Processing Paradigm Research Articles

Skimming Digits: Neuromorphic Classification of Spike-Encoded Images.

The growing demands placed upon the field of computer vision have renewed the focus on alternative visual scene representations and processing paradigms. Silicon retinea provide an alternative means of imaging the visual environment, and produce frame-free spatio-temporal data. This paper presents an investigation into event-based digit classification using N-MNIST, a neuromorphic dataset created with a silicon retina, and the Synaptic Kernel Inverse Method (SKIM), a learning method based on principles of dendritic computation. As this work represents the first large-scale and multi-class classification task performed using the SKIM network, it explores different training patterns and output determination methods necessary to extend the original SKIM method to support multi-class problems. Making use of SKIM networks applied to real-world datasets, implementing the largest hidden layer sizes and simultaneously training the largest number of output neurons, the classification system achieved a best-case accuracy of 92.87% for a network containing 10,000 hidden layer neurons. These results represent the highest accuracies achieved against the dataset to date and serve to validate the application of the SKIM method to event-based visual classification tasks. Additionally, the study found that using a square pulse as the supervisory training signal produced the highest accuracy for most output determination methods, but the results also demonstrate that an exponential pattern is better suited to hardware implementations as it makes use of the simplest output determination method based on the maximum value.

A cloud-based system framework for performing online viewing, storage, and analysis on big data of massive BIMs

Abstract This paper presents a cloud-based system framework based on Bigtable and MapReduce as the data storage and processing paradigms for providing a web-based service for viewing, storing, and analyzing massive building information models (BIMs). Cloud and Web 3D technologies were utilized to develop a BIM data center that can handle the big data of massive BIMs using multiple servers in a distributed manner and can be accessed by multiple users to concurrently submit and view BIMs online in 3D. Traditional BIM include only static information such as the geometric parameters, physical properties, and spatial relations for modeling a physical space. In this study, BIM was extended to dynamic BIM, which includes dynamic data such as historical records from the monitoring of the facility environment and usage. Owing to this extension, a dynamic BIM became a parametric model, which can be used to simulate user behaviors. On the client side, this study applied WebGL in the web interface development to achieve the display of BIMs in 3D on browsers. Users can access the services via various online devices anytime and anywhere to view the 3D model online. On the server side, this study used Apache Hadoop, which can utilize multiple servers to provide mass storage spaces in a distributed manner with Bigtable-like structured storage, to establish the BIM data center. A schema for storing the big data of massive dynamic BIMs in Bigtables was proposed. MapReduce, a Hadoop component for the parallel processing of large data sets, was utilized to process big data from dynamic BIMs. A big data analysis framework to effectively retrieve and calculate required information from dynamic BIMs in the data center for various applications by MapReduce distributed computing was proposed this study. We provide principle and architecture of the proposed framework along with its experimental assessment. The results confirmed that scalable and reliable management of massive BIMs can be achieved using the proposed framework.

First results from the generation and assimilation of 10-minute atmospheric motion vectors in the Australian region

Ten-minute interval infrared and visible satellite images, available from MTSAR-1R (Himawari-6) for a limited period over and around Australia, have been used to generate atmospheric motion vectors. These vectors—which were available every 10 minutes—have been quality controlled, error characterised and assimilated into the Australian Bureau of Meteorology's next generation operational regional forecast model as part of the new operational database. Results from this study indicate that this high temporal resolution imagery has the ability to produce high spatial and temporal density atmospheric motion vectors, and the quality of these vectors is such that they have the potential to improve numerical analysis and prognosis. In late 2013 and early 2014 the Japanese Meteorological Agency (JMA) provided several months of MTSat-1R rapid scan imagery in support of the 2014 High Altitude Ice Crystal (HAIC) – High Ice Water Content (HIWC) field campaign in the Darwin area. This 10-minute imagery was also provided in support of activities in preparation for Himawari-8. The imagery was renavigated, calibrated and used to generate atmospheric motion vectors (AMVs) over much of the Australian region including the HAIC-HIWC test area. Each vector was error characterised and assigned an expected error. In preparation for the operational assimilation of 10-minute data from the full disk footprint of Himawari-8, these high temporal and spatial resolution data generated over the Australian region test area were combined with the operational database—including high spatial but lower temporal density local Bureau and also JMA AMV data from MTSat-2—to provide forecasts using the next generation operational regional forecast model ACCESS APS2. This model will be used during the introduction of operational Himawari-8 data. First results from these tests indicate the 10-minute data have the potential to produce high density, high quality AMVs which will improve model initialisation and forecasts. This includes cases associated with extreme weather. The results have also provided indications of appropriate timing, data selection and application methods for the effective use of these high temporal resolution data. The quality and density of the 10-minute MTSat-1R data have also been compared to that available from imagery using a new processing paradigm which is to be used in future for the operational Himawari-8 AMV generation.

Evolving granular analytics for interval time series forecasting

As a paradigm of data processing, granular computation concerns processing complex data entities called granules, which arise from data abstraction and derivation of knowledge from data. This paper addresses granular computation within the framework of interval time series forecasting and evolving intelligent systems. It develops a generalized interval evolving possibilistic fuzzy modeling algorithm as an analytics tool capable to process interval data stream and to produce interval forecasts. The algorithm uses interval arithmetic in its processing steps, employs the notion of data density to adapt the current forecasting model as data are input, and computes (dis)similarity between interval data using the Hausdorff distance. Computational experiments include forecasting of an interval time series data produced by a synthetic time-varying model with parameter drift, and forecasting of financial interval time series using actual daily minimum and maximum values of the US and Brazilian main equity indexes, S&P 500 and IBOVESPA, respectively. The results suggest that the generalized interval evolving possibilistic fuzzy algorithm is highly effective to model and forecast interval time series. It has comparable or better performance than alternative evolving fuzzy and benchmark interval-based approaches.

Innerstädtische Flusslandschaften als integriertes Handlungsfeld – Planungspraktische Einflussfaktoren der Koordination und Kooperation

In urban riverscapes different significant land uses and interests come together, for instance flood protection, environmental protection, tourism or local recreation, nature or monument conservation, waterfront development or climate change mitigation. Accordingly, planning for integrated spatial developments is required for which actors of different administrative units need to collaborate. This paper first shows the institutional requirements for a collaborative approach in riverscapes. Therefore the administrative organization and the legally based objectives and planning instruments are analyzed. Based on planning culture as analytical perspective, further informal and more cultural influence factors of cooperation are included. Following the examples of integrated development planning in the riverscapes of Nuremberg and Ratisbon, some interesting characteristics of two key organizations—water management authorities and city administrations—were pointed out by highlighting their guiding principles, their organization-specific perceptions, and validations as well as their particular self-conception. It becomes apparent, that the so called common process paradigm—how the cooperation in urban riverscapes is implemented—plays a crucial role in the way planning processes take place. Last but not least it influences the spatial development and design itself. Relevant parameters for designing and steering integrated planning processes are identified.

Homoepitaxial graphene tunnel barriers for spin transport

Tunnel barriers are key elements for both charge-and spin-based electronics, offering devices with reduced power consumption and new paradigms for information processing. Such devices require mating dissimilar materials, raising issues of heteroepitaxy, interface stability, and electronic states that severely complicate fabrication and compromise performance. Graphene is the perfect tunnel barrier. It is an insulator out-of-plane, possesses a defect-free, linear habit, and is impervious to interdiffusion. Nonetheless, true tunneling between two stacked graphene layers is not possible in environmental conditions usable for electronics applications. However, two stacked graphene layers can be decoupled using chemical functionalization. Here, we demonstrate that hydrogenation or fluorination of graphene can be used to create a tunnel barrier. We demonstrate successful tunneling by measuring non-linear IV curves and a weakly temperature dependent zero-bias resistance. We demonstrate lateral transport of spin currents in non-local spin-valve structures, and determine spin lifetimes with the non-local Hanle effect. We compare the results for hydrogenated and fluorinated tunnel and we discuss the possibility that ferromagnetic moments in the hydrogenated graphene tunnel barrier affect the spin transport of our devices.

A Review of Sentiment Analysis in Twitter Data Using Hadoop

Twitter is an online social networking site which contains rich amount of data that can be a structured, semi-structured and un-structured data. In this work, a method which performs classification of tweet sentiment in Twitter is discussed. To improve its scalability and efficiency, it is proposed to implement the work on Hadoop Ecosystem, a widely-adopted distributed processing platform using the Map Reduce parallel processing paradigm. Finally, extensive experiments will be conducted on real-world data sets, with an expectation to achieve comparable or greater accuracy than the proposed techniques in literature.

In situ vapor-deposited parylene substrates for ultra-thin, lightweight organic solar cells

We fabricate the thinnest (1.3 μm) and lightest (3.6 g/m2) solar cells yet demonstrated, with weight-specific power exceeding 6 W/g, in order to illustrate the lower limits of substrate thickness and materials use achievable with a new processing paradigm. Our fabrication process uniquely starts with growth of an ultra-thin flexible polymer substrate in vacuum, followed by deposition of electrodes and photoactive layers in situ. With this process sequence, the entire cell—from transparent substrate to active layers to encapsulation—can be fabricated at room temperature without solvents and without breaking vacuum, avoiding exposure to dust and other contaminants, and minimizing damage risk associated with handling of thin substrates. We use in situ vapor-phase growth of smooth, transparent, and flexible parylene-C films to produce ultra-thin, lightweight molecular organic solar cells as thin as 2.3 μm including encapsulation with a second parylene-C film. These parylene-based devices exhibit power conversion efficiencies and fabrication yields comparable to glass-based cells. Flexible solar cells on parylene membranes can be seamlessly adhered to a variety of solid surfaces to provide additive solar power.

An on-line algorithm for cluster detection of mobile nodes through complex event processing

Clusters of mobile elements, such as vehicles and humans, are a common mobility pattern of interest for many applications. The on-line detection of them from large position streams of mobile entities is a challenging task because it requires algorithms that are capable of continuously and efficiently processing the high volume of position updates in a timely manner. Currently, the majority of approaches for cluster detection operate in batch mode, where position updates are recorded during time periods of certain length and then batch processed by an external routine, thus delaying the result of the cluster detection until the end of the time period. However, if the monitoring application requires results at a higher frequency than the one delivered by batch algorithms, then results might not reflect the current clustering state of the entities. To overcome this limitation, in this paper we propose DG2CEP, an algorithm that combines the well-known density-based clustering algorithm DBSCAN with the data stream processing paradigm Complex Event Processing (CEP) to achieve continuous, on-line detection of clusters. Our experiments with synthetic and real world datasets indicate that DG2CEP is able to detect the formation and dispersion of clusters with small latency and higher similarity to DBSCAN׳s output than batch-based approaches.

Neural Modulation in Aversive Emotion Processing: An Independent Component Analysis Study.

Emotional processing has an important role in social interaction. We report the findings about the Independent Component Analysis carried out on a fMRI set obtained with a paradigm of face emotional processing. The results showed that an independent component, mainly cerebellar-medial-frontal, had a positive modulation associated with fear processing. Also, another independent component, mainly parahippocampal-prefrontal, showed a negative modulation that could be associated with implicit reappraisal of emotional stimuli. Independent Component Analysis could serve as a method to understand complex cognitive processes and their underlying neural dynamics.

DynamoGraph: extending the Pregel paradigm for large-scale temporal graph processing

Large-scale temporal graphs can serve as a model in many application scenarios. Owing to the popularity of online social networks and increased research interest in gathering and analysing data about human behaviour and interaction reality mining temporal graphs gain attraction in social network analysis and more specifically in the analysis of dynamic processes in social networks. Originally developed for web-graph analysis, systems supporting a Pregel-style processing paradigm can be used to process large-scale social network data. However, current methods for social network analysis require data to be processed off-line, or lack support for temporal graphs, or support datasets of limited size only. In this work we present a cloud-based distributed processing framework extending the Pregel paradigm for large-scale temporal graphs. By using computing resources in the cloud this system is scalable and prepared to scale for the massive datasets that occur in social network analysis.

A Single-use Strategy to Enable Manufacturing of Affordable Biologics

The current processing paradigm of large manufacturing facilities dedicated to single product production is no longer an effective approach for best manufacturing practices. Increasing competition for new indications and the launch of biosimilars for the monoclonal antibody market have put pressure on manufacturers to produce at lower cost. Single-use technologies and continuous upstream processes have proven to be cost-efficient options to increase biomass production but as of today the adoption has been only minimal for the purification operations, partly due to concerns related to cost and scale-up. This review summarizes how a single-use holistic process and facility strategy can overcome scale limitations and enable cost-efficient manufacturing to support the growing demand for affordable biologics. Technologies enabling high productivity, right-sized, small footprint, continuous, and automated upstream and downstream operations are evaluated in order to propose a concept for the flexible facility of the future.

A distributed spatial–temporal weighted model on MapReduce for short-term traffic flow forecasting

Accurate and timely traffic flow prediction is crucial to proactive traffic management and control in data-driven intelligent transportation systems (D2ITS), which has attracted great research interest in the last few years. In this paper, we propose a Spatial–Temporal Weighted K-Nearest Neighbor model, named STW-KNN, in a general MapReduce framework of distributed modeling on a Hadoop platform, to enhance the accuracy and efficiency of short-term traffic flow forecasting. More specifically, STW-KNN considers the spatial–temporal correlation and weight of traffic flow with trend adjustment features, to optimize the search mechanisms containing state vector, proximity measure, prediction function, and K selection. Furthermore, STW-KNN is implemented on a widely adopted Hadoop distributed computing platform with the MapReduce parallel processing paradigm, for parallel prediction of traffic flow in real time. Finally, with extensive experiments on real-world big taxi trajectory data, STW-KNN is compared with the state-of-the-art prediction models including conventional K-Nearest Neighbor (KNN), Artificial Neural Networks (ANNs), Naïve Bayes (NB), Random Forest (RF), and C4.5. The results demonstrate that the proposed model is superior to existing models on accuracy by decreasing the mean absolute percentage error (MAPE) value more than 11.59% only in time domain and even achieves 89.71% accuracy improvement with the MAPEs of between 3.34% and 6.00% in both space and time domains, and also significantly improves the efficiency and scalability of short-term traffic flow forecasting over existing approaches.

English

Modern digital computers outperform humans in tasks based on precise and fast arithmetic operations. However, people are much better and faster than computers in solving complex perceptual problems, such as recognizing images, often in the presence of disturbances. Also, humans can perform complex movements with precision and grace, even in the presence of disturbances, and can generalize from past experience. In a computer, usually there exists a single processor implementing a sequence of arithmetic and logical operations, nowadays at speeds approaching billion operations per second. However this type of devices has ability neither to adapt their structure nor to learn in the way that human being does. An Artificial Neural Model (ANM) is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel mathematical structure of the information processing system. In this paper we focus on the mathematical modeling aspects of the basic unit of the human nervous system and Artificial Neural Model.

A Review of Sentiment Analysis in Twitter Data Using Hadoop

Twitter is an online social networking site which contains rich amount of data that can be a structured, semistructured and un-structured data. In this work, a method which performs classification of tweet sentiment in Twitter is discussed. To improve its scalability and efficiency, it is proposed to implement the work on Hadoop Ecosystem, a widely-adopted distributed processing platform using the MapReduce parallel processing paradigm. Finally, extensive experiments will be conducted on real-world data sets, with an expectation to achieve comparable or greater accuracy than the proposed techniques in literature.

Structural-damage detection with big data using parallel computing based on MPSoC

Life of many people and vital social activities depend on good functioning of civil structures such as nuclear power plants, large bridges, pipelines, and others especially during and after high winds, earthquakes or environmental changes. We can recognize the structural damage of existing systems through signals of the stiffness reductions of their components, and changes in their observed displacements under certain load or environmental conditions. It can be formulated as an inverse problem and solved by the learning approach using neural networks, NN. Here, network function is referred as an associative memory device capable of satisfactory diagnostics even in the presence of noisy or incomplete measurements. It is not new idea, however, to obtain, in real time; an advanced warning of the onset of durability/structural problems at a stage when preventative action is possible has emerged creating an exciting new field within civil engineering. It refers to the broad concept of assessing the ongoing in-service performance of structures using a variety of measurement techniques, e.g., global position system—real time kinematic, GPS-RTK, wireless sensor networks, WSN, and others. It leads to, in this case, the so called big data. Big data is a term applied to data sets whose size is beyond the ability of commonly used traditional computer software and hardware to undertake their acquisition, access, and analytics in a reasonable amount of time. To solve this problem, we can use the Hadoop like Batch processing system using a distributed parallel processing paradigm called MapReduce, which like system handles the volume and variety part of big data. “MapReduce” is a programming model of Google and an associated implementation for processing and generating large data sets using a group of machines. Despite continuous data (BigData) originated from GPS-RTK and WSN is not studied in this paper; only one selected data is used to determine technical state of structure in real time. However, bigdata approach is used in this paper to solve the following problems: Firstly, small number of available data for neural network training may lead to strange, not practical solutions, which are beyond the permissible area. To overcome this problem, small available data is supplemented by numerical data generated from integration of Monte Carlo simulation and finite element method—MCSFEM model. Also, the learning approach using neural networks, NN, gives effect which is guaranteed only if this network has, among the various required conditions, a complete set of training data and so good (optimal) architecture of NN. Integration of a Monte-Carlo simulation with NN—MCSNN model in order to find the optimal (global) architecture of NN is selected and formulated to solve this problem. Both MCSFEM model and MCSNN model require expensive computing, which lead to necessity of the use of parallel computing based on MultiProcessorSystem on Chip, MPSoC, architecture. Secondly, the use of MPSoC architecture led to the emergence of multiple NN named net of neural network, NoNN, rather than a single NN. The simple approach “trial and error” for a single neural network is insufficient to determine an optimal architecture of NN from NoNN. To deal with this problem we develop, in this paper, a new distributed parallel processing using the “Master-Slaver” structure, in the “MapReduce” framework rather than the Google’s Hadoop like Batch processing system using a distributed parallel processing paradigm called MapReduce. Our method is based on the “single instruction multiple data”, SIMD, technology using computer with multiprocessor systems on a chip (or computer with multiple cores, CPUs) without necessarily Google’s computer program and a group of machines. It is used for both MCSFEM model and MCSNN model in order to generate virtual dataset and to find out the optimal architecture of NN from NoNN. It enables us to create quickly many numerical data and eighty thousand architectures (using computer with eight CPUs) can be created instead of eight thousand (using computer with one CPU) only in the same computation times. Effect obtained is presented in the numerical example for the local damage detection in real time of the plane steel truss structures so that one can react quickly when problems appear or to predict new trends in the near future.

A survey of open source tools for machine learning with big data in the Hadoop ecosystem

With an ever-increasing amount of options, the task of selecting machine learning tools for big data can be difficult. The available tools have advantages and drawbacks, and many have overlapping uses. The world’s data is growing rapidly, and traditional tools for machine learning are becoming insufficient as we move towards distributed and real-time processing. This paper is intended to aid the researcher or professional who understands machine learning but is inexperienced with big data. In order to evaluate tools, one should have a thorough understanding of what to look for. To that end, this paper provides a list of criteria for making selections along with an analysis of the advantages and drawbacks of each. We do this by starting from the beginning, and looking at what exactly the term “big data” means. From there, we go on to the Hadoop ecosystem for a look at many of the projects that are part of a typical machine learning architecture and an understanding of how everything might fit together. We discuss the advantages and disadvantages of three different processing paradigms along with a comparison of engines that implement them, including MapReduce, Spark, Flink, Storm, and H2O. We then look at machine learning libraries and frameworks including Mahout, MLlib, SAMOA, and evaluate them based on criteria such as scalability, ease of use, and extensibility. There is no single toolkit that truly embodies a one-size-fits-all solution, so this paper aims to help make decisions smoother by providing as much information as possible and quantifying what the tradeoffs will be. Additionally, throughout this paper, we review recent research in the field using these tools and talk about possible future directions for toolkit-based learning.

Call for Papers—Interfaces Special Issue: Applications of Analytics and Operations Research in Big Data Analysis

The numbers and modalities of digital information sources being captured to monitor our traffic, weather, goods, manufacturing facilities, utilities, ports, health, and information technology infrastructure are growing at an incredible rate. Commercial and government organizations require the continuous analysis of this structured and unstructured data—transactions, sensor readings, text, audio, video, and images—to monitor, alert, learn from, and sometimes affect and control our surroundings. Doing this at scale has necessitated that we move away from traditional database store-and-process solutions to the emergence of the big data processing paradigm—with algorithms, systems, and applications designed to process data continuously and in a scalable, distributed manner. This includes the stream-processing paradigm for continuous analysis, and map-reduce (and similar) frameworks for scalable data processing. Big data processing lies at the intersection of several diverse disciplines, including large-scale distributed systems, programming models, signal processing, data mining, machine learning, real-time algorithms, and user interfaces. Big data processing has also heavily leveraged analytics and operations research both at the business level for analyzing the data to enable better decisions, and at the systems level for making decisions about scheduling and optimization for scaling. In this special issue, we solicit papers that describe applications of big data processing systems and algorithms at the strategic, tactical, and operational levels of business processes. Papers must belong to one or more of the following categories: • An application of big data systems and analytics with a verification letter from the company or organization benefitting from the work; • A novel literature review of analytics and operations research methods related to big data processing and distributed and continuous analysis; • A tutorial emphasizing the emergence of novel big data processing systems, algorithms, and mathematical models for operations research. Authors should review the Interfaces instructions on preparing a paper at http://pubsonline.informs.org/ page/inte/submission-guidelines. Papers must be submitted online at the ScholarOne Manuscripts website: http://mc.manuscriptcentral.com/inte. In the Author Center of ScholarOne, please designate the “Manuscript Type” (#1—“Type, Title, and Abstract”) as “Special Issue” in the dropdown list; in #4—“Details and Comments,” respond “Operations Research in Big Data Analysis” to the question, “This paper is for which special issue?” All papers will be refereed. Authors planning to write review articles and tutorials for this special issue should indicate their intent to the special issue editor by November 15, 2015. Articles must be received no later than February 15, 2016 for full consideration for publication in this special issue. Papers received after that date will be considered for publication in Interfaces, but will not be guaranteed consideration in the special issue.

Definição e desenvolvimento de competências: um paradigma no processo estratégico

This study aims to examine the issue of defining and developing of competences and their definitive insertion in the strategic process of organizations. The relevance is evident, since in the globalized world of the twenty-first century new skills are being required of workers, employers and citizens in general. The understanding of the current dynamics of competence-based management becomes important, since it can be approached as a set of knowledge, skills and attitudes that accredit an individual to perform a certain function or, on the other hand, as achievements of the citizen in his/her work, i.e., the first definition focuses on the acquired potential and the second one lies in the effected performance. The paper also identifies the methods and techniques for assessing competences, in order to obtain a more objective view of human potential that exists in certain company and what are the needs for updating. Moreover, it attempts to present the importance of emotional competence in the strategic process and its close links with the concept of emotional intelligence, as well as it aims to demonstrate the importance of the irrevocable insertion of competences as a paradigm in the strategic process. Keywords: Competences. Management. Organizations. Globalization. Strategic Process.

On the evolutionary relationship between chondrocytes and osteoblasts.

Vertebrates are the only animals that produce bone, but the molecular genetic basis for this evolutionary novelty remains obscure. Here, we synthesize information from traditional evolutionary and modern molecular genetic studies in order to generate a working hypothesis on the evolution of the gene regulatory network (GRN) underlying bone formation. Since transcription factors are often core components of GRNs (i.e., kernels), we focus our analyses on Sox9 and Runx2. Our argument centers on three skeletal tissues that comprise the majority of the vertebrate skeleton: immature cartilage, mature cartilage, and bone. Immature cartilage is produced during early stages of cartilage differentiation and can persist into adulthood, whereas mature cartilage undergoes additional stages of differentiation, including hypertrophy and mineralization. Functionally, histologically, and embryologically, these three skeletal tissues are very similar, yet unique, suggesting that one might have evolved from another. Traditional studies of the fossil record, comparative anatomy and embryology demonstrate clearly that immature cartilage evolved before mature cartilage or bone. Modern molecular approaches show that the GRNs regulating differentiation of these three skeletal cell fates are similar, yet unique, just like the functional and histological features of the tissues themselves. Intriguingly, the Sox9 GRN driving cartilage formation appears to be dominant to the Runx2 GRN of bone. Emphasizing an embryological and evolutionary transcriptomic view, we hypothesize that the Runx2 GRN underlying bone formation was co-opted from mature cartilage. We discuss how modern molecular genetic experiments, such as comparative transcriptomics, can test this hypothesis directly, meanwhile permitting levels of constraint and adaptation to be evaluated quantitatively. Therefore, comparative transcriptomics may revolutionize understanding of not only the clade-specific evolution of skeletal cells, but also the generation of evolutionary novelties, providing a modern paradigm for the evolutionary process.