Focus 26 January 2021

AI and machine learning

Computational advances have enabled the deployment of increasingly complex models, which are applied now to a broad-ranging set of fields. This editorial showcase aims at providing a snapshot of the current tools and challenges that are currently holding the promise to change lives in several ways. Herein, we also highlight research on the underlying pursuit of developing the concept of Artificial Intelligence.

Featured articles

The neurobench framework for benchmarking neuromorphic computing algorithms and systems

Brain-inspired neuromorphic algorithms and systems have shown essential advance in efficiency and capabilities of AI applications. In this Perspective, the authors introduce NeuroBench, a benchmark framework for neuromorphic approaches, collaboratively designed by researchers across industry and academia.
- Jason Yik
- Korneel Van den Berghe
- Vijay Janapa Reddi
PerspectiveOpen Access11 Feb 2025 Nature Communications
Deep reinforcement learning for active flow control in a turbulent separation bubble

The control of turbulent flows is crucial for improving efficiency in engineering systems. Here, authors show in a numerical simulation that using deep reinforcement learning to control surface actuators can successfully mitigate turbulent flow separation, paving the way for new advancements in turbulence control.
- Bernat Font
- Francisco Alcántara-Ávila
- Oriol Lehmkuhl
ArticleOpen Access7 Feb 2025 Nature Communications
Dendrites endow artificial neural networks with accurate, robust and parameter-efficient learning

Artificial neural networks, central to deep learning, are powerful but energy-consuming and prone to overfitting. The authors propose a network design inspired by biological dendrites, which offers better robustness and efficiency, using fewer trainable parameters, thus enhancing precision and resilience in artificial neural networks.
- Spyridon Chavlis
- Panayiota Poirazi
ArticleOpen Access22 Jan 2025 Nature Communications
Principled neuromorphic reservoir computing

Reservoir computing designs recurrent networks that simultaneously buffer inputs and form nonlinear features. Here, authors propose a configurable scheme with better scaling where memory buffer and nonlinear features are in separate circuits. It can be efficiently implemented in neuromorphic hardware.
- Denis Kleyko
- Christopher J. Kymn
- E. Paxon Frady
ArticleOpen Access14 Jan 2025 Nature Communications
Large Language Models lack essential metacognition for reliable medical reasoning

Large Language Models demonstrate expert-level accuracy in medical exams, supporting their potential inclusion in healthcare settings. Here, authors reveal that their metacognitive abilities are underexplored, showing significant gaps in recognizing knowledge limitations, difficulties in modulating their confidence, and challenges in identifying when a problem cannot be answered due to insufficient information.
- Maxime Griot
- Coralie Hemptinne
- Demet Yuksel
ArticleOpen Access14 Jan 2025 Nature Communications
Deep neural networks have an inbuilt Occam’s razor

Deep neural networks predict well despite having many more parameters than data points. The authors show that deep neural networks prefer simpler solutions due to an inbuilt “Occam’s razor”. Favouring simplicity avoids overfitting and captures patterns in data, explaining their success.
- Chris Mingard
- Henry Rees
- Ard A. Louis
ArticleOpen Access14 Jan 2025 Nature Communications
Local kernel renormalization as a mechanism for feature learning in overparametrized convolutional neural networks

Fully connected neural networks in the infinite-width limit often outperform finite-width models, while convolutional networks excel at finite widths. Here, the authors uncover how convolutional networks leverage local, data-dependent kernel renormalization, enabling feature learning to absent in fully connected architectures.
- R. Aiudi
- R. Pacelli
- P. Rotondo
ArticleOpen Access10 Jan 2025 Nature Communications
A scaling law to model the effectiveness of identification techniques

Advanced machine learning techniques have demonstrated the identifiability of human traces online, however, assessment of their potential risks is usually done with small-scale datasets. The authors propose a physics-based approach to evaluate the effectiveness of identification techniques from reported measurements.
- Luc Rocher
- Julien M. Hendrickx
- Yves-Alexandre de Montjoye
ArticleOpen Access9 Jan 2025 Nature Communications
A simple model for Behavioral Time Scale Synaptic Plasticity (BTSP) provides content addressable memory with binary synapses and one-shot learning

The authors propose a model for the process of one-shot learning in the brain. They show that it reproduces the repulsion effect of human memory and provides a blueprint for content-addressable in-memory computing with binary weights.
- Yujie Wu
- Wolfgang Maass
ArticleOpen Access2 Jan 2025 Nature Communications
Matching patients to clinical trials with large language models

Patient recruitment is challenging for clinical trials. Here, the authors introduce TrialGPT, an end-to-end framework for zero-shot patient-to-trial matching with large language models.
- Qiao Jin
- Zifeng Wang
- Zhiyong Lu
ArticleOpen Access18 Nov 2024 Nature Communications
SpliceTransformer predicts tissue-specific splicing linked to human diseases

RNA splicing is crucial for gene expression diversity and can be altered by mutations. Here, the authors present SpTransformer, a deep learning tool that predicts tissue-specific RNA splicing with high accuracy, revealing splicing’s role in pathogenic mutations and aiding clinical interpretations.
- Ningyuan You
- Chang Liu
- Ning Shen
ArticleOpen Access23 Oct 2024 Nature Communications
Evaluating the use of large language models to provide clinical recommendations in the Emergency Department

The emergence of large language models has the potential to transform healthcare. Here, the authors show that, when providing clinical recommendations, these models perform poorly compared to physicians and are overly cautious in their decisions.
- Christopher Y. K. Williams
- Brenda Y. Miao
- Atul J. Butte
ArticleOpen Access8 Oct 2024 Nature Communications
A framework for the emergence and analysis of language in social learning agents

Internal representations are crucial for solving tasks for natural and artificial agents. Here, using reinforcement learning and artificial neural networks, the authors present a framework to analyze the formation of individual and shared abstractions and their impact on task performance.
- Tobias J. Wieczorek
- Tatjana Tchumatchenko
- Maximilian F. Eggl
ArticleOpen Access31 Aug 2024 Nature Communications
Thermodynamics-inspired explanations of artificial intelligence

Predictive machine learning models, while powerful, are often seen as black boxes. Here, the authors introduce a thermodynamics-inspired approach for generating rationale behind their explanations across diverse domains based on the proposed concept of interpretation entropy.
- Shams Mehdi
- Pratyush Tiwary
ArticleOpen Access9 Sep 2024 Nature Communications
Evolving scientific discovery by unifying data and background knowledge with AI Hilbert

Scientific discovery is a highly relevant task in natural sciences, however generating scientifically meaningful laws and determining their consistency remains challenging. The authors introduce an approach that exploits both experimental data and underlying theory in symbolic form to generate formulas that hold scientific significance by solving polynomial optimization problems.
- Ryan Cory-Wright
- Cristina Cornelio
- Lior Horesh
ArticleOpen Access14 Jul 2024 Nature Communications
High-performance deep spiking neural networks with 0.3 spikes per neuron

To address challenges of training spiking neural networks (SNNs) at scale, the authors propose a scalable, approximation-free training method for deep SNNs using time-to-first-spike coding. They demonstrate enhanced performance and energy efficiency for neuromorphic hardware.
- Ana Stanojevic
- Stanisław Woźniak
- Wulfram Gerstner
ArticleOpen Access9 Aug 2024 Nature Communications
Fast and robust analog in-memory deep neural network training

Analog in-memory computing recent hardware implementations focused mainly on accelerating inference deployment. In this work, to improve the training process, the authors propose algorithms for supervised training of deep neural networks on analog in-memory AI accelerator hardware.
- Malte J. Rasch
- Fabio Carta
- Tayfun Gokmen
ArticleOpen Access20 Aug 2024 Nature Communications
A machine learning contest enhances automated freezing of gait detection and reveals time-of-day effects

The absence of an objective way of assessing freezing of gait in Parkinson’s Disease hinders research and care. A machine-learning contest using wearable sensor data delivered detection algorithms with high precision and identified time-of-day effects.
- Amit Salomon
- Eran Gazit
- Jeffrey M. Hausdorff
ArticleOpen Access6 Jun 2024 Nature Communications
Learning nonlinear operators in latent spaces for real-time predictions of complex dynamics in physical systems

Real-time prediction of dynamics for complex physical systems governed by partial differential equations is challenging and computationally expensive. The authors propose a framework for learning neural operators in latent spaces that allows real-time predictions of high-dimensional nonlinear systems.
- Katiana Kontolati
- Somdatta Goswami
- Michael D. Shields
ArticleOpen Access14 Jun 2024 Nature Communications
Curse of rarity for autonomous vehicles

The curse of rarity—the rarity of safety-critical events in high-dimensional variable spaces—presents significant challenges in ensuring the safety of autonomous vehicles using deep learning. Looking at it from distinct perspectives, the authors identify three potential approaches for addressing the issue.
- Henry X. Liu
- Shuo Feng
CommentOpen Access5 Jun 2024 Nature Communications
Reservoir-computing based associative memory and itinerancy for complex dynamical attractors

Artificial associative memories in neural network models have shown ability to store and retrieve static patterns of complex systems, however analysis of dynamic patterns remains challenging. The authors develop a reservoir computing based memory approach for complex multistable dynamical systems.
- Ling-Wei Kong
- Gene A. Brewer
- Ying-Cheng Lai
ArticleOpen Access6 Jun 2024 Nature Communications
Task-oriented machine learning surrogates for tipping points of agent-based models

Detecting tipping points and predicting extreme events from data remains a challenging problem in complex systems related to climate, ecology and finance. The authors propose a data-driven approach to estimate probabilities of rare events in complex systems, and detect tipping points/catastrophic shifts.
- Gianluca Fabiani
- Nikolaos Evangelou
- Ioannis G. Kevrekidis
ArticleOpen Access15 May 2024 Nature Communications
Training an Ising machine with equilibrium propagation

Ising machines have been usually applied to predefined combinatorial problems due to their distinct physical properties. The authors introduce an approach that utilizes equilibrium propagation for the training of Ising machines and achieves high accuracy performance on classification tasks.
- Jérémie Laydevant
- Danijela Marković
- Julie Grollier
ArticleOpen Access30 Apr 2024 Nature Communications
A sparse quantized hopfield network for online-continual memory

Brains and neuromorphic systems learn with local learning rules in online-continual learning scenarios. Designing neural networks that learn effectively under these conditions is challenging. The authors introduce a neural network that implements an effective, principled approach to local, online-continual learning on associative memory tasks.
- Nicholas Alonso
- Jeffrey L. Krichmar
ArticleOpen Access2 May 2024 Nature Communications
Controlling chaos using edge computing hardware

Creating accurate digital twins and controlling nonlinear systems displaying chaotic dynamics is challenging due to high system sensitivity to initial conditions and perturbations. The authors introduce a nonlinear controller for chaotic systems, based on next-generation reservoir computing, with improved accuracy, energy cost, and suitable for implementation with field-programmable gate arrays.
- Robert M. Kent
- Wendson A. S. Barbosa
- Daniel J. Gauthier
ArticleOpen Access8 May 2024 Nature Communications
TacticAI: an AI assistant for football tactics

In modern football games, data-driven analysis serves as a key driver in determining tactics. Wang, Veličković, Hennes et al. develop a geometric deep learning algorithm, named TacticAI, to solve high-dimensional learning tasks over corner kicks and suggest tactics favoured over existing ones 90% of the time.
- Zhe Wang
- Petar Veličković
- Karl Tuyls
ArticleOpen Access19 Mar 2024 Nature Communications
scButterfly: a versatile single-cell cross-modality translation method via dual-aligned variational autoencoders

Technical limitations of simultaneously multi-omics profiling lead to highly noisy multi-modal data and substantial costs. Here, authors proposed a versatile framework and data augmentation schemes, capable of single-cell cross-modality translation and multiple extensive applications.
- Yichuan Cao
- Xiamiao Zhao
- Shengquan Chen
ArticleOpen Access6 Apr 2024 Nature Communications
scCASE: accurate and interpretable enhancement for single-cell chromatin accessibility sequencing data

Single-cell chromatin accessibility sequencing (scCAS) data suffers from high sparsity and dimensionality. Here, authors propose an accurate and interpretable computational framework for enhancing scCAS data that considers cell-to-cell similarity.
- Songming Tang
- Xuejian Cui
- Shengquan Chen
ArticleOpen Access22 Feb 2024 Nature Communications
Higher-order Granger reservoir computing: simultaneously achieving scalable complex structures inference and accurate dynamics prediction

For reservoir computing, improving prediction accuracy while maintaining low computing complexity remains a challenge. Inspired by the Granger causality, Li et al. design a data-driven and model-free framework by integrating the inference process and the inferred results on high-order structures.
- Xin Li
- Qunxi Zhu
- Wei Lin
ArticleOpen Access20 Mar 2024 Nature Communications
Predicting and improving complex beer flavor through machine learning

Perception and appreciation of food flavour depends on many factors, posing a challenge for effective prediction. Here, the authors combine extensive chemical and sensory analyses of 250 commercial Belgian beers to train machine learning models that enable flavour and consumer appreciation prediction.
- Michiel Schreurs
- Supinya Piampongsant
- Kevin J. Verstrepen
ArticleOpen Access26 Mar 2024 Nature Communications
Local prediction-learning in high-dimensional spaces enables neural networks to plan

The task of planning a sequence of actions, and dynamically adjusting the plan in dependence of unforeseen circumstances, remains challenging for artificial intelligence frameworks. The authors introduce a learning approach inspired by cognitive functions, that demonstrates high flexibility and generalization capability in planning tasks, suitable for on-chip learning.
- Christoph Stöckl
- Yukun Yang
- Wolfgang Maass
ArticleOpen Access15 Mar 2024 Nature Communications
Emerging opportunities and challenges for the future of reservoir computing

Reservoir Computing has shown advantageous performance in signal processing and learning tasks due to compact design and ability for fast training. Here, the authors discuss the parallel progress of mathematical theory, algorithm design and experimental realizations of Reservoir Computers, and identify emerging opportunities as well as existing challenges for their large-scale industrial adoption.
- Min Yan
- Can Huang
- Jie Sun
PerspectiveOpen Access6 Mar 2024 Nature Communications
Predicting multiple observations in complex systems through low-dimensional embeddings

Forecasting the future behaviors based on observed data remains a challenging task especially for large nonlinear systems. The authors propose a data-driven approach combining manifold learning and delay embeddings for prediction of dynamics for all components in high-dimensional systems.
- Tao Wu
- Xiangyun Gao
- Jürgen Kurths
ArticleOpen Access12 Mar 2024 Nature Communications
Learning the intrinsic dynamics of spatio-temporal processes through Latent Dynamics Networks

Predicting the evolution of dynamical systems remains challenging, requiring high computational effort or effective reduction of the system into a low-dimensional space. Here, the authors present a data-driven approach for predicting the evolution of systems exhibiting spatiotemporal dynamics in response to external input signals.
- Francesco Regazzoni
- Stefano Pagani
- Alfio Quarteroni
ArticleOpen Access28 Feb 2024 Nature Communications
Hybrid AI-enhanced lightning flash prediction in the medium-range forecast horizon

In this work, authors propose a synergistic approach combining state-of-the-art deterministic forecasting model with artificial intelligence for predicting lightning occurrences. The strategy shows efficient predictive capabilities at medium-range forecast horizons.
- Mattia Cavaiola
- Federico Cassola
- Andrea Mazzino
ArticleOpen Access8 Feb 2024 Nature Communications
Segment anything in medical images

Segmentation is an important fundamental task in medical image analysis. Here the authors show a deep learning model for efficient and accurate segmentation across a wide range of medical image modalities and anatomies.
- Jun Ma
- Yuting He
- Bo Wang
ArticleOpen Access22 Jan 2024 Nature Communications
Connectome-based reservoir computing with the conn2res toolbox

Brain connectivity patterns shape computational capacity of biological neural networks, however mapping empirically measured connectivity to artificial networks remains challenging. The authors present a toolbox for implementing biological neural networks as artificial reservoir networks. The toolbox allows for a variety of empirical/measured connectomes and is equipped with various dynamical systems, and cognitive tasks.
- Laura E. Suárez
- Agoston Mihalik
- Bratislav Misic
ArticleOpen Access22 Jan 2024 Nature Communications
Enhancing geometric representations for molecules with equivariant vector-scalar interactive message passing

Utilising geometric information and reducing computational costs are key challenges in the molecular modelling field. Here, authors propose ViSNet, which efficiently extracts geometric features, accurately predicts molecular properties, and drives simulations with interpretability.
- Yusong Wang
- Tong Wang
- Tie-Yan Liu
ArticleOpen Access5 Jan 2024 Nature Communications
Temporal dendritic heterogeneity incorporated with spiking neural networks for learning multi-timescale dynamics

Brain-inspired spiking neural networks have shown their capability for effective learning, however current models may not consider realistic heterogeneities present in the brain. The authors propose a neuron model with temporal dendritic heterogeneity for improved neuromorphic computing applications.
- Hanle Zheng
- Zhong Zheng
- Lei Deng
ArticleOpen Access4 Jan 2024 Nature Communications
Learning few-shot imitation as cultural transmission

The modelling of human-like behaviours is one of the challenges in the field of Artificial Intelligence. Inspired by experimental studies of cultural evolution, the authors propose a reinforcement learning approach to generate agents capable of real-time third-person imitation.
- Avishkar Bhoopchand
- Bethanie Brownfield
- Lei M. Zhang
ArticleOpen Access28 Nov 2023 Nature Communications
Augmenting interpretable models with large language models during training

Prediction and interpretation tasks may be challenging in high-stakes applications, such as medical decision-making, or systems with compute-limited hardware. The authors introduce an augmented framework for leveraging the knowledge learned by Large Language Models to build interpretable models which are both accurate and efficient.
- Chandan Singh
- Armin Askari
- Jianfeng Gao
ArticleOpen Access30 Nov 2023 Nature Communications
A knowledge-guided pre-training framework for improving molecular representation learning

Accurate property prediction relies on effective molecular representation. Here, the authors introduce KPGT, a knowledge-guided self-supervised framework that improves molecular representation, leading to superior predictions of molecular properties and advancing AI-driven drug discovery.
- Han Li
- Ruotian Zhang
- Jianyang Zeng
ArticleOpen Access21 Nov 2023 Nature Communications
Differentially private knowledge transfer for federated learning

To ensure the privacy of processed data, federated learning approaches involve local differential privacy techniques which however require communicating a large amount of data that needs protection. The authors propose here a framework that uses selected small data to transfer knowledge in federated learning with privacy guarantees.
- Tao Qi
- Fangzhao Wu
- Xing Xie
ArticleOpen Access24 Jun 2023 Nature Communications
On the visual analytic intelligence of neural networks

Visual oddity tasks delve into the visual analytic intelligence of humans, which remained challenging for artificial neural networks. The authors propose here a model with biologically inspired neural dynamics and synthetic saccadic eye movements with improved efficiency and accuracy in solving the visual oddity tasks.
- Stanisław Woźniak
- Hlynur Jónsson
- Evangelos Eleftheriou
ArticleOpen Access25 Sep 2023 Nature Communications
Flight trajectory prediction enabled by time-frequency wavelet transform

Accurate flight trajectory prediction can be a challenging task in air traffic control, especially for maneuver operations. Here, authors develop a time-frequency analysis based on an encoder-decoder neural architecture to estimate wavelet components and model global flight trends and local motion details.
- Zheng Zhang
- Dongyue Guo
- Yi Lin
ArticleOpen Access29 Aug 2023 Nature Communications
Continuous estimation of power system inertia using convolutional neural networks

The increase of intermittent energy sources and renewable energy penetration generally results in reduced overall inertia, making power systems susceptible to disturbances. Here, authors develop an AI-based method to estimate inertia in real-time and test its performance on a heterogeneous power network.
- Daniele Linaro
- Federico Bizzarri
- Angelo M. Brambilla
ArticleOpen Access24 Jul 2023 Nature Communications
Data fusion and multivariate analysis for food authenticity analysis

Using two different mass spectrometric platforms, authors demonstrate how metabolomic data fusion and multivariate analysis can be used to accurately identify the geographic origin and production method of salmon.
- Yunhe Hong
- Nicholas Birse
- Christopher T. Elliott
ArticleOpen Access8 Jun 2023 Nature Communications
Learning how network structure shapes decision-making for bio-inspired computing

Better understanding of a trade-off between the speed and accuracy of decision-making is relevant for mapping biological intelligence to machines. The authors introduce a brain-inspired learning algorithm to uncover dependencies in individual fMRI networks with features of neural activity and predict inter-individual differences in decision-making.
- Michael Schirner
- Gustavo Deco
- Petra Ritter
ArticleOpen Access23 May 2023 Nature Communications
Combining data and theory for derivable scientific discovery with AI-Descartes

Automatic extraction of consistent governing laws from data is a challenging problem. The authors propose a method that takes as input experimental data and background theory and combines symbolic regression with logical reasoning to obtain scientifically meaningful symbolic formulas.
- Cristina Cornelio
- Sanjeeb Dash
- Lior Horesh
ArticleOpen Access12 Apr 2023 Nature Communications
Meta-learning biologically plausible plasticity rules with random feedback pathways

The biological plausibility of backpropagation and its relationship with synaptic plasticity remain open questions. The authors propose a meta-learning approach to discover interpretable plasticity rules to train neural networks under biological constraints. The meta-learned rules boost the learning efficiency via bio-inspired synaptic plasticity.
- Navid Shervani-Tabar
- Robert Rosenbaum
ArticleOpen Access31 Mar 2023 Nature Communications
Negotiation and honesty in artificial intelligence methods for the board game of Diplomacy

Artificial Intelligence has achieved success in a variety of single-player or competitive two-player games with no communication between players. Here, the authors propose an approach where Artificial Intelligence agents have ability to negotiate and form agreements, playing the board game Diplomacy.
- János Kramár
- Tom Eccles
- Yoram Bachrach
ArticleOpen Access6 Dec 2022 Nature Communications
Transfer learning enhanced water-enabled electricity generation in highly oriented graphene oxide nanochannels

Large-scale nanochannel integration and the multi-parameter coupling restrictive influence on electric generation are big challenges for effective energy harvesting from spontaneous water flow within artificial nanochannels. Here, authors apply transfer learning to overcome these and design optimized water-enabled generators.
- Ce Yang
- Haiyan Wang
- Liangti Qu
ArticleOpen Access10 Nov 2022 Nature Communications
Mapping global dynamics of benchmark creation and saturation in artificial intelligence

Recent studies raised concerns over the state of AI benchmarking, reporting issues such as benchmark overfitting, benchmark saturation and increasing centralization of benchmark dataset creation. To facilitate monitoring of the health of the AI benchmarking ecosystem, the authors introduce methodologies for creating condensed maps of the global dynamics of benchmark.
- Simon Ott
- Adriano Barbosa-Silva
- Matthias Samwald
ArticleOpen Access10 Nov 2022 Nature Communications
Adversarial attacks and adversarial robustness in computational pathology

Artificial Intelligence can support diagnostic workflows in oncology, but they are vulnerable to adversarial attacks. Here, the authors show that convolutional neural networks are highly susceptible to white- and black-box adversarial attacks in clinically relevant classification tasks.
- Narmin Ghaffari Laleh
- Daniel Truhn
- Jakob Nikolas Kather
ArticleOpen Access29 Sep 2022 Nature Communications

AI and machine learning

Featured articles

Search

Quick links