A Probabilistic Approach to the Primary Visual Cortex

What can the statistical structure of natural images teach us about the human brain? Even though the visual cortex is one of the most studied parts of the brain, surprisingly little is known about how exactly images are processed to leave us with a coherent percept of the world around us, so we can recognize a friend or drive on a crowded street without any effort. By constructing probabilistic models of natural images, the goal of this thesis is to understand the structure of the stimulus that is the raison d etre for the visual system. Following the hypothesis that the optimal processing has to be matched to the structure of that stimulus, we attempt to derive computational principles, features that the visual system should compute, and properties that cells in the visual system should have. Starting from machine learning techniques such as principal component analysis and independent component analysis we construct a variety of sta- tistical models to discover structure in natural images that can be linked to receptive field properties of neurons in primary visual cortex such as simple and complex cells. We show that by representing images with phase invariant, complex cell-like units, a better statistical description of the vi- sual environment is obtained than with linear simple cell units, and that complex cell pooling can be learned by estimating both layers of a two-layer model of natural images. We investigate how a simplified model of the processing in the retina, where adaptation and contrast normalization take place, is connected to the nat- ural stimulus statistics. Analyzing the effect that retinal gain control has on later cortical processing, we propose a novel method to perform gain control in a data-driven way. Finally we show how models like those pre- sented here can be extended to capture whole visual scenes rather than just small image patches. By using a Markov random field approach we can model images of arbitrary size, while still being able to estimate the model parameters from the data.

Construction of a global map of human gene expression : the process, tools and analysis

This thesis studies human gene expression space using high throughput gene expression data from DNA microarrays. In molecular biology, high throughput techniques allow numerical measurements of expression of tens of thousands of genes simultaneously. In a single study, this data is traditionally obtained from a limited number of sample types with a small number of replicates. For organism-wide analysis, this data has been largely unavailable and the global structure of human transcriptome has remained unknown. This thesis introduces a human transcriptome map of different biological entities and analysis of its general structure. The map is constructed from gene expression data from the two largest public microarray data repositories, GEO and ArrayExpress. The creation of this map contributed to the development of ArrayExpress by identifying and retrofitting the previously unusable and missing data and by improving the access to its data. It also contributed to creation of several new tools for microarray data manipulation and establishment of data exchange between GEO and ArrayExpress. The data integration for the global map required creation of a new large ontology of human cell types, disease states, organism parts and cell lines. The ontology was used in a new text mining and decision tree based method for automatic conversion of human readable free text microarray data annotations into categorised format. The data comparability and minimisation of the systematic measurement errors that are characteristic to each lab- oratory in this large cross-laboratories integrated dataset, was ensured by computation of a range of microarray data quality metrics and exclusion of incomparable data. The structure of a global map of human gene expression was then explored by principal component analysis and hierarchical clustering using heuristics and help from another purpose built sample ontology. A preface and motivation to the construction and analysis of a global map of human gene expression is given by analysis of two microarray datasets of human malignant melanoma. The analysis of these sets incorporate indirect comparison of statistical methods for finding differentially expressed genes and point to the need to study gene expression on a global level.

Exploring privacy for ubiquitous computing: Tools, methods and experiments

Ubiquitous computing is about making computers and computerized artefacts a pervasive part of our everyday lifes, bringing more and more activities into the realm of information. The computationalization, informationalization of everyday activities increases not only our reach, efficiency and capabilities but also the amount and kinds of data gathered about us and our activities. In this thesis, I explore how information systems can be constructed so that they handle this personal data in a reasonable manner. The thesis provides two kinds of results: on one hand, tools and methods for both the construction as well as the evaluation of ubiquitous and mobile systems---on the other hand an evaluation of the privacy aspects of a ubiquitous social awareness system. The work emphasises real-world experiments as the most important way to study privacy. Additionally, the state of current information systems as regards data protection is studied. The tools and methods in this thesis consist of three distinct contributions. An algorithm for locationing in cellular networks is proposed that does not require the location information to be revealed beyond the user's terminal. A prototyping platform for the creation of context-aware ubiquitous applications called ContextPhone is described and released as open source. Finally, a set of methodological findings for the use of smartphones in social scientific field research is reported. A central contribution of this thesis are the pragmatic tools that allow other researchers to carry out experiments. The evaluation of the ubiquitous social awareness application ContextContacts covers both the usage of the system in general as well as an analysis of privacy implications. The usage of the system is analyzed in the light of how users make inferences of others based on real-time contextual cues mediated by the system, based on several long-term field studies. The analysis of privacy implications draws together the social psychological theory of self-presentation and research in privacy for ubiquitous computing, deriving a set of design guidelines for such systems. The main findings from these studies can be summarized as follows: The fact that ubiquitous computing systems gather more data about users can be used to not only study the use of such systems in an effort to create better systems but in general to study phenomena previously unstudied, such as the dynamic change of social networks. Systems that let people create new ways of presenting themselves to others can be fun for the users---but the self-presentation requires several thoughtful design decisions that allow the manipulation of the image mediated by the system. Finally, the growing amount of computational resources available to the users can be used to allow them to use the data themselves, rather than just being passive subjects of data gathering.

Learning Nonlinear Visual Processing from Natural Images

The paradigm of computational vision hypothesizes that any visual function -- such as the recognition of your grandparent -- can be replicated by computational processing of the visual input. What are these computations that the brain performs? What should or could they be? Working on the latter question, this dissertation takes the statistical approach, where the suitable computations are attempted to be learned from the natural visual data itself. In particular, we empirically study the computational processing that emerges from the statistical properties of the visual world and the constraints and objectives specified for the learning process. This thesis consists of an introduction and 7 peer-reviewed publications, where the purpose of the introduction is to illustrate the area of study to a reader who is not familiar with computational vision research. In the scope of the introduction, we will briefly overview the primary challenges to visual processing, as well as recall some of the current opinions on visual processing in the early visual systems of animals. Next, we describe the methodology we have used in our research, and discuss the presented results. We have included some additional remarks, speculations and conclusions to this discussion that were not featured in the original publications. We present the following results in the publications of this thesis. First, we empirically demonstrate that luminance and contrast are strongly dependent in natural images, contradicting previous theories suggesting that luminance and contrast were processed separately in natural systems due to their independence in the visual data. Second, we show that simple cell -like receptive fields of the primary visual cortex can be learned in the nonlinear contrast domain by maximization of independence. Further, we provide first-time reports of the emergence of conjunctive (corner-detecting) and subtractive (opponent orientation) processing due to nonlinear projection pursuit with simple objective functions related to sparseness and response energy optimization. Then, we show that attempting to extract independent components of nonlinear histogram statistics of a biologically plausible representation leads to projection directions that appear to differentiate between visual contexts. Such processing might be applicable for priming, \ie the selection and tuning of later visual processing. We continue by showing that a different kind of thresholded low-frequency priming can be learned and used to make object detection faster with little loss in accuracy. Finally, we show that in a computational object detection setting, nonlinearly gain-controlled visual features of medium complexity can be acquired sequentially as images are encountered and discarded. We present two online algorithms to perform this feature selection, and propose the idea that for artificial systems, some processing mechanisms could be selectable from the environment without optimizing the mechanisms themselves. In summary, this thesis explores learning visual processing on several levels. The learning can be understood as interplay of input data, model structures, learning objectives, and estimation algorithms. The presented work adds to the growing body of evidence showing that statistical methods can be used to acquire intuitively meaningful visual processing mechanisms. The work also presents some predictions and ideas regarding biological visual processing.

Information Visualization in a Business Decision Support System

Information visualization is a process of constructing a visual presentation of abstract quantitative data. The characteristics of visual perception enable humans to recognize patterns, trends and anomalies inherent in the data with little effort in a visual display. Such properties of the data are likely to be missed in a purely text-based presentation. Visualizations are therefore widely used in contemporary business decision support systems. Visual user interfaces called dashboards are tools for reporting the status of a company and its business environment to facilitate business intelligence (BI) and performance management activities. In this study, we examine the research on the principles of human visual perception and information visualization as well as the application of visualization in a business decision support system. A review of current BI software products reveals that the visualizations included in them are often quite ineffective in communicating important information. Based on the principles of visual perception and information visualization, we summarize a set of design guidelines for creating effective visual reporting interfaces.

The visual preferences for forest regeneration and field afforestation - four case studies in Finland

The overall aim of this dissertation was to study the public's preferences for forest regeneration fellings and field afforestations, as well as to find out the relations of these preferences to landscape management instructions, to ecological healthiness, and to the contemporary theories for predicting landscape preferences. This dissertation includes four case studies in Finland, each based on the visualization of management options and surveys. Guidelines for improving the visual quality of forest regeneration and field afforestation are given based on the case studies. The results show that forest regeneration can be connected to positive images and memories when the regeneration area is small and some time has passed since the felling. Preferences may not depend only on the management alternative itself but also on the viewing distance, viewing point, and the scene in which the management options are implemented. The current Finnish forest landscape management guidelines as well as the ecological healthiness of the studied options are to a large extent compatible with the public's preferences. However, there are some discrepancies. For example, the landscape management instructions as well as ecological hypotheses suggest that the retention trees need to be left in groups, whereas people usually prefer individually located retention trees to those trees in groups. Information and psycho-evolutionary theories provide some possible explanations for people's preferences for forest regeneration and field afforestation, but the results cannot be consistently explained by these theories. The preferences of the different stakeholder groups were very similar. However, the preference ratings of the groups that make their living from forest - forest owners and forest professionals - slightly differed from those of the others. These results provide support for the assumptions that preferences are largely consistent at least within one nation, but that knowledge and a reference group may also influence preferences.

Spectral Tuning and Adaptation to Different Light Environments of Mysid Visual Pigments

In the present thesis, questions of spectral tuning, the relation of spectral and thermal properties of visual pigments, and evolutionary adaptation to different light environments were addressed using a group of small crustaceans of the genus Mysis as a model. The study was based on microspectrophotometric measurements of visual pigment absorbance spectra, electrophysiological measurements of spectral sensitivities of dark-adapted eyes, and sequencing of the opsin gene retrieved through PCR. The spectral properties were related to the spectral transmission of the respective light environments, as well as to the phylogentic histories of the species. The photoactivation energy (Ea) was estimated from temperature effects on spectral sensitivity in the long-wavelength range, and calculations were made for optimal quantum catch and optimal signal-to-noise ratio in the different light environments. The opsin amino acid sequences of spectrally characterized individuals were compared to find candidate residues for spectral tuning. The general purpose was to clarify to what extent and on what time scale adaptive evolution has driven the functional properties of (mysid) visual pigments towards optimal performance in different light environments. An ultimate goal was to find the molecular mechanisms underlying the spectral tuning and to understand the balance between evolutionary adaptation and molecular constraints. The totally consistent segregation of absorption maxima (λmax) into (shorter-wavelength) marine and (longer-wavelength) freshwater populations suggests that truly adaptive evolution is involved in tuning the visual pigment for optimal performance, driven by selection for high absolute visual sensitivity. On the other hand, the similarity in λmax and opsin sequence between several populations of freshwater M. relicta in spectrally different lakes highlights the limits to adaptation set by evolutionary history and time. A strong inverse correlation between Ea and λmax was found among all visual pigments studied in these respects, including those of M. relicta and 10 species of vertebrate pigments, and this was used to infer thermal noise. The conceptual signal-to-noise ratios thus calculated for pigments with different λmax in the Baltic Sea and Lake Pääjärvi light environments supported the notion that spectral adaptation works towards maximizing the signal-to-noise ratio rather than quantum catch as such. Judged by the shape of absorbance spectra, the visual pigments of all populations of M. relicta and M. salemaai used exclusively the A2 chromophore (3, 4-dehydroretinal). A comparison of amino acid substitutions between M. relicta and M. salemaai indicated that mysid shrimps have a small number of readily available tuning sites to shift between a shorter - and a longer -wavelength opsin. However, phylogenetic history seems to have prevented marine M. relicta from converting back to the (presumably) ancestral opsin form, and thus the more recent reinvention of marine spectral sensitivity has been accomplished by some other novel mechanism, yet to be found

Functional genes and gene array analysis as tools for monitoring hydrocarbon biodegradation

Bioremediation, which is the exploitation of the intrinsic ability of environmental microbes to degrade and remove harmful compounds from nature, is considered to be an environmentally sustainable and cost-effective means for environmental clean-up. However, a comprehensive understanding of the biodegradation potential of microbial communities and their response to decontamination measures is required for the effective management of bioremediation processes. In this thesis, the potential to use hydrocarbon-degradative genes as indicators of aerobic hydrocarbon biodegradation was investigated. Small-scale functional gene macro- and microarrays targeting aliphatic, monoaromatic and low molecular weight polyaromatic hydrocarbon biodegradation were developed in order to simultaneously monitor the biodegradation of mixtures of hydrocarbons. The validity of the array analysis in monitoring hydrocarbon biodegradation was evaluated in microcosm studies and field-scale bioremediation processes by comparing the hybridization signal intensities to hydrocarbon mineralization, real-time polymerase chain reaction (PCR), dot blot hybridization and both chemical and microbiological monitoring data. The results obtained by real-time PCR, dot blot hybridization and gene array analysis were in good agreement with hydrocarbon biodegradation in laboratory-scale microcosms. Mineralization of several hydrocarbons could be monitored simultaneously using gene array analysis. In the field-scale bioremediation processes, the detection and enumeration of hydrocarbon-degradative genes provided important additional information for process optimization and design. In creosote-contaminated groundwater, gene array analysis demonstrated that the aerobic biodegradation potential that was present at the site, but restrained under the oxygen-limited conditions, could be successfully stimulated with aeration and nutrient infiltration. During ex situ bioremediation of diesel oil- and lubrication oil-contaminated soil, the functional gene array analysis revealed inefficient hydrocarbon biodegradation, caused by poor aeration during composting. The functional gene array specifically detected upper and lower biodegradation pathways required for complete mineralization of hydrocarbons. Bacteria representing 1 % of the microbial community could be detected without prior PCR amplification. Molecular biological monitoring methods based on functional genes provide powerful tools for the development of more efficient remediation processes. The parallel detection of several functional genes using functional gene array analysis is an especially promising tool for monitoring the biodegradation of mixtures of hydrocarbons.

Molecular Tuning of Rhodopsins for High Visual Sensitivity in Different Light Environments : Variation in Absorbance Spectrum and Opsin Sequence within and between Species

Visual pigments of different animal species must have evolved at some stage to match the prevailing light environments, since all visual functions depend on their ability to absorb available photons and transduce the event into a reliable neural signal. There is a large literature on correlation between the light environment and spectral sensitivity between different fish species. However, little work has been done on evolutionary adaptation between separated populations within species. More generally, little is known about the rate of evolutionary adaptation to changing spectral environments. The objective of this thesis is to illuminate the constraints under which the evolutionary tuning of visual pigments works as evident in: scope, tempo, available molecular routes, and signal/noise trade-offs. Aquatic environments offer Nature s own laboratories for research on visual pigment properties, as naturally occurring light environments offer an enormous range of variation in both spectral composition and intensity. The present thesis focuses on the visual pigments that serve dim-light vision in two groups of model species, teleost fishes and mysid crustaceans. The geographical emphasis is in the brackish Baltic Sea area with its well-known postglacial isolation history and its aquatic fauna of both marine and fresh-water origin. The absorbance spectrum of the (single) dim-light visual pigment were recorded by microspectrophotometry (MSP) in single rods of 26 fish species and single rhabdoms of 8 opossum shrimp populations of the genus Mysis inhabiting marine, brackish or freshwater environments. Additionally, spectral sensitivity was determined from six Mysis populations by electroretinogram (ERG) recording. The rod opsin gene was sequenced in individuals of four allopatric populations of the sand goby (Pomatoschistus minutus). Rod opsins of two other goby species were investigated as outgroups for comparison. Rod absorbance spectra of the Baltic subspecies or populations of the primarily marine species herring (Clupea harengus membras), sand goby (P. minutus), and flounder (Platichthys flesus) were long-wavelength-shifted compared to their marine populations. The spectral shifts are consistent with adaptation for improved quantum catch (QC) as well as improved signal-to-noise ratio (SNR) of vision in the Baltic light environment. Since the chromophore of the pigment was pure A1 in all cases, this has apparently been achieved by evolutionary tuning of the opsin visual pigment. By contrast, no opsin-based differences were evident between lake and sea populations of species of fresh-water origin, which can tune their pigment by varying chromophore ratios. A more detailed analysis of differences in absorbance spectra and opsin sequence between and within populations was conducted using the sand goby as model species. Four allopatric populations from the Baltic Sea (B), Swedish west coast (S), English Channel (E), and Adriatic Sea (A) were examined. Rod absorbance spectra, characterized by the wavelength of maximum absorbance (λmax), differed between populations and correlated with differences in the spectral light transmission of the respective water bodies. The greatest λmax shift as well as the greatest opsin sequence difference was between the Baltic and the Adriatic populations. The significant within-population variation of the Baltic λmax values (506-511 nm) was analyzed on the level of individuals and was shown to correlate well with opsin sequence substitutions. The sequences of individuals with λmax at shorter wavelengths were identical to that of the Swedish population, whereas those with λmax at longer wavelengths additionally had substitution F261F/Y in the sixth transmembrane helix of the protein. This substitution (Y261) was also present in the Baltic common gobies and is known to redshift spectra. The tuning mechanism of the long-wavelength type Baltic sand gobies is assumed to be the co-expression of F261 and Y261 in all rods to produce ≈ 5 nm redshift. The polymorphism of the Baltic sand goby population possibly indicates ambiguous selection pressures in the Baltic Sea. The visual pigments of all lake populations of the opossum shrimp (Mysis relicta) were red-shifted by 25 nm compared with all Baltic Sea populations. This is calculated to confer a significant advantage in both QC and SNR in many humus-rich lakes with reddish water. Since only A2 chromophore was present, the differences obviously reflect evolutionary tuning of the visual protein, the opsin. The changes have occurred within the ca. 9000 years that the lakes have been isolated from the Sea after the most recent glaciation. At present, it seems that the mechanism explaining the spectral differences between lake and sea populations is not an amino acid substitution at any other conventional tuning site, but the mechanism is yet to be found.

Mutations as molecular tools : The metabolic-rate dependent molecular clock and DNA barcoding of allied species

Mutation and recombination are the fundamental processes leading to genetic variation in natural populations. This variation forms the raw material for evolution through natural selection and drift. Therefore, studying mutation rates may reveal information about evolutionary histories as well as phylogenetic interrelationships of organisms. In this thesis two molecular tools, DNA barcoding and the molecular clock were examined. In the first part, the efficiency of mutations to delineate closely related species was tested and the implications for conservation practices were assessed. The second part investigated the proposition that a constant mutation rate exists within invertebrates, in form of a metabolic-rate dependent molecular clock, which can be applied to accurately date speciation events. DNA barcoding aspires to be an efficient technique to not only distinguish between species but also reveal population-level variation solely relying on mutations found on a short stretch of a single gene. In this thesis barcoding was applied to discriminate between Hylochares populations from Russian Karelia and new Hylochares findings from the greater Helsinki region in Finland. Although barcoding failed to delineate the two reproductively isolated groups, their distinct morphological features and differing life-history traits led to their classification as two closely related, although separate species. The lack of genetic differentiation appears to be due to a recent divergence event not yet reflected in the beetles molecular make-up. Thus, the Russian Hylochares was described as a new species. The Finnish species, previously considered as locally extinct, was recognized as endangered. Even if, due to their identical genetic make-up, the populations had been regarded as conspecific, conservation strategies based on prior knowledge from Russia would not have guaranteed the survival of the Finnish beetle. Therefore, new conservation actions based on detailed studies of the biology and life-history of the Finnish Hylochares were conducted to protect this endemic rarity in Finland. The idea behind the strict molecular clock is that mutation rates are constant over evolutionary time and may thus be used to infer species divergence dates. However, one of the most recent theories argues that a strict clock does not tick per unit of time but that it has a constant substitution rate per unit of mass-specific metabolic energy. Therefore, according to this hypothesis, molecular clocks have to be recalibrated taking body size and temperature into account. This thesis tested the temperature effect on mutation rates in equally sized invertebrates. For the first dataset (family Eucnemidae, Coleoptera) the phylogenetic interrelationships and evolutionary history of the genus Arrhipis had to be inferred before the influence of temperature on substitution rates could be studied. Further, a second, larger invertebrate dataset (family Syrphidae, Diptera) was employed. Several methodological approaches, a number of genes and multiple molecular clock models revealed that there was no consistent relationship between temperature and mutation rate for the taxa under study. Thus, the body size effect, observed in vertebrates but controversial for invertebrates, rather than temperature may be the underlying driving force behind the metabolic-rate dependent molecular clock. Therefore, the metabolic-rate dependent molecular clock does not hold for the here studied invertebrate groups. This thesis emphasizes that molecular techniques relying on mutation rates have to be applied with caution. Whereas they may work satisfactorily under certain conditions for specific taxa, they may fail for others. The molecular clock as well as DNA barcoding should incorporate all the information and data available to obtain comprehensive estimations of the existing biodiversity and its evolutionary history.