Search

The combination of large open data sources with machine learning approaches presents a potentially powerful way to predict events such as protest or social unrest. However, accounting for uncertainty in such models, particularly when using diverse, unstructured datasets such as social media, is essential to guarantee the appropriate use of such methods. Here we develop a Bayesian method for predicting social unrest events in Australia using social media data. This method uses machine learning methods to classify individual postings to social media as being relevant, and an empirical Bayesian approach to calculate posterior event probabilities. We use the method to predict events in Australian cities over a period in 2017/18.

Hex is a challenging strategy board game for two players. To enhance students’ progress in acquiring understanding and practical experience with complex machine intelligence and programming concepts we developed the Machine Intelligence Hex (MIHex) project. The associated undergraduate student assignment is about designing and implementing Hex players and evaluating them in an automated tournament of all programs developed by the class. This article surveys educational aspects of the MIHex project. Additionally, fundamental techniques for game programming as well as specific concepts for Hex board evaluation are reviewed. The MIHex game server and possibilities of tournament organisation are described. We summarise and discuss our experiences from running the MIHex project assignment over four consecutive years. The impact on student motivation and learning benefits are evaluated using questionnaires and interviews.

Therapies consisting of a combination of agents are an attractive proposition, especially in the context of diseases such as cancer, which can manifest with a variety of tumor types in a single case. However uncovering usable drug combinations is expensive both financially and temporally. By employing computational methods to identify candidate combinations with a greater likelihood of success we can avoid these problems, even when the amount of data is prohibitively large. HITTING SET is a combinatorial problem that has useful application across many fields, however as it is NP-complete it is traditionally considered hard to solve exactly. We introduce a more general version of the problem (α,β,d)-HITTING SET, which allows more precise control over how and what the hitting set targets. Employing the framework of Parameterized Complexity we show that despite being NP-complete, the ((α,β,d)-HITTING SET problem is fixed-parameter tractable with a kernel of size O(αdkd) when we parameterize by the size k of the hitting set and the maximum number a of the minimum number of hits, and taking the maximum degree d of the target sets as a constant. We demonstrate the application of this problem to multiple drug selection for cancer therapy, showing the flexibility of the problem in tailoring such drug sets. The fixed-parameter tractability result indicates that for low values of the parameters the problem can be solved quickly using exact methods. We also demonstrate that the problem is indeed practical, with computation times on the order of 5 seconds, as compared to previous Hitting Set applications using the same dataset which exhibited times on the order of 1 day, even with relatively relaxed notions for what constitutes a low value for the parameters. Furthermore the existence of a kernelization for ((α,β,d)-HITTING SET indicates that the problem is readily scalable to large datasets. © 2010 Mellor et al.

Profiling of the learning classifier system XCS [11] has revealed that its execution time tends to be dominated by rule matching [8], it is therefore important for rule matching to be efficient. To date, the fastest speedups for matching have been achieved by exploiting parallelism [8], but efficient sequential approaches, such as bitset and "specificity" matching [2], can be utilised if there is no platform support for the vector instruction sets that [8] employs. Previous sequential approaches have focussed on improving the efficiency of matching individual rules; in this paper, we introduce a population-based approach that partially matches many rules simultaneously. This is achieved by maintaining the rule-base in a rooted 3-ary tree over which a backtracking depth-first search is run to find the matchset. We found that the method generally outperformed standard and specificity matching on raw matching and on several benchmarking tasks. While the bitset approach attained the best speedups on the benchmarking tasks, we give an analysis that shows that it can be the least efficient of the approaches on long rule conditions. A limitation of the new method is that it is inefficient when the proportion of "don't care" symbols in the rule conditions is very large, which could perhaps be remedied by combining the method with the specificity technique.

This article describes a learning classifier system (LCS) approach to relational reinforcement learning (RRL). The system, Foxcs-2 , is a derivative of Xcs that learns rules expressed as definite clauses over first-order logic. By adopting the LCS approach, Foxcs-2 , unlike many RRL systems, is a general, model-free and "tabula rasa" system. The change in representation from bit-strings in Xcs to first-order logic in Foxcs-2 necessitates modifications, described within, to support matching, covering, mutation and several other functions. Evaluation on inductive logic programming (ILP) and RRL tasks shows that the performance of Foxcs-2 is comparable to other systems. Further evaluation on RRL tasks highlights a significant advantage of Foxcs-2 's rule language: in some environments it is able to represent policies that are genuinely scalable; that is, policies that are independent of the size of the environment.

A methodology for validating fire spread simulation systems using historical fire data is presented. The key features of this methodology are (a) quantitative comparison between simulator-generated fire perimeters and fire perimeters from an independently produced fire reconstruction at multiple time points during the fire, and (b) extensive sensitivity analyses on simulation variables including simulation spatial resolution, weather, vegetation coverage and fire behaviour model selection to determine the effect of each simulation input on the simulation output. The methodology is demonstrated in a case study in which the ability of the Australis high-performance wildfire simulator to replicate a large wildfire in Western Australia was examined. Simulation accuracy was found to be lower in extreme fire danger conditions and exhibited under-prediction of the head fire rate of spread. This was caused by inaccuracies in at least one of wind speed data, vegetation data or the fire behaviour model applied; however, the source of the inaccuracy could not be further diagnosed with the available data. The gathering of accurate data during and after active wildfires would facilitate more rigorous simulator and fire behaviour model validation studies as well as more accurate prediction of ‘live’ wildfires.

BackgroundSeveral lines of evidence suggest that transcription factors are involved in the pathogenesis of Multiple Sclerosis (MS) but complete mapping of the whole network has been elusive. One of the reasons is that there are several clinical subtypes of MS and transcription factors that may be involved in one subtype may not be in others. We investigate the possibility that this network could be mapped using microarray technologies and contemporary bioinformatics methods on a dataset derived from whole blood in 99 untreated MS patients (36 Relapse Remitting MS, 43 Primary Progressive MS, and 20 Secondary Progressive MS) and 45 age-matched healthy controls.Methodology/principal findingsWe have used two different analytical methodologies: a non-standard differential expression analysis and a differential co-expression analysis, which have converged on a significant number of regulatory motifs that are statistically overrepresented in genes that are either differentially expressed (or differentially co-expressed) in cases and controls (e.g., V$KROX_Q6, p-value Conclusions/significanceOur analysis uncovered a network of transcription factors that potentially dysregulate several genes in MS or one or more of its disease subtypes. The most significant transcription factor motifs were for the Early Growth Response EGR/KROX family, ATF2, YY1 (Yin and Yang 1), E2F-1/DP-1 and E2F-4/DP-2 heterodimers, SOX5, and CREB and ATF families. These transcription factors are involved in early T-lymphocyte specification and commitment as well as in oligodendrocyte dedifferentiation and development, both pathways that have significant biological plausibility in MS causation.