy performance over a single unit of product as the demand rate …
-
An Online Convex Optimization Approach for Network Slicing
The emergence of 5G and beyond networks introduces a significant shift in resource allocation across a wide array of services and applications, all supported by a singular physical infrastructure. This advancement necessitates an efficient mechanism design, termed Network Slicing. This concept …
-
Modelling, inference and optimization in probabilistic machine learning
… problem. In particular, through a finite-time regret analysis we show that the regret of the proposed algorithm grows sublinearly with time. Further, we propose a hierarchical extension of Daisee to encourage exploration in the region with high uncertainty.</p> <p>The new models proposed in …
-
Risk-Averse and Distributionally Robust Optimization for Online and Offline Decision-Making
… of the CVaR gradients. Our theoretical analysis demonstrates that this algorithm achieves sublinear regret with high probability. We further enhance performance through two innovative variants: one employing a new sampling strategy that leverages samples from the previous iteration to …
-
Machine Learning Methods for Data-driven Decision Making: Contextual Optimization, Causal Inference, and Algorithmic Fairness
… approaches can actually have much slower regret-convergence rates than the far simpler "estimate-then-optimize" approach. Chapter 4 considers the online setting where we collect more data as we make decisions and studies nonparametric contextual bandits with smooth expected reward …
-
ONLINE LEARNING OF MARKOV JUMP DYNAMIC SYSTEMS
… suffer a sublinear (on the time horizon length) regret when compared against the optimal control of the system in expectation. Where, in the optimal control setting, the system's parameters and the Markov Chain are known. We perform numerical simulations of this proposed algorithm learning a …
-
LEARNING ON GRAPHS: ALGORITHMS FOR CLASSIFICATION AND SEQUENTIAL DECISIONS
… we present a rapidly emerging theme in the analysis of networked data: signed networks, graphs whose edges carry a label encoding the positive or negative nature of the relationship between the connected nodes. For example, social networks and e-commerce offer several examples of signed …
Page 1 of 2