Differentiating the Multipoint Expected Improvement for Optimal Batch Design

This work deals with parallel optimization of expensive objective functions which are modelled as sample realizations of Gaussian processes. The study is formalized as a Bayesian optimization problem, or continuous multi-armed bandit problem, where a batch of q > 0 arms is pulled in parallel at each iteration. Several algorithms have been developed for choosing batches by trading off exploitation and exploration. As of today, the maximum Expected Improvement (EI) and Upper Confidence Bound (UCB) selection rules appear as the most prominent approaches for batch selection. Here, we build upon recent work on the multipoint Expected Improvement criterion, for which an analytic expansion relying on Tallis’ formula was recently established. The computational burden of this selection rule being still an issue in application, we derive a closed-form expression for the gradient of the multipoint Expected Improvement, which aims at facilitating its maximization using gradient-based ascent algorithms. Substantial computational savings are shown in application. In addition, our algorithms are tested numerically and compared to state-of-the-art UCB-based batchsequential algorithms. Combining starting designs relying on UCB with gradient-based EI local optimization finally appears as a sound option for batch design in distributed Gaussian Process optimization.

The floating mass transducer at the round window: direct transmission or bone conduction?

The round window placement of a floating mass transducer (FMT) is a new approach for coupling an implantable hearing system to the cochlea. We evaluated the vibration transfer to the cochlear fluids of an FMT placed at the round window (rwFMT) with special attention to the role of bone conduction. A posterior tympanotomy was performed on eleven ears of seven human whole head specimens. Several rwFMT setups were examined using laser Doppler vibrometry measurements at the stapes and the promontory. In three ears, the vibrations of a bone anchored hearing aid (BAHA) and an FMT fixed to the promontory (pFMT) were compared to explore the role of bone conduction. Vibration transmission to the measuring point at the stapes was best when the rwFMT was perpendicularly placed in the round window and underlayed with connective tissue. Fixation of the rwFMT to the round window exhibited significantly lower vibration transmission. Although measurable, bone conduction from the pFMT was much lower than that of the BAHA. Our results suggest that the rwFMT does not act as a small bone anchored hearing aid, but instead, acts as a direct vibratory stimulator of the round window membrane.

The locus for bovine dilated cardiomyopathy maps to chromosome 18

Bovine dilated cardiomyopathy (BDCMP) is a severe and terminal disease of the heart muscle observed in Holstein-Friesian cattle over the last 30 years. There is strong evidence for an autosomal recessive mode of inheritance for BDCMP. The objective of this study was to genetically map BDCMP, with the ultimate goal of identifying the causative mutation. A whole-genome scan using 199 microsatellite markers and one SNP revealed an assignment of BDCMP to BTA18. Fine-mapping on BTA18 refined the candidate region to the MSBDCMP06-BMS2785 interval. The interval containing the BDCMP locus was confirmed by multipoint linkage analysis using the software loki. The interval is about 6.7 Mb on the bovine genome sequence (Btau 3.1). The corresponding region of HSA19 is very gene-rich and contains roughly 200 genes. Although telomeric of the marker interval, TNNI3 is a possible positional and a functional candidate for BDCMP given its involvement in a human form of dilated cardiomyopathy. Sequence analysis of TNNI3 in cattle revealed no mutation in the coding sequence, but there was a G-to-A transition in intron 6 (AJ842179:c.378+315G>A). The analysis of this SNP using the study's BDCMP pedigree did not conclusively exclude TNNI3 as a candidate gene for BDCMP. Considering the high density of genes on the homologous region of HSA19, further refinement of the interval on BTA18 containing the BDCMP locus is needed.

Bone conduction in Thiel-embalmed cadaver heads

INTRODUCTION Sound can reach the inner ear via at least two different pathways: air conduction and bone conduction (BC). BC hearing is used clinically for diagnostic purposes and for BC hearing aids. Research on the motion of the human middle ear in response to BC stimulation is typically conducted using cadaver models. We evaluated middle ear motion of Thiel-embalmed whole-head specimens in terms of linearity, reproducibility, and consistency with the reported middle ear motion of living subjects, fresh cadaveric temporal bones, and whole-heads embalmed with a Non-Thiel solution of salts. METHODS We used laser Doppler vibrometry to measure the displacement of the skull, the umbo, the cochlear promontory, the stapes, and the round window in seven ears from four human whole-head specimens embalmed according to Thiel's method. The ears were stimulated with a Baha(®) implanted behind the auricle. RESULTS The Thiel model shows promontory velocity similar to that reported in the literature for whole-heads embalmed with a Non-Thiel solution of salts (0- to 7-dB difference). The Thiel heads' relative velocity of the stapes with respect to the promontory was similar to that of fresh cadaver temporal bones (0- to 4-dB difference). The velocity of the umbo was comparable in Thiel-embalmed heads and living subjects (0- to 10-dB difference). The skull and all middle ear elements measured responded linearly to different stimulation levels, with an average difference less than 1 dB. The variability of repeated measurements for both short- (2 h; 4 dB) and long-term (4-16 weeks; 6 dB) repetitions in the same ear, and the difference between the two ears of the same donor (approximately 10 dB) were lower than the inter-individual difference (up to 25 dB). CONCLUSION Thiel-embalmed human whole-head specimens can be used as an alternative model for the study of human middle ear mechanics secondary to BC stimulation. At some frequencies, differences from living subjects must be considered.

Fast parallel kriging-based stepwise uncertainty reduction with application to the identification of an excursion set

Stepwise uncertainty reduction (SUR) strategies aim at constructing a sequence of points for evaluating a function  f in such a way that the residual uncertainty about a quantity of interest progressively decreases to zero. Using such strategies in the framework of Gaussian process modeling has been shown to be efficient for estimating the volume of excursion of f above a fixed threshold. However, SUR strategies remain cumbersome to use in practice because of their high computational complexity, and the fact that they deliver a single point at each iteration. In this article we introduce several multipoint sampling criteria, allowing the selection of batches of points at which f can be evaluated in parallel. Such criteria are of particular interest when f is costly to evaluate and several CPUs are simultaneously available. We also manage to drastically reduce the computational cost of these strategies through the use of closed form formulas. We illustrate their performances in various numerical experiments, including a nuclear safety test case. Basic notions about kriging, auxiliary problems, complexity calculations, R code, and data are available online as supplementary materials.

Bone conduction responses of middle ear structures in Thiel embalmed heads

Thiel-embalmed human whole-head specimens offer a promising alternative model for bone conduction (BC) studies of middle ear structures. In this work we present the Thiel model’s linearity and stability over time as well as its possible use in the study of a fixed ossicle chain. Using laser Doppler vibrometry (LDV), the motion of the retroauricular skull, the promontory, the stapes footplate and the round window (RW) were measured. A bone-anchored hearing aid stimulated the ears with step sinus tones logarithmically spread between 0.1 and 10 kHz. Linearity of the model was verified using input levels in steps of 10 dBV. The stability of the Thiel model over time was examined with measurements repeated after hours and weeks. The influence of a cement-fixed stapes was assessed. The middle ear elements measured responded linearly in amplitude for the applied input levels (100, 32.6, and 10 mV). The variability of measurements for both short- (2 h) and long-term (4-16 weeks) repetitions in the same ear was lower than the interindividual difference. The fixation of the stapes induced a lowered RW displacement for frequencies near 750 Hz (-4 dB) and an increased displacement for frequencies above 1 kHz (max. +3.7 dB at 4 kHz). LDV assessment of BC-induced middle ear motion in Thiel heads can be performed with stable results. The vibratory RW response is affected by the fixation of the stapes, indicating a measurable effect of ossicle chain inertia on BC response in Thiel embalmed heads.