Dual Input Stream Transformer for Vertical Drift Correction in Eye-tracking Reading Data.

We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17 %. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field.

Recovery via SfM photogrammetry of latent footprint impressions in carpet.

Impression evidence retained in carpet is usually recovered, if at all, in two dimensions via a vertical photograph. Here, we show that recovery is also possible via SfM photogrammetry and this gives good quality results that allow digital measurements both in the x-y plane and by depth (z-axis). This study focuses on recovery from polypropylene carpets which are widespread due to their resistance to wear and low cost. We show how traces can be recovered using both SfM photogrammetry and conventional photography with illumination provided via a crime scene light source. Experiments show that traces are retained for considerable time periods if left undisturbed, in excess of four weeks, but are quickly lost in under 8 hours by subsequent footfall. A simple simulation shows how the movement of an individual can be determined from carpet traces and the value of 3D recovery is illustrated via a set of experiments conducted with barefoot traces. We draw attention to the fact that 3D models allow a more statistical-based approach to be taken to match bare footprints at crime scenes. SfM photogrammetry is shown to provide a useful compliment to existing techniques and therefore worthy of further experimentation and potentially operational use.

Technological innovation in the recovery and analysis of 3D forensic footwear evidence: Structure from motion (SfM) photogrammetry.

The recovery of three-dimensional footwear impressions at crime scenes can be a challenge but can also yield important investigative data. Traditional methods involve casting 3D impressions but these methods have limitations: the trace is usually destroyed during capture; the process can be time consuming, with a risk of failure; and the resultant cast is bulky and therefore difficult to share and store. The use of Structure from Motion (SfM) photogrammetry has been used widely to capture fossil footprints in the geological record and while there is a small body of work advocating its use in forensic practice the full potential of this technique has yet to be realised in an operational context. The availability of affordable software is one limiting factor and here we report the availability of a bespoke freeware for SfM recovery and subsequent analysis of for footwear evidence (DigTrace). Our aim here is not to provide a rigorous comparison of SfM methods to other recovery methods, but more to illustrate the potential while also documenting the typical workflows and potential errors associated with an SfM based approach. By doing so we hope to encourage further research, experimentation and ultimately adoption by practitioners.

Do readers use character information when programming return-sweep saccades?

Reading saccades that occur within a single line of text are guided by the size of letters. However, readers occasionally need to make longer saccades (known as return-sweeps) that take their eyes from the end of one line of text to the beginning of the next. In this study, we tested whether return-sweep saccades are also guided by font size information and whether this guidance depends on visual acuity of the return-sweep target area. To do this, we manipulated the font size of letters (0.29 vs 0.39° per character) and the length of the first line of text (16 vs 26°). The larger font resulted in return-sweeps that landed further to the right of the line start and in a reduction of under-sweeps compared to the smaller font. This suggests that font size information is used when programming return-sweeps. Return-sweeps in the longer line condition landed further to the right of the line start and the proportion of under-sweeps increased compared to the short line condition. This likely reflects an increase in saccadic undershoot error with the increase in intended saccade size. Critically, there was no interaction between font size and line length. This suggests that when programming return-sweeps, the use of font size information does not depend on visual acuity at the saccade target. Instead, it appears that readers rely on global typographic properties of the text in order to maintain an optimal number of characters to the left of their first fixation on a new line.

Sexing white 2D footprints using convolutional neural networks.

Footprints are left, or obtained, in a variety of scenarios from crime scenes to anthropological investigations. Determining the sex of a footprint can be useful in screening such impressions and attempts have been made to do so using single or multi landmark distances, shape analyses and via the density of friction ridges. Here we explore the relative importance of different components in sexing two-dimensional foot impressions namely, size, shape and texture. We use a machine learning approach and compare this to more traditional methods of discrimination. Two datasets are used, a pilot data set collected from students at Bournemouth University (N = 196) and a larger data set collected by podiatrists at Sheffield NHS Teaching Hospital (N = 2677). Our convolutional neural network can sex a footprint with accuracy of around 90% on a test set of N = 267 footprint images using all image components, which is better than an expert can achieve. However, the quality of the impressions impacts on this success rate, but the results are promising and in time it may be possible to create an automated screening algorithm in which practitioners of whatever sort (medical or forensic) can obtain a first order sexing of a two-dimensional footprint.

Evidence of humans in North America during the Last Glacial Maximum.

Archaeologists and researchers in allied fields have long sought to understand human colonization of North America. Questions remain about when and how people migrated, where they originated, and how their arrival affected the established fauna and landscape. Here, we present evidence from excavated surfaces in White Sands National Park (New Mexico, United States), where multiple in situ human footprints are stratigraphically constrained and bracketed by seed layers that yield calibrated radiocarbon ages between ~23 and 21 thousand years ago. These findings confirm the presence of humans in North America during the Last Glacial Maximum, adding evidence to the antiquity of human colonization of the Americas and providing a temporal range extension for the coexistence of early inhabitants and Pleistocene megafauna.

Proof of concept for a syndromic surveillance system based on routine ambulance records in the South West of England, for the influenza season 2016/2017.

INTRODUCTION: The introduction of electronic patient records in the ambulance service provides new opportunities to monitor the population. Approximately 36% of patients presenting to English ambulance services are discharged at scene. Ambulance records are therefore an ideal data source for syndromic early event detection systems to monitor infectious disease in the pre-hospital population. It has been previously found that tympanic temperature records can be used to detect influenza outbreaks in emergency departments. This study aimed to determine whether routine tympanic temperature readings collected by ambulance crews can be used to detect seasonal influenza. METHODS: Here we show that temperature readings do allow the detection of seasonal influenza before methods applied to conventional data sources. The counts of pyretic patients were used to calculate a sliding case ratio as a measurement to detect seasonal influenza outbreaks. This method does not rely on conventional thresholds and can be adapted to the data. RESULTS: The data collected correlated with seasonal influenza. The 2016/2017 outbreak was detected up to nine weeks before other surveillance programmes. The results show that ambulance records can be a useful data source for biosurveillance systems. CONCLUSION: Temperature readings from routinely collected ambulance patient records can be used as a surveillance tool for febrile diseases.

Stat-tracks and mediotypes: powerful tools for modern ichnology based on 3D models.

Vertebrate tracks are subject to a wide distribution of morphological types. A single trackmaker may be associated with a range of tracks reflecting individual pedal anatomy and behavioural kinematics mediated through substrate properties which may vary both in space and time. Accordingly, the same trackmaker can leave substantially different morphotypes something which must be considered in creating ichnotaxa. In modern practice this is often captured by the collection of a series of 3D track models. We introduce two concepts to help integrate these 3D models into ichnological analysis procedures. The mediotype is based on the idea of using statistically-generated three-dimensional track models (median or mean) of the type specimens to create a composite track to support formal recognition of a ichno type. A representative track (mean and/or median) is created from a set of individual reference tracks or from multiple examples from one or more trackways. In contrast, stat-tracks refer to other digitally generated tracks which may explore variance. For example, they are useful in: understanding the preservation variability of a given track sample; identifying characteristics or unusual track features; or simply as a quantitative comparison tool. Both concepts assist in making ichnotaxonomical interpretations and we argue that they should become part of the standard procedure when instituting new ichnotaxa. As three-dimensional models start to become a standard in publications on vertebrate ichnology, the mediotype and stat-track concepts have the potential to help guiding a revolution in the study of vertebrate ichnology and ichnotaxonomy.

Footprints preserve terminal Pleistocene hunt? Human-sloth interactions in North America.

Predator-prey interactions revealed by vertebrate trace fossils are extremely rare. We present footprint evidence from White Sands National Monument in New Mexico for the association of sloth and human trackways. Geologically, the sloth and human trackways were made contemporaneously, and the sloth trackways show evidence of evasion and defensive behavior when associated with human tracks. Behavioral inferences from these trackways indicate prey selection and suggest that humans were harassing, stalking, and/or hunting the now-extinct giant ground sloth in the terminal Pleistocene.

Metalearning: a survey of trends and technologies.

Metalearning attracted considerable interest in the machine learning community in the last years. Yet, some disagreement remains on what does or what does not constitute a metalearning problem and in which contexts the term is used in. This survey aims at giving an all-encompassing overview of the research directions pursued under the umbrella of metalearning, reconciling different definitions given in scientific literature, listing the choices involved when designing a metalearning system and identifying some of the future research challenges in this domain.

Can we identify non-stationary dynamics of trial-to-trial variability?

Identifying sources of the apparent variability in non-stationary scenarios is a fundamental problem in many biological data analysis settings. For instance, neurophysiological responses to the same task often vary from each repetition of the same experiment (trial) to the next. The origin and functional role of this observed variability is one of the fundamental questions in neuroscience. The nature of such trial-to-trial dynamics however remains largely elusive to current data analysis approaches. A range of strategies have been proposed in modalities such as electro-encephalography but gaining a fundamental insight into latent sources of trial-to-trial variability in neural recordings is still a major challenge. In this paper, we present a proof-of-concept study to the analysis of trial-to-trial variability dynamics founded on non-autonomous dynamical systems. At this initial stage, we evaluate the capacity of a simple statistic based on the behaviour of trajectories in classification settings, the trajectory coherence, in order to identify trial-to-trial dynamics. First, we derive the conditions leading to observable changes in datasets generated by a compact dynamical system (the Duffing equation). This canonical system plays the role of a ubiquitous model of non-stationary supervised classification problems. Second, we estimate the coherence of class-trajectories in empirically reconstructed space of system states. We show how this analysis can discern variations attributable to non-autonomous deterministic processes from stochastic fluctuations. The analyses are benchmarked using simulated and two different real datasets which have been shown to exhibit attractor dynamics. As an illustrative example, we focused on the analysis of the rat's frontal cortex ensemble dynamics during a decision-making task. Results suggest that, in line with recent hypotheses, rather than internal noise, it is the deterministic trend which most likely underlies the observed trial-to-trial variability. Thus, the empirical tool developed within this study potentially allows us to infer the source of variability in in-vivo neural recordings.

Density-preserving sampling: robust and efficient alternative to cross-validation for error estimation.

Estimation of the generalization ability of a classification or regression model is an important issue, as it indicates the expected performance on previously unseen data and is also used for model selection. Currently used generalization error estimation procedures, such as cross-validation (CV) or bootstrap, are stochastic and, thus, require multiple repetitions in order to produce reliable results, which can be computationally expensive, if not prohibitive. The correntropy-inspired density-preserving sampling (DPS) procedure proposed in this paper eliminates the need for repeating the error estimation procedure by dividing the available data into subsets that are guaranteed to be representative of the input dataset. This allows the production of low-variance error estimates with an accuracy comparable to 10 times repeated CV at a fraction of the computations required by CV. This method can also be used for model ranking and selection. This paper derives the DPS procedure and investigates its usability and performance using a set of public benchmark datasets and standard classifiers.

Robust predictive modelling of water pollution using biomarker data.

This paper describes the methodology of building a predictive model for the purpose of marine pollution monitoring, based on low quality biomarker data. A step-by-step, systematic data analysis approach is presented, resulting in design of a purely data-driven model, able to accurately discriminate between various coastal water pollution levels. The environmental scientists often try to apply various machine learning techniques to their data without much success, mostly because of the lack of experience with different methods and required 'under the hood' knowledge. Thus this paper is a result of a collaboration between the machine learning and environmental science communities, presenting a predictive model development workflow, as well as discussing and addressing potential pitfalls and difficulties. The novelty of the modelling approach presented lays in successful application of machine learning techniques to high dimensional, incomplete biomarker data, which to our knowledge has not been done before and is the result of close collaboration between machine learning and environmental science communities.