Search

Deep learning (DL) generates new computational tasks that are different from those encountered in classical scientific applications. In particular, DL training and inference require general matrix multiplications (gemm) with matrix operands that are far from large and square as in other scientific fields. In addition, DL models gain arithmetic/storage complexity, and as a result, reduced precision via quantization is now mainstream for inferring DL models in edge devices. Automatic code generation addresses these new types of gemm by (1) improving portability between different hardware with only one base code; (2) supporting mixed and reduced precision; and (3) enabling auto-tuning methods that, given a base operation, perform a (costly) optimization search for the best schedule. In this paper, we rely on Apache TVM to generate an experience-guided gemm that provides performance competitive with the TVM auto-scheduler, while reducing tuning time by a factor of 48×. [ABSTRACT FROM AUTHOR]

This work presents a microwave resonant multi-parameter sensor devoted to the simultaneous extraction of three characteristics of a homogeneous solid sample: its dielectric permittivity, its loss tangent and its thickness. The device is composed of three coupled resonators in two different substrate boards, having the sample between the boards, in a sandwich configuration. Presence of the sample impacts the electrical response of the device, not only influencing resonators, but also by affecting inter-resonator couplings. A method to analyse the response of the device, allowing for the extraction of the desired characteristics of the sample is proposed, as well as an experimental calibration procedure. The model is built upon 990 simulations, calibrated with three reference-samples measurements and then tested over 18 experimental measurements, with good results, thereby validating the multi-parameter sensing approach. [ABSTRACT FROM AUTHOR]

General matrix multiplication (gemm) is a fundamental kernel in scientific computing and current frameworks for deep learning. Modern realisations of gemm are mostly written in C, on top of a small, highly tuned micro-kernel that is usually encoded in assembly. The high performance realisation of gemm in linear algebra libraries in general include a single micro-kernel per architecture, usually implemented by an expert. In this paper, we explore a couple of paths to automatically generate gemm micro-kernels, either using C++ templates with vector intrinsics or high-level Python scripts that directly produce assembly code. Both solutions can integrate high performance software techniques, such as loop unrolling and software pipelining, accommodate any data type, and easily generate micro-kernels of any requested dimension. The performance of this solution is tested on three ARM-based cores and compared with state-of-the-art libraries for these processors: BLIS, OpenBLAS and ArmPL. The experimental results show that the auto-generation approach is highly competitive, mainly due to the possibility of adapting the micro-kernel to the problem dimensions. [ABSTRACT FROM AUTHOR]

We address the efficient implementation of the convolution operator on the GAP8 parallel ultra-low power platform (PULP), a heterogeneous multi-core processor equipped with a fabric controller (FC); a cluster of eight compute cores; and a four-level memory hierarchy with scratchpads instead of conventional, hardware-assisted cache memories. Our solution for this platform transforms the convolution into a general matrix–matrix multiplication (gemm) via the lowering approach, demonstrating that it is possible to attain reasonable performance on the GAP8 by carefully adapting techniques such as tiling and loop parallelism, which are mainstream in the multi-threaded, cache-aware realization of gemm. [ABSTRACT FROM AUTHOR]

We report on the effect of a weak magnetic field applied on an iodine cell used to frequency stabilize a laser. A 1.5 μ m laser is frequency tripled in order to excite the molecular transitions at 0.51 μ m and frequency locked on a hyperfine line. With this frequency reference, we report short-term stability about 3 × 10 - 14 τ - 1 / 2 , with a minimum value of 4 × 10 - 15 at 200 s. The lower part of 10 - 15 frequency stability domain is reached, in our case, only by adding an efficient magnetic shield around the sealed quartz iodine cell. In order to quantify the Zeeman effect, we applied magnetic fields of several × 10 - 4 T on the cell containing the iodine vapour. The Zeeman effect affects the lineshape transition in such a way that we observe a modification of the laser frequency. We have measured this linear Zeeman shift at (1062 ± 6) × 10 4 Hz/T for the a1 hyperfine component of the R(34) 44-0 transition, near 514 nm by applying a magnetic field along the cell. Thus, in case of uncontrolled magnetic fields of an order of magnitude of 1 × 10 - 4 T, the frequency stability is limited in the upper of the 10 - 14 domain. [ABSTRACT FROM AUTHOR]

The use of virtual reality or augmented reality systems in billiards sports are useful tools for pure entertainment or improving the player's skills. Depending on the purpose of these systems, tracking algorithms based on computer vision must be used. These algorithms are especially useful in systems aiming to reconstruct the trajectories followed by the balls after a strike. However, depending on the billiard modality, the problem of tracking multiple small identical objects, such as balls, is a complex task. In addition, when an amateur or nontop professional player uses low-frame-rate and low-resolution devices, problems such as blurred balls, blurred contours, or fuzzy edges, among others, arise. These effects have a negative impact on ball-tracking accuracy and reconstruction quality. Thus, this work proposes two contributions. The first contribution is a new tracking algorithm called "multiobject local tracking (MOLT)". This algorithm can track balls with high precision and accuracy even with motion blur caused by low-resolution and low-frame-rate devices. Moreover, the proposed MOLT algorithm is compared with nine tracking methods and four different metrics, outperforming the rest of the methods in the majority of the cases and providing a robust solution. The second contribution is a whole system to track (using the MOLT algorithm) and reconstruct the movements of the balls on a billiard table in a 3D virtual world using computer vision. The proposed system covers all steps from image capture to 3D reconstruction. The 3D reconstruction results have been qualitatively evaluated by different users through a series of questionnaires, obtaining an overall score of 7.6 (out of 10), which indicates that the system is a promising and useful tool for training. Finally, both the MOLT algorithm and the reconstruction system are tested in three billiard modalities: blackball, carom billiards, and snooker. [ABSTRACT FROM AUTHOR]

In this article, we globalize the quasi-Newton algorithm proposed in Macías et al. (Appl Math Comput 441:127678, 2023) introducing an exact line search, we propose a polynomial approximation to the merit function and we deduce a sufficient condition for its minimization interval. We use the exact merit function and its approximation to propose two global quasi-Newton algorithms for solving matrix polynomial equations. For each algorithm, we prove that the exact line search does not affect the convergence of quasi-Newton method. In addition, we present comparative numerical tests of the algorithmic proposals in which we also compare with global Newton's method. [ABSTRACT FROM AUTHOR]

We take a step forward towards developing high-performance codes for the convolution operator, based on the Winograd algorithm, that are easy to customise for general-purpose processor architectures. In our approach, augmenting the portability of the solution is achieved via the introduction of vector instructions from Intel SSE/AVX2/AVX512 and ARM NEON/SVE to exploit the single-instruction multiple-data capabilities of current processors as well as OpenMP pragmas to exploit multi-threaded parallelism. While this comes at the cost of sacrificing a fraction of the computational performance, our experimental results on three distinct processors, with Intel Xeon Skylake, ARM Cortex A57 and Fujitsu A64FX processors, show that the impact is affordable and still renders a Winograd-based solution that is competitive when compared with the lowering gemm-based convolution. [ABSTRACT FROM AUTHOR]

In this work, we assess the performance and energy efficiency of high-performance codes for the convolution operator, based on the direct, explicit/implicit lowering and Winograd algorithms used for deep learning (DL) inference on a series of ARM-based processor architectures. Specifically, we evaluate the NVIDIA Denver2 and Carmel processors, as well as the ARM Cortex-A57 and Cortex-A78AE CPUs as part of a recent set of NVIDIA Jetson platforms. The performance–energy evaluation is carried out using the ResNet-50 v1.5 convolutional neural network (CNN) on varying configurations of convolution algorithms, number of threads/cores, and operating frequencies on the tested processor cores. The results demonstrate that the best throughput is obtained on all platforms with the Winograd convolution operator running on all the cores at their highest frequency. However, if the goal is to reduce the energy footprint, there is no rule of thumb for the optimal configuration. [ABSTRACT FROM AUTHOR]

We provide a practical demonstration that it is possible to systematically generate a variety of high-performance micro-kernels for the general matrix multiplication (gemm) via generic templates which can be easily customized to different processor architectures and micro-kernel dimensions. These generic templates employ vector intrinsics to exploit the SIMD (single instruction, multiple data) units in current general-purpose processors and, for the particular type of gemm problems encountered in deep learning, deliver a floating-point throughput rate on par with or even higher than that obtained with conventional, carefully tuned implementations of gemm in current linear algebra libraries (e.g., BLIS, AMD AOCL, ARMPL). Our work exposes the structure of the template-based micro-kernels for ARM Neon (128-bit SIMD), ARM SVE (variable-length SIMD) and Intel AVX512 (512-bit SIMD), showing considerable performance for an NVIDIA Carmel processor (ARM Neon), a Fujitsu A64FX processor (ARM SVE) and on an AMD EPYC 7282 processor (256-bit SIMD). [ABSTRACT FROM AUTHOR]

The resolution of fluorescence microscopy images is limited by the physical properties of light. In the last decade, numerous super-resolution microscopy (SRM) approaches have been proposed to deal with such hindrance. Here we present Mean-Shift Super Resolution (MSSR), a new SRM algorithm based on the Mean Shift theory, which extends spatial resolution of single fluorescence images beyond the diffraction limit of light. MSSR works on low and high fluorophore densities, is not limited by the architecture of the optical setup and is applicable to single images as well as temporal series. The theoretical limit of spatial resolution, based on optimized real-world imaging conditions and analysis of temporal image stacks, has been measured to be 40 nm. Furthermore, MSSR has denoising capabilities that outperform other SRM approaches. Along with its wide accessibility, MSSR is a powerful, flexible, and generic tool for multidimensional and live cell imaging applications. The presented Mean-Shift Super Resolution (MSSR) algorithm can extend spatial resolution within a single microscopy image. Its applicability extends across a wide range of experimental and instrumental configurations and it is compatible with other super-resolution microscopy approaches. [ABSTRACT FROM AUTHOR]

This study explored learning and generalization of a third-order conditional discrimination. Two 8-year-old children learned two auditory–visual conditional discriminations in which they selected visual Japanese syllabic symbols in response to syllables spoken by the experimenter. Then, they learned a third-order conditional discrimination in which they selected between two visual symbols after being exposed to two spoken syllables and one visual symbol. Thereafter, we probed generalization with novel symbols and names by teaching two additional conditional discriminations with Nahuatl symbols and spoken words and probing without reinforcement a new third-order conditional discrimination in which they had to select between two visual Nahuatl symbols after being exposed to two spoken Nahuatl words and one visual Nahuatl symbol. The two children responded in a predicted way to the novel third-order conditional discrimination. The emergent performance was possible because the set of relations established among the stimuli of the third-order conditional discrimination with Japanese syllables was analogous to the set of relations established among the stimuli of the third-order conditional discriminations with Nahuatl words. These results demonstrated a novel type of emergent responding in third-order conditional discrimination with arbitrary relations. [ABSTRACT FROM AUTHOR]

Parkinson's disease dementia (PDD) and dementia with Lewy bodies (DLB) are clinically and neuropathologically highly related α-synucleinopathies that collectively constitute the second leading cause of neurodegenerative dementias. Genetic and neuropathological studies directly implicate α-synuclein (αS) abnormalities in PDD and DLB pathogenesis. However, it is currently unknown how αS abnormalities contribute to memory loss, particularly since forebrain neuronal loss in PDD and DLB is less severe than in Alzheimer's disease. Previously, we found that familial Parkinson's disease-linked human mutant A53T αS causes aberrant localization of the microtubule-associated protein tau to postsynaptic spines in neurons, leading to postsynaptic deficits. Thus, we directly tested if the synaptic and memory deficits in a mouse model of α-synucleinopathy (TgA53T) are mediated by tau. TgA53T mice exhibit progressive memory deficits associated with postsynaptic deficits in the absence of obvious neuropathological and neurodegenerative changes in the hippocampus. Significantly, removal of endogenous mouse tau expression in TgA53T mice (TgA53T/mTau−/−), achieved by mating TgA53T mice to mouse tau-knockout mice, completely ameliorates cognitive dysfunction and concurrent synaptic deficits without affecting αS expression or accumulation of selected toxic αS oligomers. Among the known tau-dependent effects, memory deficits in TgA53T mice were associated with hippocampal circuit remodeling linked to chronic network hyperexcitability. This remodeling was absent in TgA53T/mTau−/− mice, indicating that postsynaptic deficits, aberrant network hyperactivity, and memory deficits are mechanistically linked. Our results directly implicate tau as a mediator of specific human mutant A53T αS-mediated abnormalities related to deficits in hippocampal neurotransmission and suggest a mechanism for memory impairment that occurs as a consequence of synaptic dysfunction rather than synaptic or neuronal loss. We hypothesize that these initial synaptic deficits contribute to network hyperexcitability which, in turn, exacerbate cognitive dysfunction. Our results indicate that these synaptic changes present potential therapeutic targets for amelioration of memory deficits in α-synucleinopathies. [ABSTRACT FROM AUTHOR]

Aquaporins are integral membrane proteins that exchange water and small solutes. They played an important role in the colonisation of terrestrial environments by tetrapod ancestors via the appearance of three exclusive paralogs. Like early tetrapods, mudskippers represent an independent case of amphibious lifestyle evolution that is unparalleled by other extant fish groups. Given this lifestyle parallelism and that aquaporins were relevant for tetrapod terrestrialisation, this study examines the aquaporins in mudskippers to investigate whether similar changes in aquaporins could have possibly occurred during their water-to-land transition. We have catalogued aquaporin genes in four mudskipper genomes and studied their diversity and molecular evolution (including detection of positive selection) in a broad phylogenetic context of vertebrates. Our genomic screening returned 55 aquaporin genes for mudskippers (none of them constituting novel paralogs) that can be assigned to 10 different known classes. We detected signatures of positive selection in AQP10a and AQP11b in mudskippers (both the entire clade and the clade containing the most terrestrial species, implying different evolutionary times). This suggests possible alteration of the molecular function of such paralogs caused by changes at specific protein sequence positions, some of them located in relatively close proximity to parts of the molecule involved in pore formation and substrate selectivity. Given the importance of aquaporins for osmotic regulation in fishes, it might be possible that these selective changes (perhaps allowing permeability to new solutes) could have played a role during the adaptation of mudskippers to an amphibious lifestyle. [ABSTRACT FROM AUTHOR]

In this work we estimated the budgetary impact of the samples produced by the biobank of the "Instituto Nacional de Cancerología" (BT-INCan) to set a recuperation fee from the perspective of the Health Ministry of Mexico. The study is an observational retrospective review of the direct medical costs (DMCs) of the processes involved in cryopreservation of the samples collected, on a per sample basis, including materials, laboratory tests, personnel, and administrative costs. Materials and labor costs were determined by information collected from the BT-INCan. DMCs were provided depending on the type of sample: plasma, tissue and biopsy; they were calculated according to the process required to preserve them. Sensitivity analysis was performed using bootstrap. Recuperation costs ranged from 130 to 155 USD. Costs were considered on a 5-year time frame for the maintenance per sample, which is the average time that a sample is kept in the BT-INCan. The cost analysis is perceived as an approximation to the most adequate recuperation fee per sample needed to guarantee the correct development of the BT-INCan. This work provides a basis and valuable information about costs, to enable several health institutions to strategically plan and manage a biobank or even motivate to establish their own biobank. [ABSTRACT FROM AUTHOR]

Weight loss dietary supplements are used with some frequency by an increasingly overweight population. Some products are not adequately regulated and may pose potential health risks. We report two new cases of acute toxic leukoencephalopathy (ATL) due to the use of a supplement marketed as a thermogenic weight loss aid. ATL is a heterogeneous clinic-radiological entity that has been associated with various compounds, such as chemotherapeutic drugs and immunomodulators. It is characterized by an often reversible periventricular and infratentorial demyelination. The commercialization of non-regulated weight loss products continues to be a health risk in our population. [ABSTRACT FROM AUTHOR]

Computer games have recently shown promise as a diagnostic and treatment tool for psychiatric rehabilitation. This paper examines the potential of combining multiple modalities for detecting affective responses of patients interacting with a simulation built on game technology, aimed at the treatment of mental diagnoses such as post traumatic stress disorder (PTSD). For that purpose, we couple game design and game technology to create a game-based tool for exposure therapy and stress inoculation training that utilizes stress detection for the automatic profiling and potential personalization of PTSD treatments. The PTSD treatment game we designed forces the player to go through various stressful experiences while a stress detection mechanism profiles the severity and type of PTSD by analyzing the physiological responses to those in-game stress elicitors in two separate modalities: skin conductance (SC) and blood volume pulse (BVP). SC is often used to monitor stress as it is connected to the activation of the sympathetic nervous system (SNS). By including BVP into the model we introduce information about para-sympathetic activation, which offers a more complete view of the psycho-physiological experience of the player; in addition, as BVP is also modulated by SNS, a multimodal model should be more robust to changes in each modality due to particular drugs or day-to-day bodily changes. Overall, the study and analysis of 14 PTSD-diagnosed veteran soldiers presented in this paper reveals correspondence between diagnostic standard measures of PTSD severity and SC and BVP responsiveness and feature combinations thereof. The study also reveals that these features are significantly correlated with subjective evaluations of the stressfulness of experiences, represented as pairwise preferences. More importantly, the results presented here demonstrate that using the modalities of SC and BVP captures a more nuanced representation of player stress responses than using SC alone. We conclude that the results support the use of the simulation as a relevant treatment tool for stress inoculation training, and suggest the feasibility of using such a tool to profile PTSD patients. The use of multiple modalities appears to be key for an accurate profiling, although further research and analysis are required to identify the most relevant physiological features for capturing user stress. [ABSTRACT FROM AUTHOR]

Purpose: The aims of this study are to characterize the frequency, density, and distribution of aberrant crypt foci (ACF) and its histological features and to determine the frequency of loss of expression of DNA mismatch repair (MMR) proteins of subjects with hereditary nonpolyposic colorectal cancer (HNPCC) and sporadic colon rectal cancer (CRC). Methods: Patients with HNPCC, first-degree relatives of subjects with HNPCC, sporadic CRC, and average risk subjects of sporadic CRC were included prospectively. Total colonoscopy with chromoendoscopy using methylene blue 0.5% and magnification in the right colon (cecum and 20 cm of the ascending colon) and in the left colon (rectum) was performed; loss of expression of MLH1 and MSH2 was evaluated by immunohistochemistry in confirmed ACF. Results: Fifty-two subjects were included. Thirty-eight of the 119 ACF detected by endoscopy were biopsied. In 14 of the 38 specimens (36.8%), ACF were confirmed by histology (Cohen's kappa, 0.44). In subjects with HNPCC, ACF were identified more frequently in the right segment of the colon than in the left (73.1% vs. 26%); in contrast, ACF predominated in the left segment of the colon (89.3% vs. 10.6%) in subjects with sporadic CRC. There was a loss of MLH1 expression in ACF in subjects with HNPCC. Conclusions: In HNPCC, we found a greater density of ACF in the right colon, and in sporadic CRC, greater density in the left. ACF present loss in the expression of DNA MMR protein and can be used as an early marker in patients with a risk of HNPCC in whom carcinogenesis appears to be accelerated. [ABSTRACT FROM AUTHOR]

Information about interactive virtual environments, such as games, is perceived by users through a virtual camera. While most interactive applications let users control the camera, in complex navigation tasks within 3D environments users often get frustrated with the interaction. In this paper, we propose inclusion of camera control as a vital component of affective adaptive interaction in games. We investigate the impact of camera viewpoints on psychophysiology of players through preference surveys collected from a test game. Data is collected from players of a 3D prey/predator game in which player experience is directly linked to camera settings. Computational models of discrete affective states of fun, challenge, boredom, frustration, excitement, anxiety and relaxation are built on biosignal (heart rate, blood volume pulse and skin conductance) features to predict the pairwise self-reported emotional preferences of the players. For this purpose, automatic feature selection and neuro-evolutionary preference learning are combined providing highly accurate affective models. The performance of the artificial neural network models on unseen data reveals accuracies of above 80% for the majority of discrete affective states examined. The generality of the obtained models is tested in different test-bed game environments and the use of the generated models for creating adaptive affect-driven camera control in games is discussed. [ABSTRACT FROM AUTHOR]

Eighteen undergraduates participated in studies designed to examine the factors that produce transfer of contextual functions to novel stimuli in second-order conditional discriminations. In Study 1, participants selected comparison B1 given sample A1 and comparison B2 given sample A2 in a matching-to-sample procedure. Contextual stimuli X1 or X2 were introduced next. In the presence of X1, selections of B1 given A1 and selections of B2 given A2 were reinforced; in the presence of X2, selections of B2 given A1 and selections of B1 given A2 were reinforced (the XAB relations). The participants then learned novel conditional discriminations with stimuli E and F. When we presented the contextual stimuli X1 and X2 with EF, all 3 participants selected the same and opposite comparisons, respectively, as those previously established in EF training. This performance demonstrated the transfer of functions of contextual stimuli to a conditional discrimination with novel stimuli. In Study 2, participants were exposed to the same teaching and testing conditions as in Study 1, except that stimuli X1 and X2 were introduced in a single phase. Of the 4 participants, 2 showed transfer to a novel conditional discrimination. In Study 3 the selections in the XAB task had no differential consequences. Study 4 replicated the conditions of Study 1, with the exception that the AB phase was not taught. Finally, in Study 5 the EF phase was not taught. All participants failed to demonstrate contextual control of the novel conditional discrimination. These results demonstrated that a complete sequence of conditional-discrimination teaching plays an important role in the acquisition and transfer of contextual control to novel conditional discriminations. These results are relevant to our understanding of instructional control. [ABSTRACT FROM AUTHOR]