Prediction Intervals for Neural Networks via Nonlinear Regression

Standard methods for computing prediction intervals in nonlinear regression can be effectively applied to neural networks when the number of training points is large. Simulations show, however, that these methods can generate unreliable prediction intervals on smaller datasets when the network is trained to convergence. Stopping the training algorithm prior to convergence, to avoid overfitting, reduces the effective number of parameters but can lead to prediction intervals that are too wide. We present an alternative approach to estimating prediction intervals using weight decay to fit the network and show via a simulation study that this method may be effective in overcoming some of the shortcomings of the other approaches.     

A semiautomated image analysis procedure for the quantification of dental microwear II

Paleontologists often examine microscopic scratches and pits that form on teeth for indications of the diets of past animals. Most researchers count and measure scratches and pits from photomicrographs to identify microwear patterns for comparison among samples. This paper describes an affordable, available semiautomated image analysis procedure for microwear quantification. An image is downloaded from a scanning electron microscope frame buffer to a microcomputer, and the user identifies microwear features with a mouse-driven pointer. Microwear feature density, dimensions, and orientations are then computed and stored as ASCII files for subsequent data analysis.     

Creep-induced local lattice parameter changes in a monocrystalline nickel-base superalloy

The γ-γ lattice mismatch of specimens of the monocrystalline nickel-base superalloy SRR 99 has been measured by a high-resolution X-ray diffraction technique for the undeformed state and after high-tem-perature creep deformation. During creep deformation beyond the minimum creep rate (total strain =0.5%), the lattice mismatches, measured in and perpendicular to the [001] stress axis, respectively, un-dergo changes in opposite directions. This reflects the buildup of a complex deformation-induced triaxial state of internal stress in the phases y and γ. The overall resolved shear stresses that act in γ and y due to the combined action of the external and internal stresses are estimated, and the conditions under which cutting of the γ phase by dislocations should occur are discussed.     

An efficient implementation of SELF, a dynamically-typed object-oriented language based on prototypes

We have developed and implemented techniques that double the performance of dynamically-typed object-oriented languages. Our SELF implementation runs twice as fast as the fastest Smalltalk implementation, despite SELF's lack of classes and explicit variables.To compensate for the absence of classes, our system uses implementation-levelmaps to transparently group objects cloned from the same prototype, providing data type information and eliminating the apparent space overhead for prototype-based systems. To compensate for dynamic typing, user-defined control structures, and the lack of explicit variables, our system dynamically compilesmultiple versions of a source method, eachcustomized according to its receiver's map. Within each version the type of the receiver is fixed, and thus the compiler can statically bind andinline all messages sent toself.Message splitting andtype prediction extract and preserve even more static type information, allowing the compiler to inline many other messages. Inlining dramatically improves performance and eliminates the need to hard-wire low-level methods such as+, ==, andifTrue:.Despite inlining and other optimizations, our system still supports interactive programming environments. The system traverses internal dependency lists to invalidate all compiled methods affected by a programming change. The debugger reconstructs inlined stack frames from compiler-generated debugging information, making inlining invisible to the SELF programmer.This work has been generously supported by National Science Foundation Presidential Young Investigator Grant #CCR-8657631, and by IBM, Texas Instruments, NCR, Tandem Computers, Apple Computer, and Sun Microsystems.This paper was originally published inOOPSLA '89 Conference Proceedings (SIGPLAN Notices, 25, 10 (1989) 49–70).     

Effect of radiation on the crystals of polyethylene and paraffins: 1. Formation of the hexagonal lattice and the destruction of crystallinity in polyethylene

In this, the first of a series of three preliminary papers on the effect of high dose radiation on the paraffinoid crystal lattice, the destruction of crystallinity in polyethylene under the influence of γ-radiation is followed by X-ray diffraction and calorimetry. The principal new finding is the occurrence of an orthorhombic-hexagonal phase transition on heating subsequent to irradiation at ambient temperature beyond a certain critical dose. This othorhombic-hexagonal transition temperature is an increasing function of crystal thickness, and for all samples decreases with increasing dose. While the lattice parameters are seen to change gradually with dose (as observed at room temperature) the destruction of crystallinity itself occurs rather catastrophically beyond a certain dose. It is envisaged that this collapse of crystallinity is taking place as the temperature of the orthorhombic-hexagonal transition, which decreases steadily with dose, approaches the irradiation temperature. The above scheme is supported by the higher radiation sensitivity of the hexagonal, or nearly hexagonal, phase as found here in paraffins. As a side-line, the present findings lead to a grading of hexagonal phases in paraffinoid substances as obtained along different routes.     

The application of the differential transform technique to point source problems of elasticity

The differential transform, being a formal integral transform in which integrations have been eliminated, is used to represent some point sources in operational form. The sources can also be represented by integrals or integral transforms. However, in order to solve some boundary value problems, it is simpler to use the operational representation than the integral or integral transform representation. Three examples illustrate the application of the differential transform and the operational representations.     

Square and Hexagonal Columnar Liquid Crystals Confined in Square and Triangular Pores

Columnar liquid crystals confined in cylindrical pores are known to orient perpendicular to the pore axis, although for potential nanoelectronic applications they would need to be aligned axially. The X‐ray diffraction study reveals that while in all three compounds examined, the columns arrange in concentric circles in circular pores, axial alignment is indeed achieved for the hexagonal phase of a discotic hexa‐peri‐hexabenzocoronene in “triangular pores,” and for the square phase of a T‐shaped amphiphile in “square pores.” For the third compound, a carbazole dendron with very soft hexagonal columns, a new three‐circle configuration is observed by atomic force microscopy (AFM) in triangular pores. The complex structure of domain walls in columnar liquid crystals is thus seen for the first time on the level of individual columns. A simple estimation of free energies confirms that it is the distortion of polygonal columnar lattice that prevents axial orientation in circular pores. The model also explains the appearance of the three‐circle arrangement of soft columns.     

X-ray line profile analysis—An ideal tool to quantify structural parameters of nanomaterials

For a long time the shift and broadening of Bragg profiles have been used to evaluate internal stresses and coherent domain sizes, i.e. the smallest crystalline region without lattice defects. Modern technology provides both enhanced detector resolution and high brilliance x-ray sources thus allowing measurements of x-ray peaks with a high resolution in space and time. In parallel to the hardware, also diffraction theories have been substantially improved so that the shape of Bragg profiles can be quantitatively evaluated not only in terms of the crystallite size and its distribution, but also in terms of the density, type and arrangement of dislocations, twins and stacking faults. Thus state-of-the-art x-ray line profile analysis enables the thorough characterization especially of nanostructured materials which also contain lattice defects. The method can be used also to prove the existence of dislocations in crystalline materials. Examples of nanostructured metals, polymers and even molecular crystals like fullerenes are given.