Featured
-
-
Article
| Open AccessCommittee machines—a universal method to deal with non-idealities in memristor-based neural networks
Designing reliable and energy-efficient memristor-based artificial neural networks remains a challenge. Here, the authors demonstrate a technology-agnostic approach, committee machines, which increases the inference accuracy of memristive neural networks that suffer from device variability, faulty devices, random telegraph noise and line resistance.
- D. Joksas
- , P. Freitas
- & A. Mehonic
-
Article
| Open AccessReconciling qualitative, abstract, and scalable modeling of biological networks
Boolean Networks are a well-established model of biological networks, but usual interpretations can preclude the prediction of behaviours observed in quantitative systems. The authors introduce Most Permissive Boolean Networks, which are shown not to miss any behaviour achievable by the corresponding quantitative model.
- Loïc Paulevé
- , Juri Kolčák
- & Stefan Haar
-
Article
| Open AccessArtificial intelligence for the detection of COVID-19 pneumonia on chest CT using multinational datasets
Chest CT is emerging as a valuable diagnostic tool for clinical management of COVID-19 associated lung disease. Here, the authors present a multinational study on the application of deep learning algorithms for COVID-19 diagnosis against multiple lung conditions as controls.
- Stephanie A. Harmon
- , Thomas H. Sanford
- & Baris Turkbey
-
Article
| Open AccessBrain-inspired replay for continual learning with artificial neural networks
One challenge that faces artificial intelligence is the inability of deep neural networks to continuously learn new information without catastrophically forgetting what has been learnt before. To solve this problem, here the authors propose a replay-based algorithm for deep learning without the need to store data.
- Gido M. van de Ven
- , Hava T. Siegelmann
- & Andreas S. Tolias
-
Article
| Open AccessSpatiotemporal data analysis with chronological networks
Extracting central information from ever-growing data generated in our lives calls for new data mining methods. Ferreira et al. show a simple model, called chronnets, that can capture frequent patterns, spatial changes, outliers, and spatiotemporal clusters.
- Leonardo N. Ferreira
- , Didier A. Vega-Oliveros
- & Elbert E. N. Macau
-
Article
| Open AccessUsing publicly available satellite imagery and deep learning to understand economic well-being in Africa
It is generally difficult to scale derived estimates and understand the accuracy across locations for passively-collected data sources, such as mobile phones and satellite imagery. Here the authors show that their trained deep learning models are able to explain 70% of the variation in ground-measured village wealth in held-out countries, outperforming previous benchmarks from high-resolution imagery with errors comparable to that of existing ground data.
- Christopher Yeh
- , Anthony Perez
- & Marshall Burke
-
Article
| Open AccessAccurate deep neural network inference using computational phase-change memory
Designing deep learning inference hardware based on in-memory computing remains a challenge. Here, the authors propose a strategy to train ResNet-type convolutional neural networks which results in reduced accuracy loss when transferring weights to in-memory computing hardware based on phase-change memory.
- Vinay Joshi
- , Manuel Le Gallo
- & Evangelos Eleftheriou
-
Article
| Open AccessPlaying games with multiple access channels
Multiple access channels model communication from multiple independent senders to a common receiver. By drawing a connection to the study of classical and quantum correlations using nonlocal games, Leditzky et al. reveal remarkably complex behaviour of the entanglement-assisted and unassisted information transmission capabilities of a multiple access channel.
- Felix Leditzky
- , Mohammad A. Alhejji
- & Graeme Smith
-
Article
| Open AccessA programmable chemical computer with memory and pattern recognition
Unconventional computing architectures might outperform current ones, but their realization has been limited to solving simple specific problems. Here, a network of interconnected Belousov-Zhabotinski reactions, operated by independent magnetic stirrers, performs encoding/decoding operations and data storage.
- Juan Manuel Parrilla-Gutierrez
- , Abhishek Sharma
- & Leroy Cronin
-
Article
| Open AccessComplexity control by gradient descent in deep networks
Understanding the underlying mechanisms behind the successes of deep networks remains a challenge. Here, the author demonstrates an implicit regularization in training deep networks, showing that the control of complexity in the training is hidden within the optimization technique of gradient descent.
- Tomaso Poggio
- , Qianli Liao
- & Andrzej Banburski
-
Article
| Open AccessProbing the physical limits of reliable DNA data retrieval
The physical limits and reliability of PCR-based random access of DNA encoded data is unknown. Here the authors demonstrate reliable file recovery from as few as ten copies per sequence, providing a data density limit of 17 exabytes per gram.
- Lee Organick
- , Yuan-Jyue Chen
- & Luis Ceze
-
Article
| Open AccessAutomated acquisition of explainable knowledge from unannotated histopathology images
Technologies for acquiring explainable features from medical images need further development. Here, the authors report a deep learning based automated acquisition of explainable features from pathology images, and show a higher accuracy of their method as compared to pathologist based diagnosis of prostate cancer recurrence.
- Yoichiro Yamamoto
- , Toyonori Tsuzuki
- & Go Kimura
-
Perspective
| Open AccessHierarchical motor control in mammals and machines
Recent research in motor neuroscience has focused on optimal feedback control of single, simple tasks while robotics and AI are making progress towards flexible movement control in complex environments employing hierarchical control strategies. Here, the authors argue for a return to hierarchical models of motor control in neuroscience.
- Josh Merel
- , Matthew Botvinick
- & Greg Wayne
-
Article
| Open AccessThe Eighty Five Percent Rule for optimal learning
Is there an optimum difficulty level for training? In this paper, the authors show that for the widely-used class of stochastic gradient-descent based learning algorithms, learning is fastest when the accuracy during training is 85%.
- Robert C. Wilson
- , Amitai Shenhav
- & Jonathan D. Cohen
-
Article
| Open AccessInferring neural signalling directionality from undirected structural connectomes
Neural signalling is directional, but non-invasive neuroimaging methods are unable to map directed connections between brain regions. Here, the authors show how network communication measures can be used to infer signalling directionality from the undirected topology of brain structural connectomes.
- Caio Seguin
- , Adeel Razi
- & Andrew Zalesky
-
Perspective
| Open AccessA critique of pure learning and what artificial neural networks can learn from animal brains
Recent gains in artificial neural networks rely heavily on large amounts of training data. Here, the author suggests that for AI to learn from animal brains, it is important to consider that animal behaviour results from brain connectivity specified in the genome through evolution, and not due to unique learning algorithms.
- Anthony M. Zador
-
Article
| Open AccessGraph dynamical networks for unsupervised learning of atomic scale dynamics in materials
Understanding local dynamical processes in materials is challenging due to the complexity of the local atomic environments. Here the authors propose a graph dynamical networks approach that is shown to learn the atomic scale dynamics in arbitrary phases and environments from molecular dynamics simulations.
- Tian Xie
- , Arthur France-Lanord
- & Jeffrey C. Grossman
-
Article
| Open AccessA soft photopolymer cuboid that computes with binary strings of white light
Some next-generation computing may be based in physical systems that respond directly and reciprocally to environmental stimuli. Here, the authors describe a photoresponsive material that autonomously performs computations with incident beams of incoherent white light.
- Alexander D. Hudson
- , Matthew R. Ponte
- & Kalaichelvi Saravanamuttu
-
Article
| Open AccessHumans can decipher adversarial images
Convolutional Neural Networks (CNNs) have reached human-level benchmarks in classifying images, but they can be “fooled” by adversarial examples that elicit bizarre misclassifications from machines. Here, the authors show how humans can anticipate which objects CNNs will see in adversarial images.
- Zhenglong Zhou
- & Chaz Firestone
-
Article
| Open AccessUnmasking Clever Hans predictors and assessing what machines really learn
Nonlinear machine learning methods have good predictive ability but the lack of transparency of the algorithms can limit their use. Here the authors investigate how these methods approach learning in order to assess the dependability of their decision making.
- Sebastian Lapuschkin
- , Stephan Wäldchen
- & Klaus-Robert Müller
-
Article
| Open AccessA continuous-time MaxSAT solver with high analog performance
Continuous-time computation paradigm could represent a viable alternative to the standard digital one when dealing with certain classes of problems. Here, the authors propose a generalised version of a continuous-time solver and simulate its performances in solving MaxSAT and two-colour Ramsey problems.
- Botond Molnár
- , Ferenc Molnár
- & Mária Ercsey-Ravasz
-
Article
| Open AccessMaster clinical medical knowledge at certificated-doctor-level with deep learning model
AI is used increasingly in medical diagnostics. Here, the authors present a deep learning model that masters medical knowledge, demonstrated by it having passed the written test of the 2017 National Medical Licensing Examination in China, and can provide help with clinical diagnosis based on electronic health care records.
- Ji Wu
- , Xien Liu
- & Ping Lv
-
Article
| Open AccessUncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference
Progressive diseases tend to be heterogeneous in their underlying aetiology mechanism, disease manifestation, and disease time course. Here, Young and colleagues devise a computational method to account for both phenotypic heterogeneity and temporal heterogeneity, and demonstrate it using two neurodegenerative disease cohorts.
- Alexandra L Young
- , Razvan V Marinescu
- & Ansgar J Furst
-
Article
| Open AccessFinding any Waldo with zero-shot invariant and efficient visual search
Visual search requires recognizing an object “invariantly”, despite changes in its appearance. Here, the authors show that humans can efficiently and invariantly search for objects in complex scenes and introduce a biologically-inspired zero-shot model that captures human eye movements during search.
- Mengmi Zhang
- , Jiashi Feng
- & Gabriel Kreiman
-
Article
| Open AccessCapacitive neural network with neuro-transistors
Though memristors can potentially emulate neuron and synapse functionality, useful signal energy is lost to Joule heating. Here, the authors demonstrate neuro-transistors with a pseudo-memcapacitive gate that actively process signals via energy-efficient capacitively-coupled neural networks.
- Zhongrui Wang
- , Mingyi Rao
- & J. Joshua Yang
-
Article
| Open AccessA molecular neuromorphic network device consisting of single-walled carbon nanotubes complexed with polyoxometalate
Neuromorphic hardware is based on principles of neuroscience, and has the potential to provide higher-level brain functions. Here, the authors develop a neuromorphic network device, constructed from single-walled carbon nanotubes and polyoxometalate, that mimics nerve impulse generation.
- Hirofumi Tanaka
- , Megumi Akai-Kasaya
- & Takuji Ogawa
-
Article
| Open AccessScalable training of artificial neural networks with adaptive sparse connectivity inspired by network science
Artificial neural networks are artificial intelligence computing methods which are inspired by biological neural networks. Here the authors propose a method to design neural networks as sparse scale-free networks, which leads to a reduction in computational time required for training and inference.
- Decebal Constantin Mocanu
- , Elena Mocanu
- & Antonio Liotta
-
Article
| Open AccessEfficient and self-adaptive in-situ learning in multilayer memristor neural networks
Memristor-based neural networks hold promise for neuromorphic computing, yet large-scale experimental execution remains difficult. Here, Xia et al. create a multi-layer memristor neural network with in-situ machine learning and achieve competitive image classification accuracy on a standard dataset.
- Can Li
- , Daniel Belkin
- & Qiangfei Xia
-
Article
| Open AccessMassive mining of publicly available RNA-seq data from human and mouse
Publicly available RNA-seq data is provided mostly in raw form, resulting in a barrier for integrative analyses. Here, Lachmann et al. develop a high-throughput processing infrastructure and search database (ARCHS4) that provides processed RNA-seq data for 187,946 publicly available mouse and human samples to support exploration and reuse.
- Alexander Lachmann
- , Denis Torre
- & Avi Ma’ayan
-
Article
| Open AccessMemory effects can make the transmission capability of a communication channel uncomputable
In information theory one is interested in how much information can be reliably sent over noisy communication channels. Here the authors show that for channels with memory the optimal rate of information transmission is uncomputable.
- David Elkouss
- & David Pérez-García
-
Article
| Open AccessInput–output maps are strongly biased towards simple outputs
Algorithmic information theory measures the complexity of strings. Here the authors provide a practical bound on the probability that a randomly generated computer program produces a given output of a given complexity and apply this upper bound to RNA folding and financial trading algorithms.
- Kamaludin Dingle
- , Chico Q. Camargo
- & Ard A. Louis
-
Article
| Open AccessOptimal compressed representation of high throughput sequence data via light assembly
Increase in high throughput sequencing (HTS) data warrants compression methods to facilitate better storage and communication. Here, Ginart et al. introduce Assembltrie, a reference-free compression tool which is guaranteed to achieve optimality for error-free reads.
- Antonio A. Ginart
- , Joseph Hui
- & David N. Tse
-
Article
| Open AccessPractical device-independent quantum cryptography via entropy accumulation
The security of DIQKD is difficult to prove, as one needs to take into account every possible attack strategy. Here, the authors develop a method to determine the entropy of a system as the sum of the entropies of its parts. Applied to DIQKD, this implies that it suffices to consider i.i.d. attacks.
- Rotem Arnon-Friedman
- , Frédéric Dupuis
- & Thomas Vidick
-
Article
| Open AccessCooperating with machines
Artificial intelligence is now superior to humans in many fully competitive games, such as Chess, Go, and Poker. Here the authors develop a machine-learning algorithm that can cooperate effectively with humans when cooperation is beneficial but nontrivial, something humans are remarkably good at.
- Jacob W. Crandall
- , Mayada Oudah
- & Iyad Rahwan
-
Article
| Open AccessMergeable nervous systems for robots
Robots that can self-assemble into different morphologies are desired to perform tasks that require different physical capabilities. Mathews et al. design robots whose bodies and control systems can merge and split to form new robots that retain full sensorimotor control and act as a single entity.
- Nithin Mathews
- , Anders Lyhne Christensen
- & Marco Dorigo
-
Article
| Open AccessQuantum vertex model for reversible classical computing
Solutions of computations can be encoded in the ground state of many-body spin models. Here the authors show that solutions to generic reversible classical computations can be encoded in the ground state of a vertex model, which can be reached without finite temperature phase transitions.
- C. Chamon
- , E. R. Mucciolo
- & Z.-C. Yang
-
Article
| Open AccessUnravelling raked linear dunes to explain the coexistence of bedforms in complex dunefields
Raked linear dunes are a rare dune type, but the mechanisms for growth have not been constrained. Here, the authors show that a tridirectional wind regime is required to enable this extremely rare dune type to develop, where the raked pattern may develop preferentially on the leeward side.
- Ping Lü
- , Clément Narteau
- & Sylvain Courrech du Pont
-
Article
| Open AccessTargeted proteomics identifies liquid-biopsy signatures for extracapsular prostate cancer
Proteomic technologies are capable of identifying thousands of proteins in biological samples, but biomarker applications are lagging. Here the authors use Multiple Reaction Monitoring Mass Spectrometry to delineate peptide signatures that accurately distinguish between defined prostate cancer patient risk groups.
- Yunee Kim
- , Jouhyun Jeon
- & Thomas Kislinger
-
Article
| Open AccessAn event-based architecture for solving constraint satisfaction problems
Constraint satisfaction problems are typically solved using conventional von Neumann computing architectures, which are however ill-suited to solving them. Here, the authors present a prototype for an event-based architecture that yield state of the art performance on random SAT problems.
- Hesham Mostafa
- , Lorenz K. Müller
- & Giacomo Indiveri