Search

There are two principle approaches for learning in artificial intelligence: error-driven global learning and neuroscience-oriented local learning. Integrating them into one network may provide complementary learning capabilities for versatile learning scenarios. At the same time, neuromorphic computing holds great promise, but still needs plenty of useful algorithms and algorithm-hardware co-designs to fully exploit its advantages. Here, we present a neuromorphic global-local synergic learning model by introducing a brain-inspired meta-learning paradigm and a differentiable spiking model incorporating neuronal dynamics and synaptic plasticity. It can meta-learn local plasticity and receive top-down supervision information for multiscale learning. We demonstrate the advantages of this model in multiple different tasks, including few-shot learning, continual learning, and fault-tolerance learning in neuromorphic vision sensors. It achieves significantly higher performance than single-learning methods. We further implement the model in the Tianjic neuromorphic platform by exploiting algorithm-hardware co-designs and prove that the model can fully utilize neuromorphic many-core architecture to develop hybrid computation paradigm., Global and local learning represent two distinct approaches to artificial intelligence. In this manuscript, Wu et al present a hybrid learning strategy, drawing from elements of both approaches, and implement it on a co-designed neuromorphic platform.

We previously showed that the adult ocellated lizard skin colour pattern is effectively generated by a stochastic cellular automaton (CA) of skin scales. We additionally suggested that the canonical continuous 2D reaction-diffusion (RD) process of colour pattern development is transformed into this discrete CA by reduced diffusion coefficients at the borders of scales (justified by the corresponding thinning of the skin). Here, we use RD numerical simulations in 3D on realistic lizard skin geometries and demonstrate that skin thickness variation on its own is sufficient to cause scale-by-scale coloration and CA dynamics during RD patterning. In addition, we show that this phenomenon is robust to RD model variation. Finally, using dimensionality-reduction approaches on large networks of skin scales, we show that animal growth affects the scale-colour flipping dynamics by causing a substantial decrease of the relative length scale of the labyrinthine colour pattern of the lizard skin., The adult ocellated lizard skin colour pattern is effectively generated by a stochastic cellular automaton (CA) of skin scales. Here authors use reaction diffusion (RD) numerical simulations in 3D on realistic lizard skin geometries and demonstrate that skin thickness variation on its own is sufficient to cause scale-by-scale coloration and CA dynamics during RD patterning.

Current machine learning classifiers have successfully been applied to whole-genome sequencing data to identify genetic determinants of antimicrobial resistance (AMR), but they lack causal interpretation. Here we present a metabolic model-based machine learning classifier, named Metabolic Allele Classifier (MAC), that uses flux balance analysis to estimate the biochemical effects of alleles. We apply the MAC to a dataset of 1595 drug-tested Mycobacterium tuberculosis strains and show that MACs predict AMR phenotypes with accuracy on par with mechanism-agnostic machine learning models (isoniazid AUC = 0.93) while enabling a biochemical interpretation of the genotype-phenotype map. Interpretation of MACs for three antibiotics (pyrazinamide, para-aminosalicylic acid, and isoniazid) recapitulates known AMR mechanisms and suggest a biochemical basis for how the identified alleles cause AMR. Extending flux balance analysis to identify accurate sequence classifiers thus contributes mechanistic insights to GWAS, a field thus far dominated by mechanism-agnostic results., Current machine learning classifiers have been applied to whole-genome sequencing data to identify determinants of antimicrobial resistance, but they lack interpretability. Here the authors present a metabolic machine learning classifier that uses flux balance analysis to estimate the biochemical effects of alleles.

Standfirst Among the existing machine learning frameworks, reservoir computing demonstrates fast and low-cost training, and its suitability for implementation in various physical systems. This Comment reports on how aspects of reservoir computing can be applied to classical forecasting methods to accelerate the learning process, and highlights a new approach that makes the hardware implementation of traditional machine learning algorithms practicable in electronic and photonic systems.

Natural creatures, from fish and cephalopods to snakes and birds, combine neural control, sensory feedback and compliant mechanics to effectively operate across dynamic, uncertain environments. In order to facilitate the understanding of the biophysical mechanisms at play and to streamline their potential use in engineering applications, we present here a versatile numerical approach to the simulation of musculoskeletal architectures. It relies on the assembly of heterogenous, active and passive Cosserat rods into dynamic structures that model bones, tendons, ligaments, fibers and muscle connectivity. We demonstrate its utility in a range of problems involving biological and soft robotic scenarios across scales and environments: from the engineering of millimeter-long bio-hybrid robots to the synthesis and reconstruction of complex musculoskeletal systems. The versatility of this methodology offers a framework to aid forward and inverse bioengineering designs as well as fundamental discovery in the functioning of living organisms., Natural creatures, from fish to snake and birds, combine neural control, sensory feedback and compliant mechanics to operate across uncertain environments. Here the authors present a versatile modeling approach to the dynamic simulation of their architectures based on the assembly of Cosserat rods.

Forecasting is beginning to be integrated into decision-making processes for infectious disease outbreak response. We discuss how technologies could accelerate the adoption of forecasting among public health practitioners, improve epidemic management, save lives, and reduce the economic impact of outbreaks.

Visual stimuli evoke heterogeneous responses across nearby neural populations. These signals must be locally integrated to contribute to perception, but the principles underlying this process are unknown. Here, we exploit the systematic organization of orientation preference in macaque primary visual cortex (V1) and perform causal manipulations to examine the limits of signal integration. Optogenetic stimulation and visual stimuli are used to simultaneously drive two neural populations with overlapping receptive fields. We report that optogenetic stimulation raises firing rates uniformly across conditions, but improves the detection of visual stimuli only when activating cells that are preferentially-tuned to the visual stimulus. Further, we show that changes in correlated variability are exclusively present when the optogenetically and visually-activated populations are functionally-proximal, suggesting that correlation changes represent a hallmark of signal integration. Our results demonstrate that information from functionally-proximal neurons is pooled for perception, but functionally-distal signals remain independent., Primary visual cortical neurons exhibit diverse responses to visual stimuli yet how these signals are integrated during visual perception is not well understood. Here, the authors show that optogenetic stimulation of neurons situated near the visually‐driven population leads to improved orientation detection in monkeys through changes in correlated variability.

Computational flattening algorithms have been successfully applied to X-ray microtomography scans of damaged historical documents, but have so far been limited to scrolls, books, and documents with one or two folds. The challenge tackled here is to reconstruct the intricate folds, tucks, and slits of unopened letters secured shut with “letterlocking,” a practice—systematized in this paper—which underpinned global communications security for centuries before modern envelopes. We present a fully automatic computational approach for reconstructing and virtually unfolding volumetric scans of a locked letter with complex internal folding, producing legible images of the letter’s contents and crease pattern while preserving letterlocking evidence. We demonstrate our method on four letterpackets from Renaissance Europe, reading the contents of one unopened letter for the first time. Using the results of virtual unfolding, we situate our findings within a novel letterlocking categorization chart based on our study of 250,000 historical letters., Here, the authors present a fully automatic computational approach for reconstructing and virtually unfolding volumetric scans of locked letters with complex internal folding, producing legible images of the letter’s contents and crease pattern while preserving letterlocking evidence.

Variational quantum algorithms (VQAs) optimize the parameters θ of a parametrized quantum circuit V(θ) to minimize a cost function C. While VQAs may enable practical applications of noisy quantum computers, they are nevertheless heuristic methods with unproven scaling. Here, we rigorously prove two results, assuming V(θ) is an alternating layered ansatz composed of blocks forming local 2-designs. Our first result states that defining C in terms of global observables leads to exponentially vanishing gradients (i.e., barren plateaus) even when V(θ) is shallow. Hence, several VQAs in the literature must revise their proposed costs. On the other hand, our second result states that defining C with local observables leads to at worst a polynomially vanishing gradient, so long as the depth of V(θ) is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathcal{O}}(\mathrm{log}\,n)$$\end{document}O(logn). Our results establish a connection between locality and trainability. We illustrate these ideas with large-scale simulations, up to 100 qubits, of a quantum autoencoder implementation., Parametrised quantum circuits are a promising hybrid classical-quantum approach, but rigorous results on their effective capabilities are rare. Here, the authors explore the feasibility of training depending on the type of cost functions, showing that local ones are less prone to the barren plateau problem.

Causal inference via Mendelian randomization requires making strong assumptions about horizontal pleiotropy, where genetic instruments are connected to the outcome not only through the exposure. Here, we present causal Graphical Analysis Using Genetics (cGAUGE), a pipeline that overcomes these limitations using instrument filters with provable properties. This is achievable by identifying conditional independencies while examining multiple traits. cGAUGE also uses ExSep (Exposure-based Separation), a novel test for the existence of causal pathways that does not require selecting instruments. In simulated data we illustrate how cGAUGE can reduce the empirical false discovery rate by up to 30%, while retaining the majority of true discoveries. On 96 complex traits from 337,198 subjects from the UK Biobank, our results cover expected causal links and many new ones that were previously suggested by correlation-based observational studies. Notably, we identify multiple risk factors for cardiovascular disease, including red blood cell distribution width., Mendelian randomization is a popular method to detect causal relationships between traits, but can be confounded by instances of horizontal pleiotropy. Here, the authors present a Mendelian randomization workflow which includes causal discovery analysis and filtering of genetic instruments based on their conditional independencies.

The computed tomography angiography (CTA) postprocessing manually recognized by technologists is extremely labor intensive and error prone. We propose an artificial intelligence reconstruction system supported by an optimized physiological anatomical-based 3D convolutional neural network that can automatically achieve CTA reconstruction in healthcare services. This system is trained and tested with 18,766 head and neck CTA scans from 5 tertiary hospitals in China collected between June 2017 and November 2018. The overall reconstruction accuracy of the independent testing dataset is 0.931. It is clinically applicable due to its consistency with manually processed images, which achieves a qualification rate of 92.1%. This system reduces the time consumed from 14.22 ± 3.64 min to 4.94 ± 0.36 min, the number of clicks from 115.87 ± 25.9 to 4 and the labor force from 3 to 1 technologist after five months application. Thus, the system facilitates clinical workflows and provides an opportunity for clinical technologists to improve humanistic patient care., Manual postprocessing of computed tomography angiography (CTA) images is extremely labor intensive and error prone. Here, the authors propose an artificial intelligence reconstruction system that can automatically achieve CTA reconstruction in healthcare services.

The physical architectures of information storage systems often dictate how information is encoded, databases are organized, and files are accessed. Here we show that a simple architecture comprised of a T7 promoter and a single-stranded overhang domain (ss-dsDNA), can unlock dynamic DNA-based information storage with powerful capabilities and advantages. The overhang provides a physical address for accessing specific DNA strands as well as implementing a range of in-storage file operations. It increases theoretical storage densities and capacities by expanding the encodable sequence space and simplifies the computational burden in designing sets of orthogonal file addresses. Meanwhile, the T7 promoter enables repeatable information access by transcribing information from DNA without destroying it. Furthermore, saturation mutagenesis around the T7 promoter and systematic analyses of environmental conditions reveal design criteria that can be used to optimize information access. This simple but powerful ss-dsDNA architecture lays the foundation for information storage with versatile capabilities., The physical architectures of information storage dictate how data is encoded, organised and accessed. Here the authors use DNA with a single-strand overhang as a physical address to access specific data and do in-storage file operations in a scalable and reusuable manner.

While the crisis of statistics has made it to the headlines, that of mathematical modelling hasn’t. Something can be learned comparing the two, and looking at other instances of production of numbers.Sociology of quantification and post-normal science can help.

In epigenome-wide association studies, the measured signals for each sample are a mixture of methylation profiles from different cell types. Current approaches to the association detection claim whether a cytosine-phosphate-guanine (CpG) site is associated with the phenotype or not at aggregate level and can suffer from low statistical power. Here, we propose a statistical method, HIgh REsolution (HIRE), which not only improves the power of association detection at aggregate level as compared to the existing methods but also enables the detection of risk-CpG sites for individual cell types., Cellular heterogeneity is one of the major confounding factors in EWAS studies. Here the authors present a statistical method, HIgh REsolution (HIRE), which enables the detection of risk-CpG sites for individual cell types.

In long-haul optical communication systems, compensating nonlinear effects through digital signal processing (DSP) is difficult due to intractable interactions between Kerr nonlinearity, chromatic dispersion (CD) and amplified spontaneous emission (ASE) noise from inline amplifiers. Optimizing the standard digital back propagation (DBP) as a deep neural network (DNN) with interleaving linear and nonlinear operations for fiber nonlinearity compensation was shown to improve transmission performance in idealized simulation environments. Here, we extend such concepts to practical single-channel and polarization division multiplexed wavelength division multiplexed experiments. We show improved performance compared to state-of-the-art DSP algorithms and additionally, the optimized DNN-based DBP parameters exhibit a mathematical structure which guides us to further analyze the noise statistics of fiber nonlinearity compensation. This machine learning-inspired analysis reveals that ASE noise and incomplete CD compensation of the Kerr nonlinear term produce extra distortions that accumulates along the DBP stages. Therefore, the best DSP should balance between suppressing these distortions and inverting the fiber propagation effects, and such trade-off shifts across different DBP stages in a quantifiable manner. Instead of the common ‘black-box’ approach to intractable problems, our work shows how machine learning can be a complementary tool to human analytical thinking and help advance theoretical understandings in disciplines such as optics., Nonlinear effects provide inherent limitations in fiber optical communications. Here, the authors experimentally demonstrate improved digital back propagation with machine learning and use the results to reveal insights in the optimization of digital signal processing.

Infectious disease modeling has played a prominent role in recent outbreaks, yet integrating these analyses into public health decision-making has been challenging. We recommend establishing ‘outbreak science’ as an inter-disciplinary field to improve applied epidemic modeling.

Self-induced electricity, including lightning, is often observed in dusty atmospheres. However, the physical mechanisms leading to this phenomenon remain elusive as they are remarkably challenging to determine due to the high complexity of the multi-phase turbulent flows involved. Using a fast multi-pole method in direct numerical simulations of homogeneous turbulence laden with hundreds of millions of inertial particles, here we show that mesoscopic electric fields can be aerodynamically created in bi-disperse suspensions of oppositely charged particles. The generation mechanism is self-regulating and relies on turbulence preferentially concentrating particles of one sign in clouds while dispersing the others more uniformly. The resulting electric field varies over much larger length scales than both the mean inter-particle spacing and the size of the smallest eddies. Scaling analyses suggest that low ambient pressures, such as those prevailing in the atmosphere of Mars, increase the dynamical relevance of this aerodynamic mechanism for electrical breakdown., How lightning occurs in dusty atmospheres remains largely unknown because of the complexity of the turbulent flows involved. Di Renzo and Urzay reveal a flow-driven mechanism of charge separation by simulating turbulence laden with hundreds of millions of electrically charged inertial particles.