Explore the vast range of titles available.

A plethora of research advances have emerged in the fields of optics and photonics that benefit from harnessing the power of machine learning. Specifically, there has been a revival of interest in optical computing hardware due to its potential advantages for machine learning tasks in terms of parallelization, power efficiency and computation speed. Diffractive deep neural networks (D2NNs) form such an optical computing framework that benefits from deep learning-based design of successive diffractive layers to all-optically process information as the input light diffracts through these passive layers. D2NNs have demonstrated success in various tasks, including object classification, the spectral encoding of information, optical pulse shaping and imaging. Here, we substantially improve the inference performance of diffractive optical networks using feature engineering and ensemble learning. After independently training 1252 D2NNs that were diversely engineered with a variety of passive input filters, we applied a pruning algorithm to select an optimized ensemble of D2NNs that collectively improved the image classification accuracy. Through this pruning, we numerically demonstrated that ensembles of N = 14 and N = 30 D2NNs achieve blind testing accuracies of 61.14 ± 0.23% and 62.13 ± 0.05%, respectively, on the classification of CIFAR-10 test images, providing an inference improvement of >16% compared to the average performance of the individual D2NNs within each ensemble. These results constitute the highest inference accuracies achieved to date by any diffractive optical neural network design on the same dataset and might provide a significant leap to extend the application space of diffractive optical image classification and machine vision systems.Diffractive networks light the way for better optical image classificationScientists in USA have demonstrated significant improvements in the performance of diffractive optical networks, marking a major step forward for their use in optics-based computation and machine learning. There is renewed interest in optical computing hardware due to its potential advantages, including parallelization, power efficiency, and computation speed. Diffractive optical networks utilize deep learning-based design of successive diffractive layers to all-optically process information as the light is transmitted from the input to the output plane. Led by Aydogan Ozcan, a team of researchers from University of California, Los Angeles has significantly improved the statistical inference performance of diffractive optical networks using feature engineering and ensemble learning. Using a pruning algorithm, they searched through 1,252 unique diffractive networks to design ensembles of desired size that substantially improve the overall system’s all-optical image classification accuracy.

Complex field imaging, which captures both the amplitude and phase information of input optical fields or objects, can offer rich structural insights into samples, such as their absorption and refractive index distributions. However, conventional image sensors are intensity-based and inherently lack the capability to directly measure the phase distribution of a field. This limitation can be overcome using interferometric or holographic methods, often supplemented by iterative phase retrieval algorithms, leading to a considerable increase in hardware complexity and computational demand. Here, we present a complex field imager design that enables snapshot imaging of both the amplitude and quantitative phase information of input fields using an intensity-based sensor array without any digital processing. Our design utilizes successive deep learning-optimized diffractive surfaces that are structured to collectively modulate the input complex field, forming two independent imaging channels that perform amplitude-to-amplitude and phase-to-intensity transformations between the input and output planes within a compact optical design, axially spanning ~100 wavelengths. The intensity distributions of the output fields at these two channels on the sensor plane directly correspond to the amplitude and quantitative phase profiles of the input complex field, eliminating the need for any digital image reconstruction algorithms. We experimentally validated the efficacy of our complex field diffractive imager designs through 3D-printed prototypes operating at the terahertz spectrum, with the output amplitude and phase channel images closely aligning with our numerical simulations. We envision that this complex field imager will have various applications in security, biomedical imaging, sensing and material science, among others.A diffractive optical imager simultaneously captures both the amplitude and phase distributions of complex-valued input objects.

Animal cognitive abilities have traditionally been studied in the lab, but studying cognition in nature could provide several benefits including reduced stress and reduced impact on life-history traits. However, it is not yet clear to what extent cognitive abilities can be properly measured in the wild. Here we present the first comparison of the cognitive performance of individuals from the same population, assessed using an identical test, but in contrasting contexts: in the wild vs. in controlled captive conditions. We show that free-ranging great tits (Parus major) perform similarly to deprived, captive birds in a successive spatial reversal-learning task using automated operant devices. In both captive and natural conditions, more than half of birds that contacted the device were able to perform at least one spatial reversal. Moreover, both captive and wild birds showed an improvement of performance over successive reversals, with very similar learning curves observed in both contexts for each reversal. Our results suggest that it is possible to study cognitive abilities of wild animals directly in their natural environment in much the same way that we study captive animals. Such methods open numerous possibilities to study and understand the evolution and ecology of cognition in natural populations.

Diffractive deep neural networks (D2NNs) are composed of successive transmissive layers optimized using supervised deep learning to all-optically implement various computational tasks between an input and output field-of-view. Here, we present a pyramid-structured diffractive optical network design (which we term P-D2NN), optimized specifically for unidirectional image magnification and demagnification. In this design, the diffractive layers are pyramidally scaled in alignment with the direction of the image magnification or demagnification. This P-D2NN design creates high-fidelity magnified or demagnified images in only one direction, while inhibiting the image formation in the opposite direction—achieving the desired unidirectional imaging operation using a much smaller number of diffractive degrees of freedom within the optical processor volume. Furthermore, the P-D2NN design maintains its unidirectional image magnification/demagnification functionality across a large band of illumination wavelengths despite being trained with a single wavelength. We also designed a wavelength-multiplexed P-D2NN, where a unidirectional magnifier and a unidirectional demagnifier operate simultaneously in opposite directions, at two distinct illumination wavelengths. Furthermore, we demonstrate that by cascading multiple unidirectional P-D2NN modules, we can achieve higher magnification factors. The efficacy of the P-D2NN architecture was also validated experimentally using terahertz illumination, successfully matching our numerical simulations. P-D2NN offers a physics-inspired strategy for designing task-specific visual processors.

Extracting the temporal structure of sequences of events is crucial for perception, decision-making, and language processing. Here, we investigate the mechanisms by which the brain acquires knowledge of sequences and the possibility that successive brain responses reflect the progressive extraction of sequence statistics at different timescales. We measured brain activity using magnetoencephalography in humans exposed to auditory sequences with various statistical regularities, and we modeled this activity as theoretical surprise levels using several learning models. Successive brain waves related to different types of statistical inferences. Early post-stimulus brain waves denoted a sensitivity to a simple statistic, the frequency of items estimated over a long timescale (habituation). Mid-latency and late brain waves conformed qualitatively and quantitatively to the computational properties of a more complex inference: the learning of recent transition probabilities. Our findings thus support the existence of multiple computational systems for sequence processing involving statistical inferences at multiple scales.

Decreasing speed and stride length over successive races have been shown to be associated with musculoskeletal injury (MSI) in racehorses, demonstrating the potential for early detection of MSI through longitudinal monitoring of changes in stride characteristics. A machine learning (ML) approach for early detection of MSI, enforced rest, and retirement events using this same horse-level, race-level, and stride characteristic data across all race sectionals was investigated. A CatBoost model using features from the two races prior to an event had the highest classification performance (sensitivity score for MSI, enforced rest and retirement equal to 0.00, 0.58, 0.76, respectively and balanced accuracy score corresponding to 0.44), with scores decreasing for models incorporating windows that included additional races further from the event. Feature importance analysis of ML models demonstrated that retirement was predicted by older age, poor performance, and longer racing career, enforced rest was predicted by younger age and better performance, but was less likely to occur when the stride length is increasing, and MSI predicted by increased number of starters, greater variation in speed and lower percentage of career time at rest. A relatively low classification performance highlights the difficulties in discerning MSI from alternate events using ML. Improved data recording through more thorough assessment and annotation of adverse events is expected to improve the predictability of MSI.

Adult olfactory neurogenesis provides waves of new neurons involved in memory encoding. However, how the olfactory bulb deals with neuronal renewal to ensure the persistence of pertinent memories and the flexibility to integrate new events remains unanswered. To address this issue, mice performed two successive olfactory discrimination learning tasks with varying times between tasks. We show that with a short time between tasks, adult-born neurons supporting the first learning task appear to be highly sensitive to interference. Furthermore, targeting these neurons using selective light-induced inhibition altered memory of this first task without affecting that of the second, suggesting that neurons in their critical period of integration may only support one memory trace. A longer period between the two tasks allowed for an increased resilience to interference. Hence, newly formed adult-born neurons regulate the transience or persistence of a memory as a function of information relevance and retrograde interference. Olfactory bulb neurogenesis raises the question of how persistent olfactory memories are retained while remaining flexible to encode new memories. Here, the authors show that new neurons can only support a single odor memory within their critical period of integration into the circuit.

PurposeThe mobilization of most available hospital resources to manage coronavirus disease 2019 (COVID-19) may have affected the safety of care for non-COVID-19 surgical patients due to restricted access to intensive or intermediate care units (ICU/IMCUs). We estimated excess surgical mortality potentially attributable to ICU/IMCUs overwhelmed by COVID-19, and any hospital learning effects between two successive pandemic waves.MethodsThis nationwide observational study included all patients without COVID-19 who underwent surgery in France from 01/01/2019 to 31/12/2020. We determined pandemic exposure of each operated patient based on the daily proportion of COVID-19 patients among all patients treated within the ICU/IMCU beds of the same hospital during his/her stay. Multilevel models, with an embedded triple-difference analysis, estimated standardized in-hospital mortality and compared mortality between years, pandemic exposure groups, and semesters, distinguishing deaths inside or outside the ICU/IMCUs.ResultsOf 1,870,515 non-COVID-19 patients admitted for surgery in 655 hospitals, 2% died. Compared to 2019, standardized mortality increased by 1% (95% CI 0.6–1.4%) and 0.4% (0–1%) during the first and second semesters of 2020, among patients operated in hospitals highly exposed to pandemic. Compared to the low-or-no exposure group, this corresponded to a higher risk of death during the first semester (adjusted ratio of odds-ratios 1.56, 95% CI 1.34–1.81) both inside (1.27, 1.02–1.58) and outside the ICU/IMCU (1.98, 1.57–2.5), with a significant learning effect during the second semester compared to the first (0.76, 0.58–0.99).ConclusionSignificant excess mortality essentially occurred outside of the ICU/IMCU, suggesting that access of surgical patients to critical care was limited.

Previous experiments found that acceptance of saccharin by rats was reduced if they had prior experience of sucrose or some other highly palatable solution. This study tested whether such successive negative contrast (SNC) effects involve acquisition of an aversion to the new taste. In three experiments, rats were switched from sucrose exposure in Stage 1 to a less palatable solution containing a new taste in Stage 2. In Experiments 1 and 2 , a novel flavor was added to a saccharin solution at the start of Stage 2. In Experiment 1 , preference tests revealed a weak aversion to the added vanilla flavor in the Suc-Sacch group, while in Experiment 2 an aversion was found in the Suc-Sacch group to the salty flavor that was used, compared with controls given access either saccharin or water in Stage 1. In Experiment 3 , the Suc-Quin group, given quinine solution in Stage 2, displayed a greater aversion to quinine than a Water-Quin control group. These results support the suggestion that taste aversion learning plays a role in the initial suppression of intakes in a qualitative consummatory SNC effect. However, in the light of other evidence, it seems that the unusual persistence of successive negative contrast when rats are switched from sucrose to saccharin is not due to a long-lasting reduction in the value of saccharin.

Most educational literature on conceptual change concerns the process by which introductory students acquire scientific knowledge. However, with modern developments in science and technology, the social significance of learning successive theories is steadily increasing, thus opening new areas of interest to discipline-based education research, e.g., quantum logic, quantum information, and communication. Here, we present an initial proposal for modeling the transition from the understanding of a theory to the understanding of its successor and explore its generative potential by applying it to a concrete case—the classical-quantum transition in physics. In pursue of such task, we make coordinated use of contributions from research not only on conceptual change in education, but also on the history and philosophy of science, on the teaching and learning of quantum mechanics, and on mathematics education. By means of analytical instruments developed for characterizing conceptual trajectories at different representational levels, we review empirical literature in the search for the connections between theory change and cognitive demands. The analysis shows a rich landscape of changes and new challenges that are absent in the traditionally considered cases of conceptual change. In order to fully disclose the educational potential of the analysis, we visualize categorical changes by means of dynamic frames, identifying recognizable patterns that answer to students’ need of comparability between the older and the new paradigm. Finally, we show how the frame representation can be used to suggest pattern-dependent strategies to promote the understanding of the new content, and may work as a guide to curricular design.