Two topographic maps for data visualisation

We review a new form of self-organizing map which is based on a nonlinear projection of latent points into data space, identical to that performed in the Generative Topographic Mapping (GTM) [Bishop et al. (1997) Neurl Comput 10(1): 215–234]. But whereas the GTM is an extension of a mixture of experts, our new model is an extension of a product of experts [Hinton (2000) Technical report GCNU TR 2000-004, Gatsby Computational Neuroscience Unit, University College, London]. We show visualisation results on some real data sets and compare with the GTM. We then introduce a second mapping based on harmonic averages and show that it too creates a topographic mapping of the data. We compare these mappings on real and artificial data sets. Responsible editor: Soumen Chakrabarti.     

Effects of vesnarinone, a novel orally active inotropic agent with an immunosuppressive action, on experimental cardiac transplantation in rats

BACKGROUND: Vesnarinone (VES) has been used for treatment of patients with congestive heart failure. In addition to inotropic effects, it seems to have immunosuppressive action. We tested the hypothesis that VES suppresses graft rejection, inotropic dysfunction caused by early rejection, and chronic coronary obstruction in a heterotopic rat cardiac transplantation model. METHODS: (1) To study acute rejection, hearts from Lewis-Brown Norway (LBN) rats were transplanted into Lewis rats, which were treated with or without VES (50 or 100 mg/kg/day orally). (2) In a functional study, LBN hearts with or without VES (100 mg/kg/ day) were isolated and perfused on day 3 after transplantation to assess inotropic response to isoproterenol (3 x 10(-8) M). (3) To study chronic rejection, Lewis hearts were transplanted into Fisher 344 rats, which were treated with or without VES (50 mg/kg/day) for 90 days. Coronary obstructive disease was assessed by morphometric analysis. There were five to six animals in each group. RESULTS: (1) VES (100 mg/kg/day) prolonged LBN heart survival (11.7 +/- 0.7 vs. 9.6 +/- 0.7 days in control; P < 0.05). (2) Left ventricular developed pressure was depressed in transplanted hearts regardless of VES treatment (84 +/- 12, 90 +/- 8 vs. 144 +/- 16 mmHg in untransplanted hearts; P < 0.01). The developed pressure after administration of isoproterenol in VES-treated hearts (184 +/- 20 mmHg) was higher than transplanted hearts without VES (118 +/- 16 mmHg; P < 0.05), and similar to untransplanted hearts (203 +/- 27 mmHg; P = NS). (3) Transplanted hearts treated with or without VES showed similar grades of rejection (2.0 +/- 0.3 vs. 2.6 +/- 0.2; P = NS), intimal area (6,996 +/- 3,186 vs. 13,441 +/- 5,165 microns2; NS), and coronary luminal obstruction (45 +/- 16% vs. 67 +/- 14%; NS). CONCLUSIONS: VES produces mild prolongation in survival of rat heart grafts, but has no significant effect on chronic graft atherosclerosis. VES preserves the positive inotropic effects of isoproterenol that are otherwise deteriorated by early acute rejection.     

Phase I trial of weekly scheduling and pharmacokinetics of titanocene dichloride in patients with advanced cancer

PURPOSE: To determine the maximum-tolerated dose (MTD) and the dose-limiting toxicities (DLTs) of a weekly schedule of titanocene dichloride (TD) and to define the pharmacokinetics of titanium in plasma and urine. PATIENTS AND METHODS: Twenty patients with a median age of 58 years received 83 courses of TD. TD was given as 1-hour infusion at escalating doses from 70 to 185 mg/m2/wk. Pharmacokinetic analysis was performed in eight patients for total plasma titanium (TPTi) and in three patients for ultrafiltrable titanium (UFTi). RESULTS: At the fifth dose level (185 mg/m2/wk), a variety of DLTs were seen in five patients: fatigue in three, bilirubinemia in one, and hypokalemia in two. A further six patients were treated at 140 mg/m2; only one had dose-limiting creatinine elevation and this dose was therefore defined as the MTD. No myelosuppression or alopecia were observed. One patient with adenocarcinoma of unknown primary had a minor response. Pharmacokinetic analysis showed that TPTi maximum concentration (Cmax) values were linear with dose and elimination of TPTi was triphasic with a long terminal half-life (t1/2; median, 165 hours; range, 89 to 592). Between 7% and 24.3% of the total of administered titanium was eliminated in urine over the first 24 hours. In contrast, UFTi elimination was described by a one-compartment model with a t1/2 of 0.41 hours; peak levels of UFTi were 5.2% +/- 2.5% those of TPTi. CONCLUSION: The MTD of TD given on a weekly schedule is 140 mg/m2, with cumulative, but reversible creatinine and bilirubin elevation being the DLTs.     

Evaluation of the quantitative reliability of the 13C CP/MAS technique for the analysis of coals and related materials

The percentages of aromatic carbon in three representative solid coal samples and one pitch sample have been determined by three different n.m.r. techniques. The first two involve the use of cross-polarization/magic-angle spinning (CP/MAS) and in one case, the magnetization curves were fitted to a theoretical expression and the fit parameters used to calculate the percentage of aromatic carbon. In the second case, the percentage of aromatic carbons was obtained from a choice of contact time in the cross-polarization pulse sequence which gave maximum signal intensity. The third technique used MAS together with a simple 90° carbon pulse sequence with a 5 × T₁, (for the carbon-13 nuclei) recycle time and was assumed to give the absolute quantitative values. It was found that the results obtained using cross-polarization techniques could be substantially different from those obtained using 90° pulses and that caution must be exercised in the application of these techniques to the quantitative analysis of solid carbonaceous fuels.     

A scale-invariant feature map

We use a simple network which uses negative feedback of activation and simple Hebbian learning to self-organize in such a way as to produce a hierarchical classification network. By adding neighbourhood relations to its learning rule, we create a feature map which has the property of retaining the angular properties of the input data, i.e. vectors of similar directions are classified similarly regardless of their magnitude. We use neither re-normalization of weights nor data preprocessing in the network despite using competition based on maximizing the neuron's activation.     

A Family of Canonical Correlation Networks

We have previously introduced [2, 5] a neural implementation of Canonical Correlation Analysis (CCA). In this paper, we re-derive the learning method from a probabilistic perspective and then show that similar networks can be derived based on the pioneering work of Becker [1] if certain simplifying assumptions are made. Becker has shown that her network is able to find depth information from an abstraction of random dot stereogram data and so finally we note the similarity of the derived methods with those of Stone [3] which was used with a smooth stereo disparity data set to extract depth information.     

Unsupervised Extraction of Structural Information from High Dimensional Visual Data

We present three unsupervised artificial neural networks for the extraction of structural information from visual data. The ability of each network to represent structured knowledge in a manner easily accessible to human interpretation is illustrated using artificial visual data. These networks are used to collectively demonstrate a variety of unsupervised methods for identifying features in visual data and the structural representation of these features in terms of orientation, temporal and topographical ordering, and stereo disparity.     

Maximum and Minimum Likelihood Hebbian Learning for Exploratory Projection Pursuit

In this paper, we review an extension of the learning rules in a Principal Component Analysis network which has been derived to be optimal for a specific probability density function. We note that this probability density function is one of a family of pdfs and investigate the learning rules formed in order to be optimal for several members of this family. We show that, whereas we have previously (Lai et al., 2000; Fyfe and MacDonald, 2002) viewed the single member of the family as an extension of PCA, it is more appropriate to view the whole family of learning rules as methods of performing Exploratory Projection Pursuit. We illustrate this on both artificial and real data sets.     

Aluminum Phosphates Derived from Alumina and Alumina-Sol–Gel Systems

Aluminum phosphate products formed by the reactions of alumina and alumina-gel systems with acidic phosphates were analyzed. Drying of alumina-gel to form microcrystalline boehmite and conversion to γ-alumina by thermal treatment was indicated by the appearance of octahedral, pentacoordinate, or tetrahedral sites, which were established using ²⁷Al magic-angle-spinning solid-state nuclear magnetic resonance spectroscopy. Crystalline aluminum phosphate products and amorphous material were identified using this technique. α-alumina and heat-treated alumina-gel that were reacted with phosphate in an Al:P ratio of 1:1 yielded dramatically different aluminum orthophosphate:aluminum metaphosphate product ratios of 8.2:1 and 1:1.1, respectively. When alumina-gel was heat-treated with phosphate, an abundance of aluminum orthophosphate, aluminum metaphosphate, and hydrated aluminum phosphate products were affected by varying conditions of temperature and time of heat treatment and by the amount of phosphate present. An α-alumina/alumina-gel composite sol–gel phase that was reacted with phosphoric acid (H₃PO₄) in a Al:P ratio of 1:1 exhibited an increased quantity of aluminum metaphosphate products compared with an α-alumina:H₃PO₄ ratio of 1:1 and a higher percentage of reaction (79%) compared with the reactions of an α-alumina:H₃PO₄ ratio of 1:1 or an alumina-gel:H₃PO₄ ratio of 1:1. The morphologies of aluminum triphosphate hydrate and aluminum metaphosphate product phases were observed using scanning electron microscopy.