Experimental comparison of representation methods and distance measures for time series data

The previous decade has brought a remarkable increase of the interest in applications that deal with querying and mining of time series data. Many of the research efforts in this context have focused on introducing new representation methods for dimensionality reduction or novel similarity measures for the underlying data. In the vast majority of cases, each individual work introducing a particular method has made specific claims and, aside from the occasional theoretical justifications, provided quantitative experimental observations. However, for the most part, the comparative aspects of these experiments were too narrowly focused on demonstrating the benefits of the proposed methods over some of the previously introduced ones. In order to provide a comprehensive validation, we conducted an extensive experimental study re-implementing eight different time series representations and nine similarity measures and their variants, and testing their effectiveness on 38 time series data sets from a wide variety of application domains. In this article, we give an overview of these different techniques and present our comparative experimental findings regarding their effectiveness. In addition to providing a unified validation of some of the existing achievements, our experiments also indicate that, in some cases, certain claims in the literature may be unduly optimistic.     

CID: an efficient complexity-invariant distance for time series

The ubiquity of time series data across almost all human endeavors has produced a great interest in time series data mining in the last decade. While dozens of classification algorithms have been applied to time series, recent empirical evidence strongly suggests that simple nearest neighbor classification is exceptionally difficult to beat. The choice of distance measure used by the nearest neighbor algorithm is important, and depends on the invariances required by the domain. For example, motion capture data typically requires invariance to warping, and cardiology data requires invariance to the baseline (the mean value). Similarly, recent work suggests that for time series clustering, the choice of clustering algorithm is much less important than the choice of distance measure used.In this work we make a somewhat surprising claim. There is an invariance that the community seems to have missed, complexity invariance. Intuitively, the problem is that in many domains the different classes may have different complexities, and pairs of complex objects, even those which subjectively may seem very similar to the human eye, tend to be further apart under current distance measures than pairs of simple objects. This fact introduces errors in nearest neighbor classification, where some complex objects may be incorrectly assigned to a simpler class. Similarly, for clustering this effect can introduce errors by “suggesting” to the clustering algorithm that subjectively similar, but complex objects belong in a sparser and larger diameter cluster than is truly warranted.We introduce the first complexity-invariant distance measure for time series, and show that it generally produces significant improvements in classification and clustering accuracy. We further show that this improvement does not compromise efficiency, since we can lower bound the measure and use a modification of triangular inequality, thus making use of most existing indexing and data mining algorithms. We evaluate our ideas with the largest and most comprehensive set of time series mining experiments ever attempted in a single work, and show that complexity-invariant distance measures can produce improvements in classification and clustering in the vast majority of cases.     

Effect of medium-chain triglycerides on calbindin-D9k expression in the intestine

These studies determined the effect of the saturated fat source in infant formula on the expression of calbindin-D9k (CaBP-9k). Piglets were fed from birth to 8 d with milk or formula containing saturated fatty acids as medium-chain triglycerides (MCT), coconut oil, palm oil (Palm 1), or synthesized triglycerides with 16∶0 directed to thesn-2 position (Palm 2). Levels of intestinal CaBP-9k mRNA were significantly (P<0.01) higher in piglets fed formula with MCT than in piglets fed the other formula or milk; and higher in piglets fed the Palm-1 than in piglets fed Palm-2 formula. This is the first evidence that MCT alter piglet intestinal CaBP-9k mRNA.     

Covalent and Ionic States of Strong Acids at the Ice Surface

The relationship between the degree of ionization and the environment of a strong acid is of basic scientific interest. Often this relationship reduces to the interdependence of ion/acid hydration and proton transfer. Despite the presence of pure water, the surface of crystalline ice, particularly at cryogenic temperatures, is one of limited (controlled?) availability of water of hydration. Here, the detailed nature of the ice surface and the states of strong acids adsorbed to ice at cryogenic temperatures are examined. These subjects are of special current interest since the ability to model the complex chemistry that occurs on the surfaces of water-rich particles in the atmosphere, particularly in the stratosphere over the polar regions, requires a valid concept of the acid-ice interface. Our combined spectroscopic and simulation studies have identified the surface of free-standing ice particles as badly disordered, with a range of water-ring sizes and an increased level of H-bond saturation relative to an ordered ice surface. FT-IR results are reported for the interaction of the surface of such ice particles with submonolayer amounts of adsorbed DCl, DBr, and HNO₃ and for multilayer exposure to DCl. The DCl and DBr adsorbed states demonstrate behavior familiar from observations on strongly bound molecular adsorbates. Two methods have been devised for exposure of the nanocrystals to HNO₃ One gives an ionic state initially, while the initial state of the other approach is molecular. In both instances, the system is observed to evolve, with time/warming, towards a common mixed molecular–ionic adsorbed state.     

A review of recent laboratory research and commercial developments in fast pyrolysis and upgrading

Robust alternative technology choices are required in the paradigm shift from the current crude oil-reliant transport fuel platform to a sustainable, more flexible transport infrastructure. In this vein, fast pyrolysis of biomass and upgrading of the product is deemed to have potential as a technology solution. The objective of this review is to provide an update on recent laboratory research and commercial developments in fast pyrolysis and upgrading techniques. Fast pyrolysis is a relatively mature technology and is on the verge of commercialisation. While upgrading of bio-oils is currently confined to laboratory and pilot scale, an increased understanding of upgrading processes has been achieved in recent times.     

Atomic force microscopy study of rubbed polyimide films

The surface of rubbed polyimide films was studied as a function of the film thickness and applied load using atomic force microscopy (AFM). Three dominant consequences of rubbing on the film topography were observed: scratches confined to the near surface of the film, tears that extended to the substrate, and strings of islands or droplets aligned parallel to the rubbing direction. Tears, found only in films less than 50 nm thick, varied in areal density and shape due to variations in the film thickness, rubbing load, and adhesion to the substrate. Strings of droplets aligned in the rubbing direction were seen on most samples without discernible dependence on the thickness or rubbing load. © 2001 John Wiley & Sons, Inc. J Appl Polym Sci 80: 1470–1477, 2001     

The effects of varied imidization conditions on rubbed polyimide film surface morphology

The thermal histories of polyimide films were varied to study the effect of the extent of imidization on the response of films to rubbing. Films rubbed prior to imidization formed tears that extended to the substrate; tears in these films formed at smaller film thicknesses than in films imidized prior to rubbing. Aligned nanoscopic islands were also seen. The alignment of these islands disappeared upon imidization. Films imidized at a low temperature showed less alignment of islands than conventionally prepared films. © 2004 Wiley Periodicals, Inc. J Appl Polym Sci 93: 1192–1197, 2004     

Finding unusual medical time-series subsequences: algorithms and applications.

In this work, we introduce the new problem of finding time series discords. Time series discords are subsequences of longer time series that are maximally different to all the rest of the time series subsequences. They thus capture the sense of the most unusual subsequence within a time series. While discords have many uses for data mining, they are particularly attractive as anomaly detectors because they only require one intuitive parameter (the length of the subsequence), unlike most anomaly detection algorithms that typically require many parameters. While the brute force algorithm to discover time series discords is quadratic in the length of the time series, we show a simple algorithm that is three to four orders of magnitude faster than brute force, while guaranteed to produce identical results. We evaluate our work with a comprehensive set of experiments on electrocardiograms and other medical datasets.     

Converting non-parametric distance-based classification to anytime algorithms

For many real world problems we must perform classification under widely varying amounts of computational resources. For example, if asked to classify an instance taken from a bursty stream, we may have anywhere from several milliseconds to several minutes to return a class prediction. For such problems an anytime algorithm may be especially useful. In this work we show how we convert the ubiquitous nearest neighbor classifier into an anytime algorithm that can produce an instant classification, or if given the luxury of additional time, can continue computations to increase classification accuracy. We demonstrate the utility of our approach with a comprehensive set of experiments on data from diverse domains. We further show the utility of our work with two deployed applications, in classifying and counting fish, and in classifying insects.