scholarly journals Communicating artificial neural networks develop efficient color-naming systems

2021 ◽  
Vol 118 (12) ◽  
pp. e2016569118
Author(s):  
Rahma Chaabouni ◽  
Eugene Kharitonov ◽  
Emmanuel Dupoux ◽  
Marco Baroni
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Message Passing ◽  
Communication Systems ◽  
Color Naming ◽  
Low Complexity ◽  
Emergent Systems ◽  
Artificial Neural ◽  
Optimal Efficiency ◽  
The Cost

Words categorize the semantic fields they refer to in ways that maximize communication accuracy while minimizing complexity. Focusing on the well-studied color domain, we show that artificial neural networks trained with deep-learning techniques to play a discrimination game develop communication systems whose distribution on the accuracy/complexity plane closely matches that of human languages. The observed variation among emergent color-naming systems is explained by different degrees of discriminative need, of the sort that might also characterize different human communities. Like human languages, emergent systems show a preference for relatively low-complexity solutions, even at the cost of imperfect communication. We demonstrate next that the nature of the emergent systems crucially depends on communication being discrete (as is human word usage). When continuous message passing is allowed, emergent systems become more complex and eventually less efficient. Our study suggests that efficient semantic categorization is a general property of discrete communication systems, not limited to human language. It suggests moreover that it is exactly the discrete nature of such systems that, acting as a bottleneck, pushes them toward low complexity and optimal efficiency.

Exploring Structural and Dynamical Properties Microtubules by Means of Artificial Neural Networks

2014 ◽  
pp. 191-203
Author(s):  
R. Pizzi ◽  
S. Fiorentini ◽  
G. Strini ◽  
M. Pregnolato
Keyword(s):  
Neural Networks ◽  
Electromagnetic Field ◽  
Artificial Neural Networks ◽  
Communication Systems ◽  
Dynamical Behaviour ◽  
Information Communication ◽  
Cellular Functions ◽  
Dynamical Properties ◽  
Physical Measures ◽  
Artificial Neural

Microtubules (MTs) are cylindrical polymers of the tubulin dimer, are constituents of all eukaryotic cells cytoskeleton and are involved in key cellular functions and are claimed to be involved as sub-cellular information or quantum information communication systems. The authors evaluated some biophysical properties of MTs by means of specific physical measures of resonance and birefringence in presence of electromagnetic field, on the assumption that when tubulin and MTs show different biophysical behaviours, this should be due to their special structural properties. Actually, MTs are the closest biological equivalent to the well-known carbon nanotubes (CNTs), whose interesting biophysical and quantum properties are due to their peculiar microscopic structure. The experimental results highlighted a physical behaviour of MTs in comparison with tubulin. The dynamic simulation of MT and tubulin subjected to electromagnetic field was performed via MD tools. Their level of self-organization was evaluated using artificial neural networks, which resulted to be an effective method to gather the dynamical behaviour of cellular and non-cellular structures and to compare their physical properties.

Exploring Structural and Dynamical Properties Microtubules by Means of Artificial Neural Networks

2013 ◽  
pp. 78-91
Author(s):  
R. Pizzi ◽  
S. Fiorentini ◽  
G. Strini ◽  
M. Pregnolato
Keyword(s):  
Neural Networks ◽  
Electromagnetic Field ◽  
Artificial Neural Networks ◽  
Communication Systems ◽  
Dynamical Behaviour ◽  
Information Communication ◽  
Cellular Functions ◽  
Dynamical Properties ◽  
Physical Measures ◽  
Artificial Neural

Microtubules (MTs) are cylindrical polymers of the tubulin dimer, are constituents of all eukaryotic cells cytoskeleton and are involved in key cellular functions and are claimed to be involved as sub-cellular information or quantum information communication systems. The authors evaluated some biophysical properties of MTs by means of specific physical measures of resonance and birefringence in presence of electromagnetic field, on the assumption that when tubulin and MTs show different biophysical behaviours, this should be due to their special structural properties. Actually, MTs are the closest biological equivalent to the well-known carbon nanotubes (CNTs), whose interesting biophysical and quantum properties are due to their peculiar microscopic structure. The experimental results highlighted a physical behaviour of MTs in comparison with tubulin. The dynamic simulation of MT and tubulin subjected to electromagnetic field was performed via MD tools. Their level of self-organization was evaluated using artificial neural networks, which resulted to be an effective method to gather the dynamical behaviour of cellular and non-cellular structures and to compare their physical properties.

Application of a Neural Manufacturing Concept to Process Modeling, Monitoring and Control

1995 ◽  
Vol 387 ◽  
Author(s):  
Chi Yung Fu ◽  
Loren Petrich ◽  
Benjamin Law
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Process Modeling ◽  
Lawrence Livermore National Laboratory ◽  
National Laboratory ◽  
Monitoring And Control ◽  
Artificial Neural ◽  
The Cost ◽  
And Control ◽  
Capital Intensive

AbstractThe cost of a fabrication line, such as one in a semiconductor house, has increased dramatically over the years, and it is possibly already past the point that some new start-up company can have sufficient capital to build a new fabrication line. Such capital-intensive manufacturing needs better utilization of resources and management of equipment to maximize its productivity. In order to maximize the return from such a capital-intensive manufacturing line, we need to work on the following: 1) increasing the yield, 2) enhancing the flexibility of the fabrication line, 3) improving quality, and finally 4) minimizing the down time of the processing equipment. Because of the significant advances now made in the fields of artificial neural networks, fuzzy logic, machine learning and genetic algorithms, we advocate the use of these new tools in manufacturing. We term the applications to manufacturing of these and other such tools that mimic human intelligence neural manufacturing. This paper describes the effort at the Lawrence Livermore National Laboratory (LLNL) [1] to use artificial neural networks to address certain semiconductor process modeling, monitoring and control questions.

A low-complexity fuzzy activation function for artificial neural networks

2003 ◽  
Vol 14 (6) ◽  
pp. 1576-1579 ◽  
Author(s):  
E. Soria-Olivas ◽  
J.D. Martin-Guerrero ◽  
G. Camps-Valls ◽  
A.J. Serrano-Lopez ◽  
J. Calpe-Maravilla ◽  
...  
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Low Complexity ◽  
Activation Function ◽  
Artificial Neural

scholarly journals Experimental Implementation of a Neural Network Optical Channel Equalizer in Restricted Hardware Using Pruning and Quantization

2022 ◽  
Author(s):  
Diego Argüello Ron ◽  
Pedro Jorge Freire De Carvalho Sourza ◽  
Jaroslaw E. Prilepsky ◽  
Morteza Kamalian-Kopae ◽  
Antonio Napoli ◽  
...  
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Computational Complexity ◽  
Communication Systems ◽  
Complexity Reduction ◽  
Edge Computing ◽  
Optical Channel ◽  
Raspberry Pi ◽  
Artificial Neural ◽  
The Impact

Abstract The deployment of artificial neural networks-based optical channel equalizers on edge-computing devices is critically important for the next generation of optical communication systems. However, this is a highly challenging problem, mainly due to the computational complexity of the artificial neural networks (NNs) required for the efficient equalization of nonlinear optical channels with large memory. To implement the NN-based optical channel equalizer in hardware, a substantial complexity reduction is needed, while keeping an acceptable performance level. In this work, we address this problem by applying pruning and quantization techniques to an NN-based optical channel equalizer. We use an exemplary NN architecture, the multi-layer perceptron (MLP), and address its complexity reduction for the 30 GBd 1000 km transmission over a standard single-mode fiber. We demonstrate that it is feasible to reduce the equalizer’s memory by up to 87.12%, and its complexity by up to 91.5%, without noticeable performance degradation. In addition to this, we accurately define the computational complexity of a compressed NN-based equalizer in the digital signal processing (DSP) sense and examine the impact of using different CPU and GPU settings on power consumption and latency for the compressed equalizer. We also verify the developed technique experimentally, using two standard edge-computing hardware units: Raspberry Pi 4 and Nvidia Jetson Nano.

Crack Width Estimation Using Feed and Cascade Forward Back Propagation Artificial Neural Networks

2018 ◽  
Vol 786 ◽  
pp. 293-301 ◽  
Author(s):  
Hesham M. Shehata ◽  
Yasser S. Mohamed ◽  
Mohamed Abdellatif ◽  
Taher H. Awad
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Crack Width ◽  
Laser Scanning ◽  
Back Propagation ◽  
Maximum Width ◽  
The Neural Networks ◽  
Maximum Crack ◽  
Artificial Neural ◽  
The Cost

Automatic crack inspection techniques that limit the necessity of human have the potential to lower the cost and time of the process. In this study, a maximum crack width estimation approach is presented. Seventy nine segments of cracks are used for training the neural networks and twenty six segments are used for examination. The maximum width for each segment is measured using laser scanning microscope and segment image is captured and magnified using the microscope camera in order to obtain the extracted crack profile number of pixels. Feed and cascade forward back propagation artificial neural networks are designed and constructed. The input and output for the networks are the crack width in terms of number of pixels and the maximum estimated crack width respectively. It is shown that, the artificial neural networks technique can effectively be used to estimate the crack width. The feedforward back propagation structure which is designed with two layers and training function TRAINLM gives the best results in examination.

Sign in / Sign up

Export Citation Format

Share Document