In recent years, many studies have been published on data collected from social media, especially microblogs such as Twitter. However, rather few of these studies have considered evaluation methodologies that take into account the statistically dependent nature of such data, which breaks the theoretical conditions for using cross-validation. Despite concerns raised in the past about using cross-validation for data of similar characteristics, such as time series, some of these studies evaluate their work using standard k-fold cross-validation. Through experiments on Twitter data collected during a two-year period that includes disastrous events, we show that by ignoring the statistical dependence of the text messages published in social media, standard cross-validation can result in misleading conclusions in a machine learning task. We explore alternative evaluation methods that explicitly deal with statistical dependence in text. Our work also raises concerns for any other data for which similar conditions might hold.

In a deterministic world, a planning agent can be certain of the consequences of its planned sequence of actions. Not so, however, in dynamic, stochastic domains where Markov decision processes are commonly used. Unfortunately these suffer from the `curse of dimensionality': if the state space is a Cartesian product of many small sets (`dimensions'), planning is exponential in the number of those dimensions. Our new technique exploits the intuitive strategy of selectively ignoring various dimensions in different parts of the state space. The resulting non-uniformity has strong implications, since the approximation is no longer Markovian, requiring the use of a modified planner. We also use a spatial and temporal proximity measure, which responds to continued planning as well as movement of the agent through the state space, to dynamically adapt the abstraction as planning progresses. We present qualitative and quantitative results across a range of experimental domains showing that an agent exploiting this novel approximation method successfully finds solutions to the planning problem using much less than the full state space. We assess and analyse the features of domains which our method can exploit.

Habitat connectivity is required at large spatial scales to facilitate movement of biota in response to climatic changes and to maintain viable populations of wide-ranging species. Nevertheless, it may require decades to acquire habitat linkages at such scales, and areas that could provide linkages are often developed before they can be reserved. Reserve scheduling methods usually consider only current threats, but threats change over time as development spreads and reaches presently secure areas. We investigated the importance of considering future threats when implementing projects to maintain habitat connectivity at a regional scale. To do so, we compared forward-looking scheduling strategies with strategies that consider only current threats. The strategies were applied to a Costa Rican case study, where many reserves face imminent isolation and other reserves will probably become isolated in the more distant future. We evaluated strategies in terms of two landscape-scale connectivity metrics, a pure connectivity metric and a metric of connected habitat diversity. Those strategies that considered only current threats were unreliable because they often failed to complete planned habitat linkage projects. The most reliable and effective strategies considered the future spread of development and its impact on the likelihood of completing planned habitat linkage projects. Our analyses highlight the critical need to consider future threats when building connected reserve networks over time.

Habitat connectivity is required at large spatial scales to facilitate movement of biota in response to climatic changes and to maintain viable populations of wide-ranging species. Nevertheless, it may require decades to acquire habitat linkages at such scales, and areas that could provide linkages are often developed before they can be reserved. Reserve scheduling methods usually consider only current threats, but threats change over time as development spreads and reaches presently secure areas. We investigated the importance of considering future threats when implementing projects to maintain habitat connectivity at a regional scale. To do so, we compared forward-looking scheduling strategies with strategies that consider only current threats. The strategies were applied to a Costa Rican case study, where many reserves face imminent isolation and other reserves will probably become isolated in the more distant future. We evaluated strategies in terms of two landscape-scale connectivity metrics, a pure connectivity metric and a metric of connected habitat diversity. Those strategies that considered only current threats were unreliable because they often failed to complete planned habitat linkage projects. The most reliable and effective strategies considered the future spread of development and its impact on the likelihood of completing planned habitat linkage projects. Our analyses highlight the critical need to consider future threats when building connected reserve networks over time.

The theory of Markov Decision Processes (MDPs) provides algorithms for generating an optimal policy. For large domains these algorithms become intractable and approximate solutions become necessary. In this paper we extend previous work on approximate planning in large stochastic domains by using automatically-generated non-uniform abstractions which exploit the structure of the state space. We consider a state space expressed as a cross product of sets, or dimensions. We obtain approximate solutions by varying the level of abstraction, selectively ignoring some of the dimensions in some parts of the state space. We describe a modification of a standard policy generation algorithm for the now non-Markovian decision process, which re-calculates values for nearby states based on a locally uniform abstraction for each state. We present methods to automatically generate an initial abstraction based on the domain structure and to automatically modify the non-uniform abstraction. The changes to the abstraction are based on both the current policy and the likelihood of encountering particular states in the future, thereby taking into account the agent's changing circumstances.