A Stochastic View of Optimal Regret through Minimax Duality

We study the regret of optimal strategies for online convex optimization games. Using von Neumann's minimax theorem, we show that the optimal regret in this adversarial setting is closely related to the behavior of the empirical minimization algorithm in a stochastic process setting: it is equal to the maximum, over joint distributions of the adversary's action sequence, of the difference between a sum of minimal expected losses and the minimal empirical loss. We show that the optimal regret has a natural geometric interpretation, since it can be viewed as the gap in Jensen's inequality for a concave functional--the minimizer over the player's actions of expected loss--defined on a set of probability distributions. We use this expression to obtain upper and lower bounds on the regret of an optimal strategy for a variety of online learning problems. Our method provides upper bounds without the need to construct a learning algorithm; the lower bounds provide explicit optimal strategies for the adversary. Peter L. Bartlett, Alexander Rakhlin

Application of an optimisation algorithm to configure an internal fixation device

Fractures of long bones are sometimes treated using various types of fracture fixation devices including internal plate fixators. These are specialised plates which are used to bridge the fracture gap(s) whilst anatomically aligning the bone fragments. The plate is secured in position by screws. The aim of such a device is to support and promote the natural healing of the bone. When using an internal fixation device, it is necessary for the clinician to decide upon many parameters, for example, the type of plate and where to position it; how many and where to position the screws. While there have been a number of experimental and computational studies conducted regarding the configuration of screws in the literature, there is still inadequate information available concerning the influence of screw configuration on fracture healing. Because screw configuration influences the amount of flexibility at the area of fracture, it has a direct influence on the fracture healing process. Therefore, it is important that the chosen screw configuration does not inhibit the healing process. In addition to the impact on the fracture healing process, screw configuration plays an important role in the distribution of stresses in the plate due to the applied loads. A plate that experiences high stresses is prone to early failure. Hence, the screw configuration used should not encourage the occurrence of high stresses. This project develops a computational program in Fortran programming language to perform mathematical optimisation to determine the screw configuration of an internal fixation device within constraints of interfragmentary movement by minimising the corresponding stress in the plate. Thus, the optimal solution suggests the positioning and number of screws which satisfies the predefined constraints of interfragmentary movements. For a set of screw configurations the interfragmentary displacement and the stress occurring in the plate were calculated by the Finite Element Method. The screw configurations were iteratively changed and each time the corresponding interfragmentary displacements were compared with predefined constraints. Additionally, the corresponding stress was compared with the previously calculated stress value to determine if there was a reduction. These processes were continued until an optimal solution was achieved. The optimisation program has been shown to successfully predict the optimal screw configuration in two cases. The first case was a simplified bone construct whereby the screw configuration solution was comparable with those recommended in biomechanical literature. The second case was a femoral construct, of which the resultant screw configuration was shown to be similar to those used in clinical cases. The optimisation method and programming developed in this study has shown that it has potential to be used for further investigations with the improvement of optimisation criteria and the efficiency of the program.

Optimal control of distributed generators and capacitors by hybrid DPSO

In this paper, a comprehensive planning methodology is proposed that can minimize the line loss, maximize the reliability and improve the voltage profile in a distribution network. The injected active and reactive power of Distributed Generators (DG) and the installed capacitor sizes at different buses and for different load levels are optimally controlled. The tap setting of HV/MV transformer along with the line and transformer upgrading is also included in the objective function. A hybrid optimization method, called Hybrid Discrete Particle Swarm Optimization (HDPSO), is introduced to solve this nonlinear and discrete optimization problem. The proposed HDPSO approach is a developed version of DPSO in which the diversity of the optimizing variables is increased using the genetic algorithm operators to avoid trapping in local minima. The objective function is composed of the investment cost of DGs, capacitors, distribution lines and HV/MV transformer, the line loss, and the reliability. All of these elements are converted into genuine dollars. Given this, a single-objective optimization method is sufficient. The bus voltage and the line current as constraints are satisfied during the optimization procedure. The IEEE 18-bus test system is modified and employed to evaluate the proposed algorithm. The results illustrate the unavoidable need for optimal control on the DG active and reactive power and capacitors in distribution networks.

Optimal-stopping control for airborne collision avoidance and return-to-course flight

This paper considers an aircraft collision avoidance design problem that also incorporates design of the aircraft’s return-to-course flight. This control design problem is formulated as a non-linear optimal-stopping control problem; a formulation that does not require a prior knowledge of time taken to perform the avoidance and return-to-course manoeuvre. A dynamic programming solution to the avoidance and return-to-course problem is presented, before a Markov chain numerical approximation technique is described. Simulation results are presented that illustrate the proposed collision avoidance and return-to-course flight approach.

Optimal online prediction in adversarial environments

In many prediction problems, including those that arise in computer security and computational finance, the process generating the data is best modelled as an adversary with whom the predictor competes. Even decision problems that are not inherently adversarial can be usefully modeled in this way, since the assumptions are sufficiently weak that effective prediction strategies for adversarial settings are very widely applicable.

Optimal online prediction in adversarial environments

In many prediction problems, including those that arise in computer security and computational finance, the process generating the data is best modelled as an adversary with whom the predictor competes. Even decision problems that are not inherently adversarial can be usefully modeled in this way, since the assumptions are sufficiently weak that effective prediction strategies for adversarial settings are very widely applicable.

Optimal strategies and minimax lower bounds for online convex games

A number of learning problems can be cast as an Online Convex Game: on each round, a learner makes a prediction x from a convex set, the environment plays a loss function f, and the learner’s long-term goal is to minimize regret. Algorithms have been proposed by Zinkevich, when f is assumed to be convex, and Hazan et al., when f is assumed to be strongly convex, that have provably low regret. We consider these two settings and analyze such games from a minimax perspective, proving minimax strategies and lower bounds in each case. These results prove that the existing algorithms are essentially optimal.