Matt Tom

• University of Stirling

Autism is a psychiatric/neurological condition in which alterations in social interaction (among other symptoms) are diagnosed by behavioral psychiatric methods. The main goal of this study was to determine how the neural representations and meanings of social concepts (such as to insult) are altered in autism. A second goal was to determine whethe...

The question of whether the neural encodings of objects are similar across different people is one of the key questions in cognitive neuroscience. This article examines the commonalities in the internal representation of objects, as measured with fMRI, across individuals in two complementary ways. First, we examine the commonalities in the internal...

The goal of the study was to identify the neural representation of a noun's meaning in one language based on the neural representation of that same noun in another language. Machine learning methods were used to train classifiers to identify which individual noun bilingual participants were thinking about in one language based solely on their brain...

Automated methods can now extract brain-image coordinates appearing in hundreds of publications in targeted topic areas and then integrate these data to form computational models that classify new brain-image data.

In this work we explore whether the patterns of brain activity associated with thinking about concrete objects are dependent on stimulus presentation format, whether an object is referred to by a written or pictorial form. Multi-voxel pattern analysis methods were applied to brain imaging (fMRI) data to identify the item category associated with br...

We consider the problem of semi-supervised learning to extract categories (e.g., academic fields, athletes) and relations (e.g., PlaysSport(athlete, sport)) from web pages, starting with a handful of labeled training examples of each category or relation, plus hundreds of millions of unlabeled web documents. Semi-supervised training using only a fe...

Comparing factor analysis outcomes with traditional GLM contrasts for taxonomic categories. (0.03 MB DOC)

Locations of the multiple voxel clusters associated with the four factors. Shelter-related voxels are shown in blue, manipulation-related voxels in red, eating-related in green, and word length in yellow. (5.09 MB TIF)

Comparison of the locations of activation in taxonomic-category-based GLM contrasts to the factor locations. (0.12 MB DOC)

Taxonomic-category-specific GLM-derived clusters that have matching factor locations. The clusters that match shelter locations are shown in blue; the cluster that matches one of the manipulation locations is shown in red, and the cluster that matches the word-length location is shown in yellow. (1.20 MB TIF)

This article describes the discovery of a set of biologically-driven semantic dimensions underlying the neural representation of concrete nouns, and then demonstrates how a resulting theory of noun representation can be used to identify simple thoughts through their fMRI patterns. We use factor analysis of fMRI brain imaging data to reveal the biol...

Something important is changing in how we as a society use computers to mine data. In the past decade, machine-learning algorithms have helped to analyze historical data, often revealing trends and patterns too subtle for humans to detect. Examples include mining credit card data to discover activity patterns that suggest fraud, and mining scientif...

A key question regarding the future of the semantic web is “how will we acquire structured information to populate the semantic web on a vast scale?” One approach is to enter this information manually. A second approach is to take advantage of pre-existing databases, and to develop common ontologies, publishing standards, and reward systems to make...

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 w...

Recent advances in functional Magnetic Resonance Imaging (fMRI) offer a significant new approach to studying semantic represen- tations in humans by making it possible to di- rectly observe brain activity while people comprehend words and sentences. In this study, we investigate how humans compre- hend adjective-noun phrases (e.g. strong dog) while...

We report research toward a never-ending language learning system, focusing on a first implementation which learns to classify occurrences of noun phrases according to lexical categories such as "city" and "uni- versity." Our experiments suggest that the accuracy of classifiers produced by semi-supervised learning can be improved by coupling the le...

We consider semi-supervised learning of information extraction methods, especially for extracting instances of noun categories (e.g., 'athlete,' 'team') and relations (e.g., 'playsForTeam(athlete,team)'). Semi- supervised approaches using a small number of labeled examples together with many un- labeled examples are often unreliable as they frequen...

The question of how the human brain represents conceptual knowledge has been debated in many scientific fields. Brain imaging studies have shown that different spatial patterns of neural activation are associated with thinking about different semantic categories of pictures and words (for example, tools, buildings, and animals). We present a comput...

We study methods of efficiently leveraging massive textual corpora through n-gram statistics. Specifically, we explore algorithms that use a database of frequency counts for sequences of tokens in a teraword Web corpus to correct spelling mistakes and to extract a list of instances of some category given only the name of the target category. For sp...

Previous studies have succeeded in identifying the cognitive state corresponding to the perception of a set of depicted categories, such as tools, by analyzing the accompanying pattern of brain activity, measured with fMRI. The current research focused on identifying the cognitive state associated with a 4s viewing of an individual line drawing (1...

Identification accuracies of object categories based on the patterns of functional activity of that or other participants. Observed accuracies, number of voxels, and the p-value based on permutation distribution with 1,000 permutations are reported. (0.04 MB DOC)

Identification accuracies of object exemplars based on the patterns of functional activity of that or other participants. Observed accuracies, number of voxels, and the p-value based on permutation distribution with 1,000 permutations are reported. (0.04 MB DOC)

Organizing data into hierarchies is natural for humans. However, there is little work in machine learning that ex- plores human-machine mixed-initiative approaches to orga- nizing data into hierarchical clusters. In this paper we con- sider mixed-initiative clustering of a user's email, in which the machine produces (initial and re-trained) hierarc...

Machine Learning techniques have been used quite widely for the task of predicting cognitive processes from fMRI data. However, these models do not describe well the fMRI signal when it is generated by multiple cognitive processes that are simultaneously active. In this paper we consider the problem of accurately modeling the fMRI signal of a human...

The task of learning models for many real-world problems requires incorporating domain knowl- edge into learning algorithms, to enable accurate learning from a realistic volume of training data. Domain knowledge can come in many forms. For example, expert knowledge about the relevance of variables relative to a certain problem can help per- form be...

Mixed-initiative clustering is a machine learning task that integrates a machine's clustering capability and a user's guidance in order to obtain the user's desired result. This task is different from traditional autonomous clustering tasks by introducing user crite-rion, a user's understanding of data and pur-pose of sorting. We propose a framewor...

A long-standing goal of AI is the development of intelligent workstation-based personal agents to assist users in their daily lives. A key impediment to this goal is the unrealistic cost of developing and maintaining a detailed knowledge base describing the user's different activities, and which people, meetings, emails, etc. are affiliated with ea...

The task of learning models for many real-world problems requires incorporating do- main knowledge into learning algorithms, to enable accurate learning from a realistic vol- ume of training data. This paper considers a variety of types of domain knowledge for constraining parameter estimates when learning Bayesian Networks. In particular, we consi...

We introduce Hidden Process Models (HPMs), a class of probabilistic models for multivariate time series data. The design of HPMs has been motivated by the challenges of modeling hidden cognitive processes in the brain, given functional Magnetic Resonance Imaging (fMRI) data. fMRI data is sparse, high-dimensional, non-Markovian, and often involves p...

So far, climate change mitigation pathways focus mostly on CO2 and a limited number of climate targets. Comprehensive studies of emission implications have been hindered by the absence of a flexible method to generate multi-gas emissions pathways, user-definable in shape and the climate target. The presented method ‘Equal Quantile Walk’ (EQW) is in...

Building accurate models from a small amount of available training data can sometimes prove to be a great challenge. Expert domain knowledge can often be used to alleviate this burden. Parameter Sharing is one such important form of domain knowledge. Graphical models like HMMs, DBNs and Module Networks use difierent forms of Parameter Shar- ing to...

One key to providing intelligent assistance to workstation users is to construct machine-understandable descriptions of the user's ongoing projects, or activities, (e.g., their committee memberships, writing projects, conference organization activities), and indices describing which emails, meetings, and colleagues relate to which activity. This pa...

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 subject's brain activity every half second, at a spatial resolution of a few millimeter...

We consider learning to classify cognitive states of human subjects, based on their brain activity observed via functional Magnetic Resonance Imaging (fMRI). This problem is important because such classifiers constitute "virtual sensors" of hidden cognitive states, which may be useful in cognitive science research and clinical applications. In rece...

Assume a uniform, multidimensional grid of bivariate data, where each cell of the grid has a count ci and a baseline bi. Our goal is to find spatial regions (d-dimensional rectangles) where the ci are significantly higher than expected given bi. We focus on two applications: detection of clusters of disease cases from epidemiological data (emergenc...

We consider the problem of detecting the instantaneous cognitive state of a human subject based on their observed functional Magnetic Resonance Imaging (fMRI) data. Whereas fMRI has been widely used to determine average activation in different brain regions, our problem of automatically decoding instantaneous cognitive states has received little at...

We consider learning to classify cognitive states of human subjects, based on their brain activity observed via functional Magnetic Resonance Imaging (fMRI). This problem is important because such classifiers con- stitute "virtual sensors" of hidden cognitive states, which may be useful in cognitive science research and clinical applications. In re...

Is it feasible to train cross-subject classifiers to decode the cognitive states of human subjects based on functional Magnetic Resonance Imag-ing (fMRI) data observed over a single time interval? If so, these trained classifiers could be used as virtual sensors to detect cognitive states that apply across multiple human subjects. This problem is r...

We consider learning to classify cognitive states of human subjects, based on their brain activity observed via functional Magnetic Resonance Imaging (fMRI). This problem is important because such classifiers con- stitute "virtual sensors" of hidden cognitive states, which may be useful in cognitive science research and clinical applications. In re...

To evaluate a diagnostic protocol incorporating helical computed tomographic pulmonary angiography (CTPA) and lung perfusion scintigraphy in the detection or exclusion of pulmonary embolism (PE) in routine clinical practice. A prospective observational study of 808 consecutive patients with suspected acute PE was undertaken over a 23-month period....

We present a method for distinguishing two subtly different mental states, on the basis of the underlying brain activation measured with fMRI. The method uses a classifier to learn to distinguish between brain activation in a set of selected voxels (volume elements) during the processing of two types of sentences, namely ambiguous versus un- ambigu...

Methods Participants were presented with four nouns, eight two-word sentences, and two verbs, with seven repetitions each. Participants were instructed to think of the same properties consistently at each presentation of a word or a sentence. Functional images were recorded on a Siemens 3T Allegra scanner. Each stimulus was presented for 3s, follow...

We use Hidden Process Models (HPMs) to evaluate different models of a functional Magnetic Resonance Imaging (fMRI) study in which sub- jects decide whether stimuli match. We demonstrate the ability of HPMs to simultaneously estimate the hemodynamic response functions and the onset times of a set of cognitive processes underlying an fMRI time se- ri...

In this paper we propose a method to automatically find useful abstrac- tions of the fMRI data using a new neural network clustering technique. The purpose of these data abstractions is to alleviate the computational burden by reducing dimensionality, to minimize the risk of overfitting by reducing the number of free model parameters, and to uncove...

We present an approach to integrate multiple fMRI datasets in the context of predictive fMRI data analysis. The approach utilizes canonical correlation analysis (CCA) to find common dimensions among the different datasets, and it does not require that the multiple fMRI datasets be spatially normalized. We apply the approach to the task of predictin...

Thesis (Ph. D.)--University of Washington. Bibliography: l. 299-308.