Learning the learning rate for prediction with expert advice

Most standard algorithms for prediction with expert advice depend on a parameter called the learning rate. This learning rate needs to be large enough to fit the data well, but small enough to prevent overfitting. For the exponential weights algorithm, a sequence of prior work has established theoretical guarantees for higher and higher data-dependent tunings of the learning rate, which allow for increasingly aggressive learning. But in practice such theoretical tunings often still perform worse (as measured by their regret) than ad hoc tuning with an even higher learning rate. To close the gap between theory and practice we introduce an approach to learn the learning rate. Up to a factor that is at most (poly)logarithmic in the number of experts and the inverse of the learning rate, our method performs as well as if we would know the empirically best learning rate from a large range that includes both conservative small values and values that are much higher than those for which formal guarantees were previously available. Our method employs a grid of learning rates, yet runs in linear time regardless of the size of the grid.

Open problem: Shifting experts on easy data

A number of online algorithms have been developed that have small additional loss (regret) compared to the best “shifting expert”. In this model, there is a set of experts and the comparator is the best partition of the trial sequence into a small number of segments, where the expert of smallest loss is chosen in each segment. The regret is typically defined for worst-case data / loss sequences. There has been a recent surge of interest in online algorithms that combine good worst-case guarantees with much improved performance on easy data. A practically relevant class of easy data is the case when the loss of each expert is iid and the best and second best experts have a gap between their mean loss. In the full information setting, the FlipFlop algorithm by De Rooij et al. (2014) combines the best of the iid optimal Follow-The-Leader (FL) and the worst-case-safe Hedge algorithms, whereas in the bandit information case SAO by Bubeck and Slivkins (2012) competes with the iid optimal UCB and the worst-case-safe EXP3. We ask the same question for the shifting expert problem. First, we ask what are the simple and efficient algorithms for the shifting experts problem when the loss sequence in each segment is iid with respect to a fixed but unknown distribution. Second, we ask how to efficiently unite the performance of such algorithms on easy data with worst-case robustness. A particular intriguing open problem is the case when the comparator shifts within a small subset of experts from a large set under the assumption that the losses in each segment are iid.

Numerical algorithms for time-fractional subdiffusion equation with second-order accuracy

This article aims to fill in the gap of the second-order accurate schemes for the time-fractional subdiffusion equation with unconditional stability. Two fully discrete schemes are first proposed for the time-fractional subdiffusion equation with space discretized by finite element and time discretized by the fractional linear multistep methods. These two methods are unconditionally stable with maximum global convergence order of $O(\tau+h^{r+1})$ in the $L^2$ norm, where $\tau$ and $h$ are the step sizes in time and space, respectively, and $r$ is the degree of the piecewise polynomial space. The average convergence rates for the two methods in time are also investigated, which shows that the average convergence rates of the two methods are $O(\tau^{1.5}+h^{r+1})$. Furthermore, two improved algorithms are constrcted, they are also unconditionally stable and convergent of order $O(\tau^2+h^{r+1})$. Numerical examples are provided to verify the theoretical analysis. The comparisons between the present algorithms and the existing ones are included, which show that our numerical algorithms exhibit better performances than the known ones.

High order unconditionally stable difference schemes for the Riesz space-fractional telegraph equation

In this paper, a class of unconditionally stable difference schemes based on the Pad´e approximation is presented for the Riesz space-fractional telegraph equation. Firstly, we introduce a new variable to transform the original dfferential equation to an equivalent differential equation system. Then, we apply a second order fractional central difference scheme to discretise the Riesz space-fractional operator. Finally, we use (1, 1), (2, 2) and (3, 3) Pad´e approximations to give a fully discrete difference scheme for the resulting linear system of ordinary differential equations. Matrix analysis is used to show the unconditional stability of the proposed algorithms. Two examples with known exact solutions are chosen to assess the proposed difference schemes. Numerical results demonstrate that these schemes provide accurate and efficient methods for solving a space-fractional hyperbolic equation.

A low-cost attack on branch-based software watermarking schemes

In 2005, Ginger Myles and Hongxia Jin proposed a software watermarking scheme based on converting jump instructions or unconditional branch statements (UBSs) by calls to a fingerprint branch function (FBF) that computes the correct target address of the UBS as a function of the generated fingerprint and integrity check. If the program is tampered with, the fingerprint and integrity checks change and the target address will not be computed correctly. In this paper, we present an attack based on tracking stack pointer modifications to break the scheme and provide implementation details. The key element of the attack is to remove the fingerprint and integrity check generating code from the program after disassociating the target address from the fingerprint and integrity value. Using the debugging tools that give vast control to the attacker to track stack pointer operations, we perform both subtractive and watermark replacement attacks. The major steps in the attack are automated resulting in a fast and low-cost attack.

Australian Commercial Law

Australian Commercial Law offers a concise yet comprehensive introduction to commercial law in Australia. The textbook provides a thorough and detailed discussion of a variety of topics in commercial law such as agency, bailment, the sale of goods, the transfer of property and the Personal Property Securities Act. The book also offers a detailed overview of topics within the Australian Consumer Law that are now relevant to commercial practice such as unconscionable conduct, consumer guarantees, and misleading and deceptive conduct. Written in a clear and accessible style, each chapter features key points and further reading to enhance students' understanding. Significant cases are discussed in detail and include excerpts from judgments to illustrate points of law. Australian Commercial Law is an indispensable resource for students who are seeking a comprehensive understanding of commercial law.

A hybrid of modified PSO and local search on a multi-robot search system

Particle swarm optimization (PSO), a new population based algorithm, has recently been used on multi-robot systems. Although this algorithm is applied to solve many optimization problems as well as multi-robot systems, it has some drawbacks when it is applied on multi-robot search systems to find a target in a search space containing big static obstacles. One of these defects is premature convergence. This means that one of the properties of basic PSO is that when particles are spread in a search space, as time increases they tend to converge in a small area. This shortcoming is also evident on a multi-robot search system, particularly when there are big static obstacles in the search space that prevent the robots from finding the target easily; therefore, as time increases, based on this property they converge to a small area that may not contain the target and become entrapped in that area.Another shortcoming is that basic PSO cannot guarantee the global convergence of the algorithm. In other words, initially particles explore different areas, but in some cases they are not good at exploiting promising areas, which will increase the search time.This study proposes a method based on the particle swarm optimization (PSO) technique on a multi-robot system to find a target in a search space containing big static obstacles. This method is not only able to overcome the premature convergence problem but also establishes an efficient balance between exploration and exploitation and guarantees global convergence, reducing the search time by combining with a local search method, such as A-star.To validate the effectiveness and usefulness of algorithms,a simulation environment has been developed for conducting simulation-based experiments in different scenarios and for reporting experimental results. These experimental results have demonstrated that the proposed method is able to overcome the premature convergence problem and guarantee global convergence.

Active learning: A step towards automating medical concept extraction

Objective This paper presents an automatic active learning-based system for the extraction of medical concepts from clinical free-text reports. Specifically, (1) the contribution of active learning in reducing the annotation effort, and (2) the robustness of incremental active learning framework across different selection criteria and datasets is determined. Materials and methods The comparative performance of an active learning framework and a fully supervised approach were investigated to study how active learning reduces the annotation effort while achieving the same effectiveness as a supervised approach. Conditional Random Fields as the supervised method, and least confidence and information density as two selection criteria for active learning framework were used. The effect of incremental learning vs. standard learning on the robustness of the models within the active learning framework with different selection criteria was also investigated. Two clinical datasets were used for evaluation: the i2b2/VA 2010 NLP challenge and the ShARe/CLEF 2013 eHealth Evaluation Lab. Results The annotation effort saved by active learning to achieve the same effectiveness as supervised learning is up to 77%, 57%, and 46% of the total number of sequences, tokens, and concepts, respectively. Compared to the Random sampling baseline, the saving is at least doubled. Discussion Incremental active learning guarantees robustness across all selection criteria and datasets. The reduction of annotation effort is always above random sampling and longest sequence baselines. Conclusion Incremental active learning is a promising approach for building effective and robust medical concept extraction models, while significantly reducing the burden of manual annotation.

Educational research and strategies of world-forming: The globe, the unconscious, and the child

The passion to eradicate alterity from the earth is also the passion for the home, the country, the dwelling, that authorizes this desire and rewards it. In its nationalism, parochialism and racism it constitutes a public and private neurosis. So, unwinding the rigid understanding of place that apparently permits me to speak, that guarantees my voice, my power, is not simply to disperse my locality within the wider coordinates of an ultimate planetary context. That would merely absolve me of responsibility in the name of an abstract and generic globalism, permitting my inheritance to continue uninterrupted in the vagaries of a new configuration. There is something altogether more precise and more urgent involved. For in the horror of the unhomely pulses the dread for the dispersal of Western humankind: the dread of a rationality confronted with what exceeds and slips its grasp. (Chambers, 2001, p. 196)

CCD: Connected Component Descriptor for Robust Mosaicing of Camera-Captured Document Images

We propose a robust method for mosaicing of document images using features derived from connected components. Each connected component is described using the Angular Radial Tran. form (ART). To ensure geometric consistency during feature matching, the ART coefficients of a connected component are augmented with those of its two nearest neighbors. The proposed method addresses two critical issues often encountered in correspondence matching: (i) The stability of features and (ii) Robustness against false matches due to the multiple instances of characters in a document image. The use of connected components guarantees a stable localization across images. The augmented features ensure a successful correspondence matching even in the presence of multiple similar regions within the page. We illustrate the effectiveness of the proposed method on camera captured document images exhibiting large variations in viewpoint, illumination and scale.

G(its)(2) VSR: An Information Theoretic Secure Verifiable Secret Redistribution Protocol for Long-term Archival Storage

Protocols for secure archival storage are becoming increasingly important as the use of digital storage for sensitive documents is gaining wider practice. Wong et al.[8] combined verifiable secret sharing with proactive secret sharing without reconstruction and proposed a verifiable secret redistribution protocol for long term storage. However their protocol requires that each of the receivers is honest during redistribution. We proposed[3] an extension to their protocol wherein we relaxed the requirement that all the recipients should be honest to the condition that only a simple majority amongst the recipients need to be honest during the re(distribution) processes. Further, both of these protocols make use of Feldman's approach for achieving integrity during the (redistribution processes. In this paper, we present a revised version of our earlier protocol, and its adaptation to incorporate Pedersen's approach instead of Feldman's thereby achieving information theoretic secrecy while retaining integrity guarantees.

Refinement of defection strategies stabilizes cooperation

Cooperation among unrelated individuals is an enduring evolutionary riddle and a number of possible solutions have been suggested. Most of these suggestions attempt to refine cooperative strategies, while little attention is given to the fact that novel defection strategies can also evolve in the population. Especially in the presence of punishment to the defectors and public knowledge of strategies employed by the players, a defecting strategy that avoids getting punished by selectively cooperating only with the punishers can get a selective benefit over non-conditional defectors. Furthermore, if punishment ensures cooperation from such discriminating defectors, defectors who punish other defectors can evolve as well. We show that such discriminating and punishing defectors can evolve in the population by natural selection in a Prisoner’s Dilemma game scenario, even if discrimination is a costly act. These refined defection strategies destabilize unconditional defectors. They themselves are, however, unstable in the population. Discriminating defectors give selective benefit to the punishers in the presence of non-punishers by cooperating with them and defecting with others. However, since these players also defect with other discriminators they suffer fitness loss in the pure population. Among the punishers, punishing cooperators always benefit in contrast to the punishing defectors, as the latter not only defect with other punishing defectors but also punish them and get punished. As a consequence of both these scenarios, punishing cooperators get stabilized in the population. We thus show ironically that refined defection strategies stabilize cooperation. Furthermore, cooperation stabilized by such defectors can work under a wide range of initial conditions and is robust to mistakes.

Latitudinal and seasonal effects on growth of the Australian eastern king prawn (Melicertus plebejus)

The growth of the Australian eastern king prawn (Melicertus plebejus) is understood in greater detail by quantifying the latitudinal effect. The latitudinal effect is the change in the species' growth rate during migration. Mark-recapture data (N = 1635, latitude 22.21 degrees S-34.00 degrees S) presents northerly movement of the eastern king prawn, with New South Wales prawns showing substantial average movement of 140 km (standard deviation: 176 km) north. A generalized von Bertalanffy growth model framework is used to incorporate the latitudinal effect together with the canonical seasonal effect. Applying this method to eastern king prawn mark-recapture data guarantees consistent estimates for the latitudinal and seasonal effects. For M. plebejus, it was found that growth rate peaks on 25 and 29 January for males and females, respectively; is at a minimum on 27 and 31 July, respectively; and that the shape parameter, k (per year), changes by -0.0236 and -0.0556 every 1 degree of latitude south increase for males and females, respectively.

Paraxial-wave optics and relativistic front description .1. The scalar theory

With the extension of the work of the preceding paper, the relativistic front form for Maxwell's equations for electromagnetism is developed and shown to be particularly suited to the description of paraxial waves. The generators of the Poincaré group in a form applicable directly to the electric and magnetic field vectors are derived. It is shown that the effect of a thin lens on a paraxial electromagnetic wave is given by a six-dimensional transformation matrix, constructed out of certain special generators of the Poincaré group. The method of construction guarantees that the free propagation of such waves as well as their transmission through ideal optical systems can be described in terms of the metaplectic group, exactly as found for scalar waves by Bacry and Cadilhac. An alternative formulation in terms of a vector potential is also constructed. It is chosen in a gauge suggested by the front form and by the requirement that the lens transformation matrix act locally in space. Pencils of light with accompanying polarization are defined for statistical states in terms of the two-point correlation function of the vector potential. Their propagation and transmission through lenses are briefly considered in the paraxial limit. This paper extends Fourier optics and completes it by formulating it for the Maxwell field. We stress that the derivations depend explicitly on the "henochromatic" idealization as well as the identification of the ideal lens with a quadratic phase shift and are heuristic to this extent.

Nuori konservatiivi : Tuure Junnila ja hänen poliittinen toimintansa vuoteen 1956

Tuure Junnila, PhD (1910-1999) was one of Finland's most renowned conservative politicians of the post-war period. Junnila is remembered primarily as a persistent opponent of Urho Kekkonen, a long-term Member of Parliament, a conspicuous opposition member and a prolific political writer. Junnila's ideologies and political views were conservative, and he is one of the most outstanding figures in the history of the National Coalition Party. Junnila also made an extensive career outside of politics, first as an economist and then as an executive of Finland's leading commercial bank Kansallis-Osake-Pankki. The Young Conservative is a partial biography written using traditional historical research methods, which examines Junnila's personal history and his activity in public life up to 1956. The study begins by investigating Junnila's background through his childhood, school years, university studies and early professional career. It also looks at Junnila's work as an economist and practical banker. Particular attention is paid to Junnila's political work, constantly focusing on the following five often overlapping areas: (1) economic policy, (2) domestic policy, (3) foreign and security policy, (4) Junnila and Urho Kekkonen, (5) Junnila, the Coalition Party and Finnish conservatism. In his economic policy, Junnila emphasised the importance of economic stability, opposed socialisation and the growth of public expenditure, defended the free market system and private entrepreneurship, and demanded tax cuts. This policy was very popular within the Coalition Party during the early 1950s, making Junnila the leading conservative economic politician of the time. In terms of domestic policy, Junnila demanded as early as the 1940s that a "third force" should be established in Finland to counterbalance the agrarian and labour parties by uniting conservative and liberal ideologies under the same roof. Foreign and security policy is the area of Junnila's political activity which is most clearly situated after the mid-1950s. However, Junnila's early speeches and writings already show a striving towards the unconditional neutrality modelled by Switzerland and Sweden and a strong emphasis on Finland's right to internal self-determination. Junnila, as did the Coalition Party as a whole, adopted a consistently critical approach towards Urho Kekkonen between 1951 and 1956, but this attitude was not as bluntly negative and all-round antagonistic as many previous studies have implied. Junnila was one of the leading Finnish conservatives of the early 1950s and in all essence his views were analogous to the general alignment of the Coalition Party at the time: conservative in ideology and general policy, and liberal in economic policy.