A generalized learning paradigm exploiting the structure offeedforward neural networks

In this paper a general class of fast learning algorithms for feedforward neural networks is introduced and described. The approach exploits the separability of each layer into linear and nonlinear blocks and consists of two steps. The first step is the descent of the error functional in the space of the outputs of the linear blocks (descent in the neuron space), which can be performed using any preferred optimization strategy. In the second step, each linear block is optimized separately by using a least squares (LS) criterion. To demonstrate the effectiveness of the new approach, a detailed treatment of a gradient descent in the neuron space is conducted. The main properties of this approach are the higher speed of convergence with respect to methods that employ an ordinary gradient descent in the weight space backpropagation (BP), better numerical conditioning, and lower computational cost compared to techniques based on the Hessian matrix. The numerical stability is assured by the use of robust LS linear system solvers, operating directly on the input data of each layer. Experimental results obtained in three problems are described, which confirm the effectiveness of the new method.     

Spatial Charge Separation as the Origin of Anomalous Stark Effect in Fluorous 2D Hybrid Perovskites

2D hybrid perovskites (2DP) are versatile materials, whose electronic and optical properties can be tuned through the nature of the organic cations (even when those are seemingly electronically inert). Here, it is demonstrated that fluorination of the organic ligands yields glassy 2DP materials featuring long‐lived correlated electron–hole pairs. Such states have a marked charge‐transfer character, as revealed by the persistent Stark effect in the form of a second derivative in electroabsorption. Modeling shows that electrostatic effects associated with fluorination, combined with the steric hindrance due to the bulky side groups, drive the formation of spatially dislocated charge pairs with reduced recombination rates. This work enriches and broadens the current knowledge of the photophysics of 2DP, which will hopefully guide synthesis efforts toward novel materials with improved functionalities.     

Scintigraphic imaging of pituitary adenomas: an in vivo evaluation of somatostatin receptors

We have performed pituitary scintigraphy with the somatostatin (SS) analog pentetreotidean by (111In-P) in patients with GH-secreting adenoma or with "clinically non functioning" adenoma (NFA) to evaluate the presence and the functionality of SS receptors (SS-R). 111In-P pituitary accumulation was expressed as Activity Ratio (AR): the ratio between the uptake of radioactivity by the adenoma and that of the normal brain tissue. In subjects without pituitary disease, AR ranged from 1.6 to 2.2 and a value lower than 2.2 was thus arbitrarily considered as normal. In 15 out of the 17 patients with GH-secreting adenoma, an accumulation of the radioligand was shown. Median AR was 3.8 (range 1-6.9; in 14 AR were greater that 2.2) and ARs were directly correlated (r = 0.54; p < 0.05) with the suppressibility of plasma GH levels by octreotide (OC) acute administration. In two patients who repeated scintigraphy during chronic OC treatment, AR values were reduced. In all the 22 patients with NFA an accumulation of 111In-P at the pituitary level was observed and median AR was 3.0 (range 1.5-20; in 14 greater that 2.2). In vitro autoradiography of surgical specimens in 6 NFA patients revealed SS-R in 4 cases with high scintigraphic AR and negative results in two cases with low AR. Scintiscan was repeated during chronic OC treatment in 5 patients with high score: AR decreased in one patient, increased in three, and did not change in the other patient. No changes in tumor size were shown in any of these patients. A total of 8 patients (3 GH secreting and 5 NFA) had "normal" AR values. CONCLUSIONS: In acromegaly scintigraphy with 111In-P visualizes functioning pituitary SS-R coupled to intracellular events that control hormonal hypersecretion and tumor growth. In contrast, in spite of the positivity of 111In-P imaging in most patients with NFA, their receptors might have a defect in the coupling-transduction process, as they are not inhibited by OC treatment and no tumor shrinkage is observed.     

Nonorthogonal PEEC formulation for time- and frequency-domain EM and circuit modeling

Electromagnetic solvers based on the partial element equivalent circuit (PEEC) approach have proven to be well suited for the solution of combined circuit and EM problems. The inclusion of all types of Spice circuit elements is possible. Due to this, the approach has been used in many different tools. Most of these solvers have been based on a rectangular or Manhattan representation of the geometries. In this paper, we systematically extend the PEEC formulation to nonorthogonal geometries since many practical EM problems require a more general formulation. Importantly, the model given in this paper is consistent with the classical PEEC model for rectangular geometries. Some examples illustrating the application of the approach are given for both the time and frequency domain.     

Quantifying SMT decoupling capacitor placement in dc power-busdesign for multilayer PCBs

Noise on a dc power-bus that results from device switching, as well as other potential mechanisms, is a primary source of many signal integrity (SI) and electromagnetic interference (EMI) problems. Surface mount technology (SMT) decoupling capacitors are commonly used to mitigate this power-bus noise. A critical design issue associated with this common practice in high-speed digital designs is placement of the capacitors with respect to the integrated circuits (ICs). Local decoupling, namely, placing SMT capacitors in proximity to ICs, is investigated in this study. Multilayer PCB designs that employ entire layers or area fills for power and ground in a parallel plate structure are considered. The results demonstrate that local decoupling can provide high-frequency benefits for certain PCB geometries through mutual inductive coupling between closely spaced vias. The associated magnetic flux linkage is between the power and ground layers. Numerical modeling using an integral equation formulation with circuit extraction is used to quantify the local decoupling phenomenon. Local decoupling can effectively reduce high-frequency power-bus noise, though placing capacitors adjacent to ICs may limit routing flexibility, and tradeoffs need to be made based on design requirements. Design curves are generated as a function of power-bus layer thickness and SMT capacitor/IC spacing using the modeling approach to quantify the power-bus noise reduction for decoupling capacitors located adjacent to devices. Measurement data is provided to corroborate the modeling approach     

Validation of a Three-Dimensional Transmission Line Matrix (TLM) Model Implementation of a Mode-Stirred Reverberation Chamber

Reverberation chambers are attractive electromagnetic compatibility test facilities, both economically and technically. Careful design and analysis of these facilities are important, if the results obtained are to be treated with a high level of confidence. Numerical modeling is an important part of the process of reverberation chamber design and analysis. Hence, it is important that the modeling techniques to be used are appropriately validated. Much of the published work to date takes either a statistical or a deterministic view of validation. This paper provides validation evidence for a low-resolution transmission line matrix (TLM) model of a reverberation chamber in a manner approximating the way in which the chamber is used, i.e., validating based on the effects of a simple device under test. A variety of statistical and heuristic approaches have been used to quantify the level of agreement, intending to set the likely lower bound for the quality of comparisons between simulations and measurements. While not drawing any "universal" conclusions about the veracity of the TLM technique, the paper concludes that a relatively simple model of a reverberation chamber provides a useful analysis of the chamber with close comparisons between modeled and measured data.     

Equivalent network synthesis for via holes discontinuities

A methodology is presented for the synthesis of the passive equivalent circuit of via holes in multilayer printed circuit boards. The paper describes the network synthesis starting from the extraction of the poles and residues from the driving point transfer functions for two port networks. The via hole is partitioned into elementary structures assumed not electromagnetically coupled and the scattering parameters for each one of them are evaluated by using a numerical approach suitably validated by comparison with results computed by other independent numerical methods. The equivalent circuit of the complete via hole is given by the cascading of the circuits of the elementary structures. The proposed technique is validated by comparing the computed scattering parameters with those from measurements for real test boards.     

Secondary acute myeloid leukemia following treatment with VP16-containing regimens for non-Hodgkin's lymphoma

We report on two patients who developed a secondary acute myeloid leukemia (sAL) after treatment for non-Hodgkin's lymphoma (NHL) with regimens containing low to intermediate doses of VP16. Clinical and hematologic features in these two patients were consistent with epipodophyllotoxin-associated sAL. In one case, a rearrangement of chromosome band 11q23 was detected.