4 research outputs found
Fifty Years of Programming, Twenty Years of STL
No full text
Software development has come a long way since 1963, and C++ has come a long
way since 1993. Dr. Plauger offers highlights of his journey from writing
Fortran for the Princeton Cyclotron Group, through mastering the C++
Standard Template Library, to his current activities as author, C/C++
libraries vendor, and standards wonk.
About the speaker
P.J. Plauger has earned a living for over 50 years writing software and
writing about software, professions for which he has no formal training. His
hobby is physics, having earned a Ph.D. in nuclear physics.
</p
FIGURE 5 in A new species of Cephalaeschna Selys, 1883 (Odonata: Anisoptera: Aeshnidae) from Neora Valley National Park, West Bengal, India, with notes on C. acanthifrons Joshi & Kunte, 2017 and C. viridifrons (Fraser, 1922)
No full textFIGURE 5: A. Cephalaeschna acanthifrons holotype from Arunachal Pradesh, thorax [Photo by Subhajit Mazumder]; B. Cephalaeschna viridifrons from Neora Valley National Park, West Bengal, India [Photo by the author]; C. C. acanthifrons, face [Photo by Subhajit Mazumder]; D. C. viridifrons from Neora Valley National Park, face [Photo by the author]; E. C. viridifrons from Neora Valley National Park, abdomen dorsal view [Photo by the author]; F. C. acanthifrons holotype, anal appendages [Photo by Shantanu Joshi, NCBS]; G. C. viridifrons from Assam, anal appendages (reproduced from Asahina 1981a); H. C. viridifrons from Nepal, anal appendages (reproduced from Asahina 1981a); I. C. viridifrons from Neora Valley National Park, anal appendages [Photo by the author].Published as part of Dawn, Prosenjit, 2021, A new species of Cephalaeschna Selys, 1883 (Odonata: Anisoptera: Aeshnidae) from Neora Valley National Park, West Bengal, India, with notes on C. acanthifrons Joshi & Kunte, 2017 and C. viridifrons (Fraser, 1922), pp. 371-380 in Zootaxa 4949 (2) on page 378, DOI: 10.11646/zootaxa.4949.2.10, http://zenodo.org/record/463619
Projection of acoustic fields using the Fourier transform
No full textA method is presented for projecting acoustic fields using the Fourier transform. It is shown that the source velocity distribution can be represented by a number of sinusoids. Each sinusoid gives rise to a plane wave whose direction is given by the wavelength of the sinusoid. The plane waves are summed at the plane of interest to obtain the resulting pressure field. Errors are introduced when projecting pressure fields by large distances. These are circumvented by observing that the farfield pressure pattern can be found by simply taking the Fourier transform of the nearfield pressure distribution. A rectangular source is simulated on the computer; the Fourier transform technique of projecting fields is compared to a direct integration technique. The Fourier transform technique is used to back project a measured pressure pattern to detect defects on the transducer face. Measurements of pressure are made in the nearfield of a circular transducer. These measurements are forward and back projected to give the pressure and velocity at other planes
From deterministic methods to a Bayesian approximation: towards reliable segmentation of multiple sclerosis Lesions in magnetic resonance imaging
No full textEn este trabajo, se presenta una metodología para la segmentación de lesiones de esclerosis múltiple (EM) en imágenes de resonancia magnética (IRM) que aborda las limitaciones de los modelos deterministas mediante la incorporación de la estimación de incertidumbre. Se compara una arquitectura U-Net 3D determinista con una versión modificada que emplea una aproximación bayesiana con Monte Carlo Dropout (MCD) para cuantificar la incertidumbre epistémica. Los resultados demuestran que, si bien ambos modelos alcanzan un rendimiento competitivo en términos de las métricas estándar de segmentación de imágenes médicas, la estimación de incertidumbre proporciona información valiosa sobre la fiabilidad de las predicciones, especialmente en regiones desafiantes como los bordes de las lesiones. Esto tiene el potencial de mejorar la aplicabilidad clínica de la segmentación automática al permitir a los usuarios médicos evaluar la confianza en los resultados y enfocar su revisión en áreas de mayor incertidumbre.In this work, we present a methodology for the segmentation of multiple sclerosis (MS) lesions in magnetic resonance imaging (MRI) that addresses the limitations of deterministic models by incorporating uncertainty estimation. We compare a deterministic 3D U-Net architecture with a modified version that employs a Bayesian approximation with Monte Carlo Dropout (MCD) to quantify epistemic uncertainty. The results demonstrate that while both models achieve competitive performance in terms of standard medical image segmentation metrics, the uncertainty estimation provides valuable information on the reliability of the predictions, especially in challenging regions such as lesion borders. This has the potential to improve the clinical applicability of automatic segmentation by allowing medical users to assess confidence in the results and focus their review on areas of higher uncertainty
