Rebecca A. Hutchinson

Learning to Decode Cognitive States from Brain Images

Over the past decade, functional Magnetic Resonance Imaging (fMRI) has emerged as a powerful new instrument to collect vast quantities of data about activity in the human brain. A typical fMRI experiment can produce a three-dimensional image related to the human subjects brain activity every half second, at a spatial resolution of a few millimeters. As in other modern empirical sciences, this new instrumentation has led to a flood of new data, and a corresponding need for new data analysis methods. We describe recent research applying machine learning methods to the problem of classifying the cognitive state of a human subject based on fRMI data observed over a single time interval. In particular, we present case studies in which we have successfully trained classifiers to distinguish cognitive states such as (1) whether the human subject is looking at a picture or a sentence, (2) whether the subject is reading an ambiguous or non-ambiguous sentence, and (3) whether the word the subject is viewing is a word describing food, people, buildings, etc. This learning problem provides an interesting case study of classifier learning from extremely high dimensional (105 features), extremely sparse (tens of training examples), noisy data. This paper summarizes the results obtained in these three case studies, as well as lessons learned about how to successfully apply machine learning methods to train classifiers in such settings.

Modeling fMRI data generated by overlapping cognitive processes with unknown onsets using Hidden Process Models

We present a new method for modeling fMRI time series data called Hidden Process Models (HPMs). Like several earlier models for fMRI analysis, Hidden Process Models assume that the observed data is generated by a sequence of underlying mental processes that may be triggered by stimuli. HPMs go beyond these earlier models by allowing for processes whose timing may be unknown, and that might not be directly tied to specific stimuli. HPMs provide a principled, probabilistic framework for simultaneously learning the contribution of each process to the observed data, as well as the timing and identities of each instantiated process. They also provide a framework for evaluating and selecting among competing models that assume different numbers and types of underlying mental processes. We describe the HPM framework and its learning and inference algorithms, and present experimental results demonstrating its use on simulated and real fMRI data. Our experiments compare several models of the data using cross-validated data log-likelihood in an fMRI study involving overlapping mental processes whose timings are not fully known.

Modeling Experts and Novices in Citizen Science Data for Species Distribution Modeling

Citizen scientists, who are volunteers from the community that participate as field assistants in scientific studies [3], enable research to be performed at much larger spatial and temporal scales than trained scientists can cover. Species distribution modeling [6], which involves understanding species-habitat relationships, is a research area that can benefit greatly from citizen science. The eBird project [16] is one of the largest citizen science programs in existence. By allowing birders to upload observations of bird species to an online database, eBird can provide useful data for species distribution modeling. However, since birders vary in their levels of expertise, the quality of data submitted to eBird is often questioned. In this paper, we develop a probabilistic model called the Occupancy-Detection-Expertise (ODE) model that incorporates the expertise of birders submitting data to eBird. We show that modeling the expertise of birders can improve the accuracy of predicting observations of a bird species at a site. In addition, we can use the ODE model for two other tasks: predicting birder expertise given their history of eBird checklists and identifying bird species that are difficult for novices to detect.

The macroecology of infectious diseases: a new perspective on global-scale drivers of pathogen distributions and impacts

Identifying drivers of infectious disease patterns and impacts at the broadest scales of organisation is one of the most crucial challenges for modern science, yet answers to many fundamental questions remain elusive. These include what factors commonly facilitate transmission of pathogens to novel host species, what drives variation in immune investment among host species, and more generally what drives global patterns of parasite diversity and distribution? Here we consider how the perspectives and tools of macroecology, a field that investigates patterns and processes at broad spatial, temporal and taxonomic scales, are expanding scientific understanding of global infectious disease ecology. In particular, emerging approaches are providing new insights about scaling properties across all living taxa, and new strategies for mapping pathogen biodiversity and infection risk. Ultimately, macroecology is establishing a framework to more accurately predict global patterns of infectious disease distribution and emergence.

Penalized likelihood methods improve parameter estimates in occupancy models

Summary Occupancy models are employed in species distribution modelling to account for imperfect detection during field surveys. While this approach is popular in the literature, problems can occur when estimating the model parameters. In particular, the maximum likelihood estimates can exhibit bias and large variance for data sets with small sample sizes, which can result in estimated occupancy probabilities near 0 and 1 (‘boundary estimates’). In this paper, we explore strategies for estimating parameters based on maximizing a penalized likelihood. Penalized likelihood methods augment the usual likelihood with a penalty function that encodes information about what parameter values are undesirable. We introduce penalties for occupancy models that have analogues in ridge regression and Bayesian approaches, and we compare them to a penalty developed for occupancy models in prior work. We examine the bias, variance and mean squared error of parameter estimates obtained from each method on synthetic data. Across all of the synthetic data sets, the penalized estimation methods had lower mean squared error than the maximum likelihood estimates. We also provide an example of the application of these methods to point counts of avian species. Penalized likelihood methods show similar improvements when tested using empirical bird point count data. We discuss considerations for choosing among these methods when modelling occupancy. We conclude that penalized methods may be of practical utility for fitting occupancy models with small sample sizes, and we are releasing R code that implements these methods.

Distinguishing distribution dynamics from temporary emigration using dynamic occupancy models

1. Dynamic occupancy models are popular for estimating dynamic distribution rates (colonization and extinction) from repeated presence/absence surveys of unmarked animals. This approach assumes closure among repeated samples within primary periods, allowing estimation of dynamic rates between these periods. However, the impact of temporary emigration (reversible changes in sampling availability) on dynamic rate estimates, has not been tested. 2. Using simulated data, we investigated the degree to which temporary emigration could mislead researchers interested in quantifying dynamics. We then compared results from three avian point count datasets to evaluate the likelihood that temporary emigration confounds estimates of dynamics for 19 species under a popular sampling protocol. 3. Simulated experiments indicated that when secondary periods were open to temporary emigration, presence of dynamics was correctly identified ≥ 95.1% of the time, and dynamic rate estimates were accurate. However, dynamic rate estimates were biased when secondary periods were closed to temporary emigration. In empirical datasets, dynamic occupancy models had greater support than closed models for all species when secondary sampling periods occurred in immediate succession (i.e., 3 samples within 10 minutes); however, our results suggest that this is because these estimates were heavily influenced by temporary emigration. When counts within a primary period were separated by 24-48 hours, we found evidence of dynamics for less than half of these species. We recommend an alternative sampling approach that allows accurate estimation of dynamic rates when temporary emigration is of no interest, and introduce a novel model for estimating both processes simultaneously in rare cases where they are both of biological interest. 4. Concern for violating the occupancy modeling closure assumption has led to widespread recommendations that samples within primary periods be conducted extremely close in time. However, this may not be the best approach when interest is in quantifying dynamic rates. While dynamic occupancy models provide estimates of ‘colonization’ and ‘extinction,’ these values do not inherently represent dynamics unless temporary emigration has been explicitly modeled, or accounted for with sampling design. Naivete to this fact can result in incorrect conclusions about biological processes. This article is protected by copyright. All rights reserved.

eButterfly: Leveraging Massive Online Citizen Science for Butterfly Conservation

Data collection, storage, analysis, visualization, and dissemination are changing rapidly due to advances in new technologies driven by computer science and universal access to the internet. These technologies and web connections place human observers front and center in citizen science-driven research and are critical in generating new discoveries and innovation in such fields as astronomy, biodiversity, and meteorology. Research projects utilizing a citizen science approach address scientific problems at regional, continental, and even global scales otherwise impossible for a single lab or even a small collection of academic researchers. Here we describe eButterfly an integrative checklist-based butterfly monitoring and database web-platform that leverages the skills and knowledge of recreational butterfly enthusiasts to create a globally accessible unified database of butterfly observations across North America. Citizen scientists, conservationists, policy makers, and scientists are using eButterfly data to better understand the biological patterns of butterfly species diversity and how environmental conditions shape these patterns in space and time. eButterfly in collaboration with thousands of butterfly enthusiasts has created a near real-time butterfly data resource producing tens of thousands of observations per year open to all to share and explore.

Machine learning for computational sustainability

To avoid ecological collapse, we must manage Earths ecosystems sustainably. Viewed as a control problem, the two central challenges of ecosystem management are to acquire a model of the system that is sufficient to guide good decision making and then optimize the control policy against that model. This paper describes three efforts aimed at addressing the first of these challenges-machine learning methods for modeling ecosystems. The first effort focuses on automated quality control of environmental sensor data. Next, we consider the problem of learning species distribution models from citizen science observational data. Finally, we describe a novel approach to modeling the migration of birds. A major challenge for all of these methods is to scale up to large, spatially-distributed systems.