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.

Shifting : one-inclusion mistake bounds and sample compression

We present new expected risk bounds for binary and multiclass prediction, and resolve several recent conjectures on sample compressibility due to Kuzmin and Warmuth. By exploiting the combinatorial structure of concept class F, Haussler et al. achieved a VC(F)/n bound for the natural one-inclusion prediction strategy. The key step in their proof is a d=VC(F) bound on the graph density of a subgraph of the hypercube—one-inclusion graph. The first main result of this report is a density bound of n∙choose(n-1,≤d-1)/choose(n,≤d) < d, which positively resolves a conjecture of Kuzmin and Warmuth relating to their unlabeled Peeling compression scheme and also leads to an improved one-inclusion mistake bound. The proof uses a new form of VC-invariant shifting and a group-theoretic symmetrization. Our second main result is an algebraic topological property of maximum classes of VC-dimension d as being d-contractible simplicial complexes, extending the well-known characterization that d=1 maximum classes are trees. We negatively resolve a minimum degree conjecture of Kuzmin and Warmuth—the second part to a conjectured proof of correctness for Peeling—that every class has one-inclusion minimum degree at most its VC-dimension. Our final main result is a k-class analogue of the d/n mistake bound, replacing the VC-dimension by the Pollard pseudo-dimension and the one-inclusion strategy by its natural hypergraph generalization. This result improves on known PAC-based expected risk bounds by a factor of O(log n) and is shown to be optimal up to a O(log k) factor. The combinatorial technique of shifting takes a central role in understanding the one-inclusion (hyper)graph and is a running theme throughout

A unifying view of multiple kernel learning

Recent research on multiple kernel learning has lead to a number of approaches for combining kernels in regularized risk minimization. The proposed approaches include different formulations of objectives and varying regularization strategies. In this paper we present a unifying general optimization criterion for multiple kernel learning and show how existing formulations are subsumed as special cases. We also derive the criterion's dual representation, which is suitable for general smooth optimization algorithms. Finally, we evaluate multiple kernel learning in this framework analytically using a Rademacher complexity bound on the generalization error and empirically in a set of experiments.

Matrix regularization techniques for online multitask learning

In this paper we examine the problem of prediction with expert advice in a setup where the learner is presented with a sequence of examples coming from different tasks. In order for the learner to be able to benefit from performing multiple tasks simultaneously, we make assumptions of task relatedness by constraining the comparator to use a lesser number of best experts than the number of tasks. We show how this corresponds naturally to learning under spectral or structural matrix constraints, and propose regularization techniques to enforce the constraints. The regularization techniques proposed here are interesting in their own right and multitask learning is just one application for the ideas. A theoretical analysis of one such regularizer is performed, and a regret bound that shows benefits of this setup is reported.

Closing the gap between bandit and full-information online optimization : high-probability regret bound

We demonstrate a modification of the algorithm of Dani et al for the online linear optimization problem in the bandit setting, which allows us to achieve an O( \sqrt{T ln T} ) regret bound in high probability against an adaptive adversary, as opposed to the in expectation result against an oblivious adversary of Dani et al. We obtain the same dependence on the dimension as that exhibited by Dani et al. The results of this paper rest firmly on those of Dani et al and the remarkable technique of Auer et al for obtaining high-probability bounds via optimistic estimates. This paper answers an open question: it eliminates the gap between the high-probability bounds obtained in the full-information vs bandit settings.

Adaptive online gradient descent

We study the rates of growth of the regret in online convex optimization. First, we show that a simple extension of the algorithm of Hazan et al eliminates the need for a priori knowledge of the lower bound on the second derivatives of the observed functions. We then provide an algorithm, Adaptive Online Gradient Descent, which interpolates between the results of Zinkevich for linear functions and of Hazan et al for strongly convex functions, achieving intermediate rates between [square root T] and [log T]. Furthermore, we show strong optimality of the algorithm. Finally, we provide an extension of our results to general norms.

Multitask learning with expert advice

We consider the problem of prediction with expert advice in the setting where a forecaster is presented with several online prediction tasks. Instead of competing against the best expert separately on each task, we assume the tasks are related, and thus we expect that a few experts will perform well on the entire set of tasks. That is, our forecaster would like, on each task, to compete against the best expert chosen from a small set of experts. While we describe the "ideal" algorithm and its performance bound, we show that the computation required for this algorithm is as hard as computation of a matrix permanent. We present an efficient algorithm based on mixing priors, and prove a bound that is nearly as good for the sequential task presentation case. We also consider a harder case where the task may change arbitrarily from round to round, and we develop an efficient approximate randomized algorithm based on Markov chain Monte Carlo techniques.

Learning the Kernel Matrix with Semi-Definite Programming

Kernel-based learning algorithms work by embedding the data into a Euclidean space, and then searching for linear relations among the embedded data points. The embedding is performed implicitly, by specifying the inner products between each pair of points in the embedding space. This information is contained in the so-called kernel matrix, a symmetric and positive definite matrix that encodes the relative positions of all points. Specifying this matrix amounts to specifying the geometry of the embedding space and inducing a notion of similarity in the input space -- classical model selection problems in machine learning. In this paper we show how the kernel matrix can be learned from data via semi-definite programming (SDP) techniques. When applied to a kernel matrix associated with both training and test data this gives a powerful transductive algorithm -- using the labelled part of the data one can learn an embedding also for the unlabelled part. The similarity between test points is inferred from training points and their labels. Importantly, these learning problems are convex, so we obtain a method for learning both the model class and the function without local minima. Furthermore, this approach leads directly to a convex method to learn the 2-norm soft margin parameter in support vector machines, solving another important open problem. Finally, the novel approach presented in the paper is supported by positive empirical results.

The calculation of acoustic indices to characterise acoustic recordings of the environment

This technical report describes the methods used to obtain a list of acoustic indices that are used to characterise the structure and distribution of acoustic energy in recordings of the natural environment. In particular it describes methods for noise reduction from recordings of the environment and a fast clustering algorithm used to estimate the spectral richness of long recordings.

Estimating the probability of an invalid match for matching using the rank transform

A fundamental problem faced by stereo matching algorithms is the matching or correspondence problem. A wide range of algorithms have been proposed for the correspondence problem. For all matching algorithms, it would be useful to be able to compute a measure of the probability of correctness, or reliability of a match. This paper focuses in particular on one class for matching algorithms, which are based on the rank transform. The interest in these algorithms for stereo matching stems from their invariance to radiometric distortion, and their amenability to fast hardware implementation. This work differs from previous work in that it derives, from first principles, an expression for the probability of a correct match. This method was based on an enumeration of all possible symbols for matching. The theoretical results for disparity error prediction, obtained using this method, were found to agree well with experimental results. However, disadvantages of the technique developed in this chapter are that it is not easily applicable to real images, and also that it is too computationally expensive for practical window sizes. Nevertheless, the exercise provides an interesting and novel analysis of match reliability.

Google Android : a comprehensive introduction

Google Android, Google's new product and its first attempt to enter the mobile market, might have an equal impact on mobile users like Apple's hyped product, the iPhone. In this Technical report we are going to present the Google Android platform, what Android is, describe why it might be considered as a worthy rival to Apple's iPhone. We will describe parts of its internals, take a look "under the hood" while explaining components of the underlying operating system. We will show how to develop applications for this platform, which difficulties a developer might have to face, and how developers can possibly use other programming languages to develop for Android than the propagated language Java.

Hydrogeological Features of the Lockyer Valley

The Lockyer Valley is situated 80 km west of Brisbane and is bounded on the sou th and west by the Great Dividing Range. The valley is a major western sub - catchment of the larger Brisbane River drainage system and is drained by the Lockyer Creek. The Lockyer catchment forms approximately 20% of the total Brisbane River catchment and has an area of around 2900 km2. The Lockyer Creek is an ephemeral drainage system, and the stream and associated alluvium are the main source for irrigation water supply in the Lockyer Valley. The catchment is comprised of a number of well -defined, elongate tributaries in the south, and others in the north, which are more meandering in nature.

Calibration of the electron volt spectrometer, a deep inelastic neutron scattering spectrometer at the ISIS pulsed neutron spallation source

The electron Volt Spectrometer (eVS) is an inverse geometry filter difference spectrometer that has been optimised to measure the single atom properties of condensed matter systems using a technique known as Neutron Compton Scattering (NCS) or Deep Inelastic Neutron Scattering (DINS). The spectrometer utilises the high flux of epithermal neutrons that are produced by the ISIS neutron spallation source enabling the direct measurement of atomic momentum distributions and ground state kinetic energies. In this paper the procedure that is used to calibrate the spectrometer is described. This includes details of the method used to determine detector positions and neutron flight path lengths as well as the determination of the instrument resolution. Examples of measurements on 3 different samples are shown, ZrH2, 4He and Sn which show the self-consistency of the calibration procedure.

Noise removal from wave-forms and spectrograms derived from natural recordings of the environment

The work described in this technical report is part of an ongoing project to build practical tools for the manipulation, analysis and visualisation of recordings of the natural environment. This report describes the methods we use to remove background noise from spectrograms. It updates techniques previously described in Towsey and Planitz (2011), Technical report: acoustic analysis of the natural environment, downloadable from: http://eprints.qut.edu.au/41131/. It also describes noise removal from wave-forms, a technique not described in the above 2011 technical report.

An algorithm to cluster the spectra in a spectrogram

The work described in this technical report is part of an ongoing project at QUT to build practical tools for the manipulation, analysis and visualisation of recordings of the natural environment. This report describes the algorithm we use to cluster the spectra in a spectrogram. The report begins with a brief description of the signal processing that prepares the spectrograms.