Code instrumentation is vital for analysing software behaviour and facilitating cloud computing observability and monitoring, especially in microservices and containers. Despite its benefits, instrumentation introduces complexity and performance overhead, which may inadvertently slow down systems and cause unexpected or erratic behaviour. In this study, we examine the effect of automated code instrumentation on the performance of containerised microservices by comparing instrumented systems against a baseline without instrumentation. Our experimental framework is based on key performance metrics, including response time, latency, throughput, and error percentage. It is executed using a rigorous methodology with a warm-up strategy to mitigate cold-start effects. Over 5000 experiments were conducted on 70 microservice APIs drawn from two open-source applications hosted on AWS and Azure to compare the results with baseline data. The experimental analysis comprises three stages: a pilot study on AWS, a case study on AWS and Azure, and an outlier analysis of the experimental results. Overall throughput decreased by up to 8.40 %, with some individual cases experiencing up to a 30 % reduction compared to the baseline, and response time and latency dropped by 20–49 %. Moreover, the results show more outlier cases in instrumentation results than in the baseline. Additionally, the results reveal more outlier cases in the instrumentation results compared to the baseline. The instrumentation has led to unexpected or erratic behaviour, as indicated by higher variations in response time, latency, and throughput values, along with increased error rates and occasional outlier values that were not observed in the non-instrumented run. This indicates that the performance differences we observed are attributable to overhead introduced by instrumentation, rather than inherent inefficiencies within the APIs themselves. Furthermore, statistical analysis utilised the Wilcoxon Signed-Rank test and mean ratios, with multiple approaches validating significant performance differences between instrumented and baseline conditions for both cloud services. A significance analysis using Cohen’s d indicates that the throughput and response time reductions in both platforms are not only statistically significant but also suggest considerable operational impact. These findings offer insights into automated code instrumentation's performance and impact on containerised microservices. It highlights the need to develop better and less impactful instrumentation techniques, and possibly towards the development of a new approach for large-scale software development and deployment in cloud environments that facilitates efficient instrumentation by design.

Support vector machines (SVM) are often applied in the context of machine learning analysis of various data. Given the nature of SVMs, these operate always in the sub-interpolation range as a machine learning method. Here we explore the impact of structural complexity on the performance and statistical reliability of SVMs applied for text mining. We set a theoretical framework for our analysis. We found experimentally that the statistical reliability and performance reduce exponentially with the increase of the structural complexity of the SVMs. This is an important result for the understanding of how the prediction error of SVM predictive data models behaves.

A growing sense of unfairness permeates our quasi-digital society. Despite drivers supporting and motivating ethical practice in the digital technology ecosystem, there are compounding barriers to fairness that, at every level, impact technology innovation, delivery and access. Amongst these are barriers and omissions at the earliest stages of technology intentionality and design; systemic inadequacies in sensing systems that deteriorate performance for individuals based on ethnicity, age and physicality; system design, co-requisite and interface decisions that limit access; biases and inequities in datasets and algorithms; and limiting factors in system function and security. Additionally, there are concerns about unethical and illegal practices amongst digital technology providers: for example, in planned obsolescence and anti-competitive behaviors, failings in data practices and security, and in responses to problematic use and behaviors. It is critical that these failings are identified and addressed to better evolve a fairer future digital technology ecosystem. This paper contributes a perspective on technological stewardship and innovation; it identifies the compounding nature of barriers to fairness in the current digital technology ecosystem, and contrasts these with the non-compounding fairness drivers that, in general, establish minimum requirements.

Federated learning (FL) is a well established method for performing machine learning tasks over massively distributed data. However in settings where data is distributed in a non-iid (not independent and identically distributed) fashion - as is typical in real world situations - the joint model produced by FL suffers in terms of test set accuracy and/or communication costs compared to training on iid data. We show that learning a single joint model is often not optimal in the presence of certain types of non-iid data. In this work we present a modification to FL by introducing a hierarchical clustering step (FL+HC) to separate clusters of clients by the similarity of their local updates to the global joint model. Once separated, the clusters are trained independently and in parallel on specialised models. We present a robust empirical analysis of the hyperparameters for FL+HC for several iid and non-iid settings. We show how FL+HC allows model training to converge in fewer communication rounds (significantly so under some non-iid settings) compared to FL without clustering. Additionally, FL+HC allows for a greater percentage of clients to reach a target accuracy compared to standard FL. Finally we make suggestions for good default hyperparameters to promote superior performing specialised models without modifying the the underlying federated learning communication protocol.

Neural networks are often used to approximate functions defined over high-dimensional data spaces (e.g. text data, genomic data, multi-sensor data). Such approximation tasks are usually difficult due to the curse of dimensionality and improved methods are needed to deal with them effectively and efficiently. Since the data generally resides on a lower dimensional manifold various methods have been proposed to project the data first into a lower dimension and then build the neural network approximation over this lower dimensional projection data space. Here we follow this approach and combine it with the idea of weak learning through the use of random projections of the data. We show that random projection of the data works well and the approximation errors are smaller than in the case of approximation of the functions in the original data space. We explore the random projections with the aim to optimize this approach.

Measuring and testing the performance of cloud-based software services is critically important in the context of rapid growth of cloud computing. Scalability, elasticity and efficiency are interrelated aspects of performance of cloud-based software services. Here we present a work that is focused on measuring the scalability of cloud-based software services in technical terms. We introduce technical scalability metrics inspired by earlier technical metrics of elasticity.

We describe the questions and discussions raised at the First Workshop on Social Learning and Cultural Evolution held at theArtificial Life Conference 2016 in Cancún, Mexico in July 2016. The purpose of the workshop was to assemble artificial life researchers interested in social learning and cultural evolution into one group so that we could focus on recent work and interesting open questions. Our discussion related to both the mechanisms of social learning and cultural evolution and the consequences and influence of social learning and cultural evolution on living systems. We present the contributions of our workshop presenters and conclude with a discussion of the more important open questions in this area.

Approximation of high-dimensional functions is a challenge for neural networks due to the curse of dimensionality. Often the data for which the approximated function is defined resides on a low-dimensional manifold and in principle the approximation of the function over this manifold should improve the approximation performance. It has been show that projecting the data manifold into a lower dimensional space, followed by the neural network approximation of the function over this space, provides a more precise approximation of the function than the approximation of the function with neural networks in the original data space. However, if the data volume is very large, the projection into the low-dimensional space has to be based on a limited sample of the data. Here, we investigate the nature of the approximation error of neural networks trained over the projection space. We show that such neural networks should have better approximation performance than neural networks trained on high-dimensional data even if the projection is based on a relatively sparse sample of the data manifold. We also find that it is preferable to use a uniformly distributed sparse sample of the data for the purpose of the generation of the low-dimensional projection. We illustrate these results considering the practical neural network approximation of a set of functions defined on high-dimensional data including real world data as well.

The increasing volume of malicious content in social networks requires automated methods to detect and eliminate such content. This paper describes a supervised machine learning classification model that has been built to detect the distribution of malicious content in online social networks (ONSs). Multisource features have been used to detect social network posts that contain malicious Uniform Resource Locators (URLs). These URLs could direct users to websites that contain malicious content, drive-by download attacks, phishing, spam, and scams. For the data collection stage, the Twitter streaming application programming interface (API) was used and VirusTotal was used for labelling the dataset. A random forest classification model was used with a combination of features derived from a range of sources. The random forest model without any tuning and feature selection produced a recall value of 0.89. After further investigation and applying parameter tuning and feature selection methods, however, we were able to improve the classifier performance to 0.92 in recall.

Background:: Statistical validity and model complexity are both important concepts to enhanced understanding and correctness assessment of computational models. However, information about these are often missing from publications applying machine learning.

Social learning plays a key role in the evolution of cooperation in humans and other animals.It has also been shown both theoretically and experimentally that environmental adversity is also a key determinant of the evolution of cooperation among individuals.Here we investigate the impact of social learning on the evolution of cooperation in the context of a range of levels of environmental adversity.We used an agent-based simulated world of asexual individuals that communicate and play a probabilistic version of the Prisoner's Dilemma game.We considered simulated worlds either with or without random spreading of the offspring and two variants of social learning, either copying to some extent all communication rules or copying fully some of the communication rules of the best performing neighbor individual.The results show that in the case of spreading of the offspring, social learning increases the level of cooperation and reverses the association between this and the level of environmental adversity, i.e. low adversity with social learning implies higher level of cooperation.Copying fully some communication rules also increases the steady-state level of communication complexity in the simulated agent communities.The results suggest that the level of cooperation in communities of individuals may get boosted alternately by highly adverse environments and by layers of social learning in low adversity environments.

Background: Since the introduction of the systematic review process to Software Engineering in 2004, researchers have investigated a number of ways to mitigate the amount of effort and time taken to filter through large volumes of literature.

Understanding the mechanisms of restoration of activity in biological neural systems following exposure to damage is key for design of future neuro-prosthetic devices and restorative treatments. The pyloric rhythm network within the crustacean stomatogastric ganglion is a biological neural system that shows spontaneous restoration of activity following the stopping of inputs from higher control ganglia. We model the restoration of the activity in this network using conductance-based models of neurons and the alteration of conductance parameters of the model. Our analysis shows that this approach works only if some of the conductance values remain constrained following the stopping of higher inputs. Our model also shows that in order to model the restoration of the activity in this network it is not necessary to rely on complicated alterations of the mechanisms of Calcium ionic currents in the model, which was proposed previously.

Management of Parkinson's Disease (PD) could be improved significantly if reliable, objective information about fluctuations in disease severity can be obtained in ecologically valid surroundings such as the private home. Although automatic assessment in PD has been studied extensively, so far no approach has been devised that is useful for clinical practice. Analysis approaches common for the field lack the capability of exploiting data from realistic environments, which represents a major barrier towards practical assessment systems. The very unreliable and infrequent labelling of ambiguous, low resolution movement data collected in such environments represents a very challenging analysis setting, where advances would have significant societal impact in our ageing population. In this work we propose an assessment system that abides practical usability constraints and applies deep learning to differentiate disease state in data collected in naturalistic settings. Based on a large data-set collected from 34 people with PD we illustrate that deep learning outperforms other approaches in generalisation performance, despite the unreliable labelling characteristic for this problem setting, and how such systems could improve current clinical practice.

Neural network approximation of high-dimensional nonlinear functions is difficult due to the sparsity of the data in the high-dimensional data space and the need for good coverage of the data space by the `receptive fields' of the neurons. However, high-dimensional data often resides around a much lower dimensional supporting manifold. Given that a low dimensional approximation of the target function is likely to be more precise than a high-dimensional approximation, if we can find a mapping of the data points onto a lower-dimensional space corresponding to the supporting manifold, we expect to be able to build neural network approximations of the target function with improved precision and generalization ability. Here we use the local linear embedding (LLE) method to find the low-dimensional manifold and show that the neural networks trained on the transformed data achieve much better function approximation performance than neural networks trained on the original data.

Function approximation is one of the core tasks that are solved using neural networks in the context of many engineering problems. However, good approximation results need good sampling of the data space, which usually requires exponentially increasing volume of data as the dimensionality of the data increases. At the same time, often the high-dimensional data is arranged around a much lower dimensional manifold. Here we propose the breaking of the function approximation task for high-dimensional data into two steps: (1) the mapping of the high-dimensional data onto a lower dimensional space corresponding to the manifold on which the data resides and (2) the approximation of the function using the mapped lower dimensional data. We use over-complete self-organizing maps (SOMs) for the mapping through unsupervised learning, and single hidden layer neural networks for the function approximation through supervised learning. We also extend the two-step procedure by considering support vector machines and Bayesian SOMs for the determination of the best parameters for the nonlinear neurons in the hidden layer of the neural networks used for the function approximation. We compare the approximation performance of the proposed neural networks using a set of functions and show that indeed the neural networks using combined unsupervised and supervised learning outperform in most cases the neural networks that learn the function approximation using the original high-dimensional data.

Hybrid bio-silicon networks are difficult to implement in practice due to variations of biological neuron bursting frequency. This causes the hybrid network to have inaccuracies and unreliability. The network may produce irregular bursts or incorrect spiking phase relationships if the electrical neuron bursting frequency is not suitable for biological neurons. To solve this potentially vital problem, a novel adaptive control system based on dynamic clamp is proposed. Biological measurement is combined with an adaptive controller to control to silicon neuron bursting periods in real time. We use a hybrid pyloric network which contains three real neurons and one electronic neuron as a case study. Simulation results indicate that the silicon neuron can follow the biological neuron bursting frequency in real time to achieve hybrid network functionalities. System settling time can be achieved in 303 milliseconds and percentage overshoot kept to 1%. We believe that our methodology is scalable to various larger bio-silicon hybrid neural networks.

First Person Shooter (FPS) is a popular genre in online gaming, unfortunately not everyone plays the game fairly, and this hinders the growth of the industry. The aiming robot (aimbot) is a common cheating mechanism employed in this genre, it differs from many other common online bots in that there is a human operating alongside the bot, and thus the in-game data exhibit both human and bot-like behaviour. The aimbot users can aim much better than the average player. However, there are also a large number of highly skilled players who can aim much better than the average player, some of these players have in the past been banned from servers due to false accusations from their peers. Therefore, it would be interesting to find out if and where the honest player's and the bot user's behaviour differ. In this paper we investigate the difference between the aiming abilities of aimbot users and honest human players. We introduce two novel features and have conducted an experiment using a modified open source FPS game. Our data shows that there is significant difference between behaviours of honest players and aimbot users. We propose a voting scheme to improve aimbot detection in FPS based on distribution matching, and have achieved approximately 93% in both True positive and True negative rates with one of our features.

The emergence and evolution of cooperation among selfish individuals is a key question of theoretical biology. Uncertainty of outcomes of interactions between individuals is an important determinant of cooperative behavior. Here we describe a model that allows the analysis of the effects of such uncertainty on the level of cooperation. We show that in iterated cooperation games the level of cooperation increases with the level of outcome uncertainty. We show that this is the case if the individuals communicate about their cooperation intentions and also if they do not communicate their intentions.

In a distributed virtual environment interest management schemes attempt to ensure virtual world objects exchange messages only if they are interacting. This provides an opportunity to allow such systems to scale to support many hundreds, even thousands, of virtual world objects by reducing the need to send and process unnecessary messages. However, arriving at the optimum configuration for the interest management scheme is a challenging and complex problem. The customary approach is to configure parameters in an ad-hoc manner (supported by experience). However, such an approach is unlikely to yield optimum system performance. Here we propose a means of finding optimal parameter values by way of a simulation tool equipped with evolutionary optimization capabilities. We provide a series of experiments that indicate that our simulator can aid in configuring interest management schemes to gain better system performance.

One hypothesis about how biological neural systems work suggests that they use attractor dynamics to define their behaviour. Such behaviour can be modelled using recurrent neural network models. It has been shown that such systems can perform a wide range of computational tasks by learning abstract grammars. Here we show that chaotic neural dynamics in recurrent neural systems is advantageous in the sense that it facilitates the encoding of grammars describing complex behaviour. This result may explain why it is common the observation of chaotic dynamics in biological neural systems.

The majority of faults and consequent errors and failures in computer systems stem from the complexity of the system itself according to B. Schneier (2000). Yet complexity as a non-functional property is largely disregarded from initial development phases or at best is considered "manageable" by the process life cycle. Although at large the property is considered as having being conquered, valid sources inform us as stated in W. B. Kernighan (1987) that complexity as a property - borne by the system or emergent - is the cause of most failures. Furthermore the second most frequent case of failures is that of interaction with the system. In this paper we are looking at one aspect of process interactions. We are presenting a method for modelling and analyzing information system process interactions for the purpose of enhancing security. The paper introduces our work in the area of complex systems and gives a general introduction to our modelling methodology and its usage.

Understanding how biological neural systems work needs appropriate models that focus on relevant features of neural activities and ignore irrelevant ones. Here, we propose a new way of describing biological neural systems in terms of abstract communication systems. Such systems use pattern languages formulated as probabilistic continuation rules that determine which communications follow other communications. We apply this framework to define neural communication systems. The formal analysis is followed by a brief interpretation of how the crab stomatogastric ganglion works in the context of the proposed framework. We also discuss further implications of the description of neural systems as abstract communication systems.

There is some experimental evidence that the neural information processing in some small neural systems with complex behaviour (e.g., olfactory bulb, crab stomatogastric ganglion) can be best described in terms of activity patterns. Recently the pattern languages were proposed as an interpretation and analysis tool for such complex systems. Here we describe a model that shows how spatio-temporal neural activity patterns can perform abstract computation. We show how such activity patterns can be interpreted as a language of the neural system, and how neural information processing may be done in terms of pattern languages.

We investigate the combination of the Kohonen networks with the kernel methods in the context of classification. We use the idea of kernel functions to handle products of vectors of arbitrary dimension. We indicate how to build Kohonen networks with robust classification performance by transformation of the original data vectors into a possibly infinite dimensional space. The resulting Kohonen networks preserve a non-Euclidean neighborhood structure of the input space that fits the properties of the data. We show how to optimize the transformation of the data vectors in order to obtain higher classification performance. We compare the kernel-Kohonen networks with the regular Kohonen networks in the context of a classification task.

The approximation of highly irregular decision regions is a challenging problem in pattern recognition and classification. Existing neural networks require many neurons for approximating irregular decision regions. A new tree-structured neural network algorithm is proposed that does not suffer from this limitation. The network approximates irregular regions parsimoniously by using receptive fields having a special overlapping structure The performance of the proposed network is evaluated on an approximation task involving a highly irregular decision region defined by the Mandelbrot set. The results show that the tree-structured neural network approximates decision regions much more parsimoniously than Kohonen and reduced Coulomb-potential networks.