Learning from Scarce Information: Using Synthetic Data to Classify Roman Fine Ware Pottery

In this article, we consider a version of the challenging problem of learning from datasets whose size is too limited to allow generalisation beyond the training set. To address the challenge, we propose to use a transfer learning approach whereby the model is first trained on a synthetic dataset replicating features of the original objects. In this study, the objects were smartphone photographs of near-complete Roman terra sigillata pottery vessels from the collection of the Museum of London. Taking the replicated features from published profile drawings of pottery forms allowed the integration of expert knowledge into the process through our synthetic data generator. After this first initial training the model was fine-tuned with data from photographs of real vessels. We show, through exhaustive experiments across several popular deep learning architectures, different test priors, and considering the impact of the photograph viewpoint and excessive damage to the vessels, that the proposed hybrid approach enables the creation of classifiers with appropriate generalisation performance. This performance is significantly better than that of classifiers trained exclusively on the original data, which shows the promise of the approach to alleviate the fundamental issue of learning from small datasets.     

Shadowing of directional noise sources by finite noise barriers

The present study investigates the shadowing effect of barriers of infinite or finite length in the presence of directional noise sources. The diffraction model termed [Directive Line Source Model (DLSM) J. Acoust. Soc. Am. 107 (2000) 2973-2986] is employed. DLSM is appropriately modified and extended to include the effects of ground reflection, diffraction by the side edges of a finite length barrier, and diffraction by directional noise sources. Results obtained are shown to be in good agreement with available experimental data and known analytical solutions. An application of the enhanced DLSM is illustrated using helicopter type noise, which is highly directive. The noise source is modeled as a directional point source with far field directivity data and the enhanced DLSM is employed to compare the noise field with and without the barrier present for three different directivity patterns, various source locations and orientations, as well as, for various barrier lengths.     

Perron's method for quasilinear hyperbolic systems: Part I

We define a notion of viscosity solution (sub-, supersolution) for these systems, prove a comparison principle and we prove existence of viscosity solutions using a Perron like method. In Part I, we do all the above except prove existence using the Perron method.     

Lateral flow devices for nucleic acid analysis exploiting quantum dots as reporters

There is a growing interest in the development of biosensors in the form of simple lateral flow devices that enable visual detection of nucleic acid sequences while eliminating several steps required for pipetting, incubation and washing out the excess of reactants. In this work, we present the first dipstick-type nucleic acid biosensors based on quantum dots (QDs) as reporters. The biosensors enable sequence confirmation of the target DNA by hybridization and simple visual detection of the emitted fluorescence under a UV lamp. The ‘diagnostic’ membrane of the biosensor contains a test zone (TZ) and a control zone (CZ). The CZ always fluoresces in order to confirm the proper function of the biosensor. Fluorescence is emitted from the TZ, only when the specific nucleic acid sequence is present. We have developed two general types of QD-based nucleic acid biosensors, namely, Type I and Type II, in which the TZ consists of either immobilized streptavidin (Type I) or immobilized oligodeoxynucleotides (Type II). The control zone consists of immobilized biotinylated albumin. No purification steps are required prior to the application of the DNA sample on the strip. The QD-based nucleic acid biosensors performed accurately and reproducibly when applied to (a) the visual detection of PCR amplification products and (b) visual genotyping of single nucleotide polymorphisms (SNPs) in human genomic DNA from clinical samples. As low as 1.5 fmol of double-stranded DNA were clearly detected by naked eye and the dynamic range extended to 200 fmol. The %CV were estimated to be 4.3–8.2.     

Tricarbonyl (triquinacene)-molybdenum and-tungsten

Triquinacene reacts with hexacarbonylmolybdenum to give tricarbonyl(triquinacene)molybdenum, and with tris(acetonitrile)tricarbonyltungsten to give tricarbonyl(triquinacene)tungsten, whereas efforts to synthesize the corresponding chromium complex, tricarbonyl(triquinacene)chromium, were unsuccessful.The molybdenum complex was characterized by ¹H and ¹³C NMR spectroscopy, mass spectra, and a single crystal X-ray structure determination. The tungsten complex is thermally less stable and more susceptible to oxidation than its molybdenum analogue and was characterized by the mass spectrum and the ¹H NMR spectrum. The crystal structure of the tricarbonyl(triquinacene)molybdenum is compared to that for the free ligand. Besides the expected lengthening of the CC bonds, the complex shows a deepening of the triquinacene “basket”, presumably to give better overlap of the CC bonds with molybdenum orbitals.