-
First numerical analysis of runaway electron generation in tungsten-rich plasmas towards ITER
Authors:
J. Walkowiak,
M. Hoppe,
I. Ekmark,
A. Jardin,
J. Bielecki,
K. Król,
Y. Savoye-Peysson,
D. Mazon,
D. Dworak,
M. Scholz
Abstract:
The disruption and runaway electron analysis model code was extended to include tungsten impurities in disruption simulations with the aim of studying the runaway electron (RE) generation. This study investigates RE current sensitivity on the following plasma parameters and modelling choices: tungsten concentration, magnetic perturbation strength, electron modelling, thermal quench time and tokama…
▽ More
The disruption and runaway electron analysis model code was extended to include tungsten impurities in disruption simulations with the aim of studying the runaway electron (RE) generation. This study investigates RE current sensitivity on the following plasma parameters and modelling choices: tungsten concentration, magnetic perturbation strength, electron modelling, thermal quench time and tokamak geometry: ITER-like or ASDEX-like. Our investigation shows that a tungsten concentration below 10-3 does not cause significant RE generation on its own. However, at higher concentrations it is possible to reach a very high RE current. Out of the two tested models of electrons in plasma: fluid and isotropic (kinetic), results from the fluid model are more conservative, which is useful when it comes to safety analysis. However, these results are overly pessimistic when compared to the isotropic model, which is based on a more reliable approach. Our results also show that the hot-tail RE generation mechanism is dominant as a primary source of RE in tungsten induced disruptions, usually providing orders of magnitude higher RE seed than Dreicer generation. We discuss best practices for simulations with tungsten-rich plasma, present the dependence of the safety limits on modelling choices and highlight the biggest shortcoming of the current simulation techniques. The obtained results pave the way for a wider analysis of tungsten impact on the disruption dynamics, including the mitigation techniques for ITER in the case of strong contamination of the plasma with tungsten.
△ Less
Submitted 23 April, 2024;
originally announced April 2024.
-
Scaling Laws for Fine-Grained Mixture of Experts
Authors:
Jakub Krajewski,
Jan Ludziejewski,
Kamil Adamczewski,
Maciej Pióro,
Michał Krutul,
Szymon Antoniak,
Kamil Ciebiera,
Krystian Król,
Tomasz Odrzygóźdź,
Piotr Sankowski,
Marek Cygan,
Sebastian Jaszczur
Abstract:
Mixture of Experts (MoE) models have emerged as a primary solution for reducing the computational cost of Large Language Models. In this work, we analyze their scaling properties, incorporating an expanded range of variables. Specifically, we introduce a new hyperparameter, granularity, whose adjustment enables precise control over the size of the experts. Building on this, we establish scaling la…
▽ More
Mixture of Experts (MoE) models have emerged as a primary solution for reducing the computational cost of Large Language Models. In this work, we analyze their scaling properties, incorporating an expanded range of variables. Specifically, we introduce a new hyperparameter, granularity, whose adjustment enables precise control over the size of the experts. Building on this, we establish scaling laws for fine-grained MoE, taking into account the number of training tokens, model size, and granularity. Leveraging these laws, we derive the optimal training configuration for a given computational budget. Our findings not only show that MoE models consistently outperform dense Transformers but also highlight that the efficiency gap between dense and MoE models widens as we scale up the model size and training budget. Furthermore, we demonstrate that the common practice of setting the size of experts in MoE to mirror the feed-forward layer is not optimal at almost any computational budget.
△ Less
Submitted 12 February, 2024;
originally announced February 2024.
-
MoE-Mamba: Efficient Selective State Space Models with Mixture of Experts
Authors:
Maciej Pióro,
Kamil Ciebiera,
Krystian Król,
Jan Ludziejewski,
Michał Krutul,
Jakub Krajewski,
Szymon Antoniak,
Piotr Miłoś,
Marek Cygan,
Sebastian Jaszczur
Abstract:
State Space Models (SSMs) have become serious contenders in the field of sequential modeling, challenging the dominance of Transformers. At the same time, Mixture of Experts (MoE) has significantly improved Transformer-based Large Language Models, including recent state-of-the-art open models. We propose that to unlock the potential of SSMs for scaling, they should be combined with MoE. We showcas…
▽ More
State Space Models (SSMs) have become serious contenders in the field of sequential modeling, challenging the dominance of Transformers. At the same time, Mixture of Experts (MoE) has significantly improved Transformer-based Large Language Models, including recent state-of-the-art open models. We propose that to unlock the potential of SSMs for scaling, they should be combined with MoE. We showcase this on Mamba, a recent SSM-based model that achieves remarkable performance. Our model, MoE-Mamba, outperforms both Mamba and baseline Transformer-MoE. In particular, MoE-Mamba reaches the same performance as Mamba in $2.35\times$ fewer training steps while preserving the inference performance gains of Mamba against Transformer.
△ Less
Submitted 26 February, 2024; v1 submitted 8 January, 2024;
originally announced January 2024.
-
Credibility of Automatic Appraisal of Domain Names
Authors:
Karol Król,
Artur Strzelecki,
Dariusz Zdonek
Abstract:
Both domain names and entire websites are increasingly frequently treated as assets, the value of which can be appraised. The objective of the present thesis was to verify the credibility of domain name appraisals obtained using generally available web applications in an automated, algorithmic way. In conclusions section, it was mentioned that the terms domain name appraisal and website appraisal…
▽ More
Both domain names and entire websites are increasingly frequently treated as assets, the value of which can be appraised. The objective of the present thesis was to verify the credibility of domain name appraisals obtained using generally available web applications in an automated, algorithmic way. In conclusions section, it was mentioned that the terms domain name appraisal and website appraisal are frequently equated. It was also shown that algorithms used in the tested applications consider parameters characterising websites. Thus, they cannot be used to verify the value of domain names themselves. Moreover, during the analysis of the pattern of operation of the appraisal websites it was noticed that they were not made available with domain name or website appraisals in mind. Their objective was to acquire and intercept online traffic. Such applications also left cookie files on recipients' devices, which were then used by advertising systems based on the re-marketing concept.
△ Less
Submitted 29 October, 2018;
originally announced November 2018.
-
"`They brought in the horrible key ring thing!" Analysing the Usability of Two-Factor Authentication in UK Online Banking
Authors:
Kat Krol,
Eleni Philippou,
Emiliano De Cristofaro,
M. Angela Sasse
Abstract:
To prevent password breaches and guessing attacks, banks increasingly turn to two-factor authentication (2FA), requiring users to present at least one more factor, such as a one-time password generated by a hardware token or received via SMS, besides a password. We can expect some solutions -- especially those adding a token -- to create extra work for users, but little research has investigated u…
▽ More
To prevent password breaches and guessing attacks, banks increasingly turn to two-factor authentication (2FA), requiring users to present at least one more factor, such as a one-time password generated by a hardware token or received via SMS, besides a password. We can expect some solutions -- especially those adding a token -- to create extra work for users, but little research has investigated usability, user acceptance, and perceived security of deployed 2FA.
This paper presents an in-depth study of 2FA usability with 21 UK online banking customers, 16 of whom had accounts with more than one bank. We collected a rich set of qualitative and quantitative data through two rounds of semi-structured interviews, and an authentication diary over an average of 11 days. Our participants reported a wide range of usability issues, especially with the use of hardware tokens, showing that the mental and physical workload involved shapes how they use online banking. Key targets for improvements are (i) the reduction in the number of authentication steps, and (ii) removing features that do not add any security but negatively affect the user experience.
△ Less
Submitted 19 January, 2015;
originally announced January 2015.
-
A Black--Scholes Model with Long Memory
Authors:
John A. D. Appleby,
John A. Daniels,
Katja Krol
Abstract:
This note develops a stochastic model of asset volatility. The volatility obeys a continuous-time autoregressive equation. Conditions under which the process is asymptotically stationary and possesses long memory are characterised. Connections with the class of ARCH($\infty$) processes are sketched.
This note develops a stochastic model of asset volatility. The volatility obeys a continuous-time autoregressive equation. Conditions under which the process is asymptotically stationary and possesses long memory are characterised. Connections with the class of ARCH($\infty$) processes are sketched.
△ Less
Submitted 24 February, 2012;
originally announced February 2012.
-
Two-dimensional point spread matrix of layered metal-dielectric imaging elements
Authors:
Rafal Kotynski,
Tomasz Antosiewicz,
Karol Krol,
Krassimir Panajotov
Abstract:
We describe the change of the spatial distribution of the state of polarisation occurring during two-dimensional imaging through a multilayer and in particular through a layered metallic flat lens. Linear or circular polarisation of incident light is not preserved due to the difference in the amplitude transfer functions for the TM and TE polarisations. In effect, the transfer function and the poi…
▽ More
We describe the change of the spatial distribution of the state of polarisation occurring during two-dimensional imaging through a multilayer and in particular through a layered metallic flat lens. Linear or circular polarisation of incident light is not preserved due to the difference in the amplitude transfer functions for the TM and TE polarisations. In effect, the transfer function and the point spread function that characterize 2D imaging through a multilayer both have a matrix form and cross-polarisation coupling is observed for spatially modulated beams with a linear or circular incident polarisation. The point spread function in a matrix form is used to characterise the resolution of the superlens for different polarisation states. We demonstrate how the 2D PSF may be used to design a simple diffractive nanoelement consisting of two radial slits. The structure assures the separation of non-diffracting radial beams originating from two slits in the mask and exhibits an interesting property of a backward power flow in between the two rings.
△ Less
Submitted 2 December, 2010;
originally announced December 2010.
-
Long Memory in a Linear Stochastic Volterra Differential Equation
Authors:
John A. D. Appleby,
Katja Krol
Abstract:
In this paper we consider a linear stochastic Volterra equation which has a stationary solution. We show that when the kernel of the fundamental solution is regularly varying at infinity with a log-convex tail integral, then the autocovariance function of the stationary solution is also regularly varying at infinity and its exact pointwise rate of decay can be determined. Moreover, it can be shown…
▽ More
In this paper we consider a linear stochastic Volterra equation which has a stationary solution. We show that when the kernel of the fundamental solution is regularly varying at infinity with a log-convex tail integral, then the autocovariance function of the stationary solution is also regularly varying at infinity and its exact pointwise rate of decay can be determined. Moreover, it can be shown that this stationary process has either long memory in the sense that the autocovariance function is not integrable over the reals or is subexponential. Under certain conditions upon the kernel, even arbitrarily slow decay rates of the autocovariance function can be achieved. Analogous results are obtained for the corresponding discrete equation.
△ Less
Submitted 7 September, 2010;
originally announced September 2010.
