Search

We introduce a mathematical model of cultural evolution to study cultural traits that shape how individuals exchange information. Current theory focuses on traits that influence the reception of information (receiver traits), such as evaluating whether information represents the majority or stems from a trusted source. Our model shifts the focus from the receiver to the sender of cultural information and emphasizes the role of sender traits, such as communicability or persuasiveness. Here, we show that sender traits are probably a stronger driving force in cultural evolution than receiver traits. While receiver traits evolve to curb cultural transmission, sender traits can amplify it and fuel the self-organization of systems of mutually supporting cultural traits, including traits that cannot be maintained on their own. Such systems can reach arbitrary complexity, potentially explaining uniquely human practical and mental skills, goals, knowledge and creativity, independent of innate factors. Our model incorporates social and individual learning throughout the lifespan, thus connecting cultural evolutionary theory with developmental psychology. This approach provides fresh insights into the trait-individual duality, that is, how cultural transmission of single traits is influenced by individuals, who are each represented as an acquired system of cultural traits.

Background. Attempted intrusions at companies, either from an insider threat orotherwise, is increasing in frequency. Most commonly used is static analysis and filters to stop specific attacks. Utilising machine learning in order to detect behaviouralanomalies in the access flow of an isolated system can aid in detecting, and stopping, attacks faster than previous methods. Objectives. In this thesis, four algorithms were selected to be compared againsteach other using three different metrics. These metrics were chosen for their importance in an isolated domain. All algorithms will be trained on the same dataset, from which anomalies are created that are used to test each model. Methods. A dataset created for anomaly detection is preprocessed to fit the scenario that was explored. After which the dataset was split per user and only the user with the most samples was used for training the models. In order to test and evaluate the models, anomalies were forged from a profile created out of the metadata belonging to the chosen user. These anomalies, alongside a part of the benign samples were used to evaluate the F1 score of each model, which was compared. The better performing model according to the F1 score was then subjected to hyperparameter tuning to improve the performance further. Afterwards, the speed of which the model was loaded, and a single sample was predicted and the memory consumption of each action was measured. Results. The results showed that two algorithms were relatively close, all depending on the strictness of memory consumption. Local Outlier Factor, which used four times the memory (44 MB) of the other models, proved to be the better option when looking at F1 score, at 90.91% after having undergone hyperparameter tuning. However, Elliptic Envelope was a close second at 86.61% without undergoing hyperparameter tuning, while consuming less memory (11 MB) than the others. The speed of loading the models were 26.68 ms and 2.01 ms, with predicting one, Bakgrund. Försök till intrång i företag, antingen från insiderhot eller på annat håll ökar i frekvens. Vanligtvis används statisk analys, eller olika filter för att motverka dessa attacker. Genom att använda maskininlärning för att upptäcka beteendeavikelser i ett loggflöde inuti ett isolerat system kan hjälpa till att upptäcka, och stoppa, attacker snabbare än tidigare metoder. Syfte. I det här arbetet har fyra algoritmer valts att jämföras med varandra genom att titta på tre olika mätvärden. Dessa mätvärden har valts på grund av dess betydelse i system placerade i en isolerad domän. Alla algoritmer tränades på samma dataset, och testas på avvikelser som har skapats från att tillverka en profil utifrån datasetet. Metod. Ett dataset som skapades för att upptäcka avvikelser i en åtkomstlogg har behandlats så att den ska passa scenariot som ska utforskas. Sedan så delades datasetet upp per användare, och enbart den användare med flest loggar har använts för att träna modellerna.För att testa modellerna, så har en profil byggts upp ifrån metadatan för att sedan generera anomala tillfällen för den valda användaren. Dessa avvikelser, tillsammans med en del utav de normala fallen har använts för att beräkna modellernas F1 värde. Sedan har tiden som krävts för att ladda modellen till minne från disk, tiden det tog för en gissning utav modellen, samt det datorminne som krävs för detta sparats. Dessa tre mätvärden har sedan satts emot varandra i jämförelsen. Den modell som presterade bäst i F1 värde genomgick hyperparameterjustering för att förbättra detta värde. Resultat. Resultatet visade att två algoritmer är någorlunda nära i hur de presterade. Skillnaden är att ena algoritmen, Local Outlier Factor, har ett lite högre F1 värde på 90.91% efter hyperparameterjustering, men kräver fyra gånger så mycket minne (44 MB). Dess tid att ladda ifrån disk var 26.68 ms, medans en gissning utav den tog 1.87 ms. Till skillnad från Elliptic Envelope som enbart krävde 11 MB för att ladda

Many cultural phenomena cannot be understood by studying traits in isolation. Instead, they are embedded in webs of relations, layered with rich social meaning, and sequentially acquired and evaluated, filtered, in light of previously acquired traits. Incorporating multiple traits and their relations is needed to understand the organization, dynamics, and emergent products of cultural evolution. This chapter argues that a systems approach offers the potential for: (i) a more thorough understanding of the source and character of emergent phenomena, crucial for understanding the origin and historical development of culture; (ii) an increased scope of cultural evolution, whereby faithful transmission, self-organization and filtering of culture can be accounted for by systems dynamics, without the need for strong assumptions about innate machinery; and (iii) a more robust theoretical connection between cultural evolution and processes of development as they are studied in developmental psychology and related fields.

Small shallow cumulus clouds (<1 km) over the tropical oceans appear to possess the ability to self-organize into mesoscale (10–100 km) patterns. To better understand the processes leading to such self-organized convection, we present Cloud Botany, an ensemble of 103 large-eddy simulations on domains of 150 km, produced by the Dutch Atmospheric Large Eddy Simulation model on supercomputer Fugaku. Each simulation is run in an idealized, fixed, larger-scale environment, controlled by six free parameters. We vary these over characteristic ranges for the winter trades, including parameter combinations observed during the EUREC4A (Elucidating the role of clouds–circulation coupling in climate) field campaign. In contrast to simulation setups striving for maximum realism, Cloud Botany provides a platform for studying idealized, and therefore more clearly interpretable causal relationships between conditions in the larger-scale environment and patterns in mesoscale, self-organized shallow convection. We find that any simulation that supports cumulus clouds eventually develops mesoscale patterns in their cloud fields. We also find a rich variety in these patterns as our control parameters change, including cold pools lined by cloudy arcs, bands of cross-wind clouds and aggregated patches, sometimes topped by thin anvils. Many of these features are similar to cloud patterns found in nature. The published data set consists of raw simulation output on full 3D grids and 2D cross-sections, as well as post-processed quantities aggregated over the vertical (2D), horizontal (1D) and all spatial dimensions (time-series). The data set is directly accessible from Python through the use of the EUREC4A intake catalog.

Animations and graphs of the Cloud Botany ensemble of large eddy simulations The data is stored as a single zip file. When unpacked, the visualizations can be navigated in a web browser - start by opening index.html in the top directory. The simulations were performed with DALES, the Dutch Atmospheric Large Eddy Simulation, on Supercomputer Fugaku. For access to the data itself, see How To EUREC4A. Additional material: ERA5 data for the experiment region for the spring 2020 and JOANNE dropsonde data from the EUREC4A campaign, used to define the ranges for the Cloud Botany ensemble parameter space.

Human Activity Recognition (HAR) is a growing field of research concerned with classifying human activities from sensor data. Modern smartphones contain numerous sensors that could be used to identify the physical activities of the smartphone wearer, which could have applications in sectors such as healthcare, eldercare, and fitness. This project aims to use smartphone sensor data together with machine learning to perform HAR on the following human locomotion activities: standing, walking, running, ascending stairs, descending stairs, and biking. The classification was done using a random forest classifier. Furthermore, in the special case of walking, an algorithm that can count the number of steps in a given data sequence was developed. The step counting algorithm was not based on a previous implementation and could therefore be considered novel. The step counter achieved a testing accuracy of 99.1\% and the HAR classifier a testing accuracy of 100\%. It is speculated that the abnormally high accuracies can be attributed primarily to the lack of data diversity, as in both cases only two persons collected the data., Mänsklig aktivitetsigenkänning är ett växande forskningsområde som handlar om att klassificera mänskliga aktiviteter från sensordata. Moderna mobiltelefoner innehåller många sensorer som kan användas för att identifiera de fysiska aktiviteterna som bäraren utför, vilket har tillämpningar inom sektorer som sjukvård, äldreomsorg och personlig hälsa. Detta projekt använder sensordata från mobiltelefoner tillsammans med maskininlärning för att utföra aktivitetsigenkänning på följande aktiviteter: stå, gå, springa, gå uppför trappor, gå nedför trappor och cykla. Klassificeringen gjordes med hjälp av en ``random forest''-klassificerare. Vidare utvecklades en algoritm som kan räkna antalet steg i en given datasekvens som samlats in när användaren går. Stegräkningsalgoritmen baserades inte på en tidigare implementering och kan därför betraktas som ny. Stegräknaren uppnådde en testnoggrannhet på 99,1\% och aktivitetsigenkänningen en testnoggrannhet på 100\%. De oväntat höga noggrannheterna antas främst bero på bristen av diversitet i datan, eftersom den endast samlades in av två personer i båda fallen., Kandidatexjobb i elektroteknik 2022, KTH, Stockholm

This thesis examines resource allocation for Federated Learning in wireless networks. In Federated learning a server and a number of users exchange neural network parameters during training. This thesis aims to create a realistic simulation of a Federated Learning process by creating a channel model and using compression when channel capacity is insufficient. In the thesis we learn that Federated learning can handle high ratios of sparsification compression. We will also investigate how the choice of users and scheduling schemes affect the convergence speed and accuracy of the training process. This thesis will conclude that the choice of scheduling schemes will depend on the distributed data distribution.

Objective: Political polarisation is a common problem that has gained increasing attention, and opinions are often divided along party lines. Moral foundations theorysuggests that there are some foundations, appealing to individual autonomy, that are part of everyone’s morality, while other foundations, concerning bounds to yourgroup, appeal only to conservatives. We here test the hypothesis that arguments that build on individualising foundations can persuade everyone, while binding argumentscan persuade only conservatives. This gives a prediction of what we can eventually agree on, and can explain why moral values tend to become more liberal/progressiveover time, despite polarisation. Methods: We used a classic experimental design (N=379) where respondents from the US were given moral arguments for nine moral stances. The respondents were firstasked for their opinion, and a week later, they were given individualising, binding or no arguments for the stances, and were asked for their opinion again. We measuredaverage opinion change across the issues depending on treatment and the political ideology of the respondent. Results: The results were in line with the hypothesis. There was a general significant effect across the political spectrum from being given individualising arguments, whilebinding arguments for the same moral stances had an effect only when given to conservative respondents. Those in the control condition who received no arguments didnot change their opinion significantly. Conclusion: In line with cognitive dissonance theory, the moral foundation support of respondents predicts the type of arguments to which they are susceptible. Thesefindings suggests that there are tools for finding common ground in polarising issues. Further, along with previous studies on which type of moral argument supports whichmoral position in the public debate, these findings provide a mechanistic explanation for public opinion change, and in particular for the empirical obs

Introduction Cultural traits are embedded in webs of relations and sequentially acquired and evaluated – filtered – in light of previously acquired traits. Incorporating multiple traits and their relations is needed to understand the organisation, dynamics, and emergent products of cultural evolution. Approach We introduce a modelling framework of individual filtering of interdependent traits and explore the effects on cultural transmission and emergent phenomena, through mathematical treatment and simulations. Findings Sequential acquisition of compatible traits leads to culturally evolved filters. These organise culture into mostly consistent and stable systems. Depending on the structure of trait relations, they modulate rates of cultural change, exhibit path dependence and punctuated equilibria, and can generate convergent transmission, clustering and polarisation. Discussion A systems view of culture can give a fuller understanding of the source and character of emergent phenomena. It increases the scope of cultural evolution, whereby faithful transmission, self-organisation and filters can be accounted for by systems dynamics without strong assumptions on innate machinery.

A new generation of operational atmospheric models operating at horizontal resolutions in the range 200 m ∼ 2 km is becoming increasingly popular for operational use in numerical weather prediction and climate applications. Such grid spacings are becoming sufficiently fine to resolve a fraction of the turbulent transports. Here we analyze Large-eddy simulation results of a convective boundary layer obtained by coarsening horizontal grid spacings up to 800 m. The aim is to explore the dependency of the mean state and turbulent fluxes on the grid resolution. Both isotropic and anisotropic eddy diffusion approaches are evaluated, where in the latter case the horizontal and vertical eddy diffusivities differ in accord with their horizontal and vertical grid spacings. For coarsening horizontal grid sizes entrainment at the top of the boundary layer tends to get slightly enhanced for isotropic diffusion, whereas for the anisotropic diffusion approach the vertically well-mixed boundary-layer structure becomes severely degraded. An analysis of the energy spectrum shows that anisotropic diffusion causes relatively more dissipation of variance at smaller length scales. This leads, in turn, to a shift of spectral energy toward larger length scales that also becomes apparent from a rather different kind of spatial organization of convection. The present study therefore suggests that details with regards to the representation of processes at small scales might impact the organization at length scales much larger than the smallest scales that can be resolved by the model.

A specific goal of the field of cultural evolution is to understand how processes of transmission and selection at the individual level lead to population-wide patterns of cultural diversity and change. Models of cultural evolution have typically assumed that traits are independent of one another and essentially exchangeable. But culture has a structure: traits bear relationships to one another that affect the transmission and selection process itself. Here, we introduce a modelling framework to explore the effect of interdependencies on the process of learning. Through simulations, we find that introducing a simple structure changes the cultural dynamics. Based on a basic filtering mechanism for parsing trait relationships, more elaborate cultural filters emerge. In a mostly incompatible cultural domain of traits, these filters organize culture into mostly (but not fully) consistent and stable systems. Incompatible domains produce small homogeneous cultures, while more compatibility increases size, diversity and group divergence. When individuals copy based on a trait's features (here, its compatibility relationships), they produce more homogeneous cultures than when they copy based on the agent carrying the cultural trait. We discuss the implications of considering cultural systems and filters in the dynamics of cultural change. This article is part of the theme issue 'Foundations of cultural evolution'.

Cultural evolution theory has long been inspired by evolutionary biology. Conceptual analogies between biological and cultural evolution have led to the adoption of a range of formal theoretical approaches from population dynamics and genetics. However, this has resulted in a research programme with a strong focus on cultural transmission. Here, we contrast biological with cultural evolution, and highlight aspects of cultural evolution that have not received sufficient attention previously. We outline possible implications for evolutionary dynamics and argue that not taking them into account will limit our understanding of cultural systems. We propose 12 key questions for future research, among which are calls to improve our understanding of the combinatorial properties of cultural innovation, and the role of development and life history in cultural dynamics. Finally, we discuss how this vibrant research field can make progress by embracing its multidisciplinary nature. This article is part of the theme issue 'Foundations of cultural evolution'.