73 research outputs found
Interpretable pairwise distillations for generative protein sequence models
No full textMany different types of generative models for protein sequences have been proposed in literature. Their uses include the prediction of mutational effects, protein design and the prediction of structural properties. Neural network (NN) architectures have shown great performances, commonly attributed to the capacity to extract non-trivial higher-order interactions from the data. In this work, we analyze two different NN models and assess how close they are to simple pairwise distributions, which have been used in the past for similar problems. We present an approach for extracting pairwise models from more complex ones using an energy-based modeling framework. We show that for the tested models the extracted pairwise models can replicate the energies of the original models and are also close in performance in tasks like mutational effect prediction. In addition, we show that even simpler, factorized models often come close in performance to the original models
The loyalty of the literary reviser : author, source text, target text or reader?
No full textCITATION: Feinauer, I. & Lourens, A. 2017. The loyalty of the literary reviser : author, source text, target text or reader?. Stellenbosch Papers in Linguistics Plus, 53:97-118,
doi:10.5842/53-0-735.The original publication is available at http://spilplus.journals.ac.za/The processes of revision and translation, according to Mossop (2010:112-113), can address the problem of conflicting interests, goals and needs by taking different approaches. Translation, he suggests, should seek to achieve a balance between loyalty to the source text author and to the target text readers, whereas revision should serve the interests of the future readers of the text. As a result, revising activities will steer away from a linguistic or a text-based approach in order to prioritise the needs of the reader. The question, however, is whether revisers in literary translation processes do follow the suggested approach and prioritise the needs of the target readers. An empirical analysis of the metatextual discourse among the agents involved in three different literary translation processes seeks to answer this question. During the course of this analysis, a second question presents itself, namely whether self-revision and other-revision (Mossop 2010:167, 174) should be discerned as two distinct types of revision or whether this distinction could be refined. The results of the subsequent analysis give rise to the presumption that such a binary projection of the act of revision can be contested. It is instead suggested that revision can be plotted on a continuum, with self-revision by the translator as the one end and revision by others as the other end. The analysis of three Afrikaans novels translated into English by the same award-winning translator suggests that self-revision by the translator may find itself moving away from true self-revision (a process that is not influenced by feedback from agents other than the translator) initially to a second phase of self-revision that is shaped by revision by others (e.g. the author, reviser and editor).http://spilplus.journals.ac.za/pub/article/view/735Publisher's versio
Entropic gradient descent algorithms and wide flat minima
No full textThe properties of flat minima in the empirical risk landscape of neural networks have been debated for some time. Increasing evidence suggests they possess better generalization capabilities with respect to sharp ones. In this work we first discuss the relationship between alternative measures of flatness: The local entropy, which is useful for analysis and algorithm development, and the local energy, which is easier to compute and was shown empirically in extensive tests on state-of-the-art networks to be the best predictor of generalization capabilities. We show semi-analytically in simple controlled scenarios that these two measures correlate strongly with each other and with generalization. Then, we extend the analysis to the deep learning scenario by extensive numerical validations. We study two algorithms, Entropy-SGD and Replicated-SGD, that explicitly include the local entropy in the optimization objective. We devise a training schedule by which we consistently find flatter minima (using both flatness measures), and improve the generalization error for common architectures (e.g. ResNet, EfficientNet)
Entropic gradient descent algorithms and wide flat minima
No full textThe properties of flat minima in the empirical risk landscape of neural networks have been debated for some time. Increasing evidence suggests they possess better generalization capabilities with respect to sharp ones. In this work we first discuss the relationship between alternative measures of flatness: the local entropy, which is useful for analysis and algorithm development, and the local energy, which is easier to compute and was shown empirically in extensive tests on state-of-the-art networks to be the best predictor of generalization capabilities. We show semi-analytically in simple controlled scenarios that these two measures correlate strongly with each other and with generalization. Then, we extend the analysis to the deep learning scenario by extensive numerical validations. We study two algorithms, entropy-stochastic gradient descent and replicated-stochastic gradient descent, that explicitly include the local entropy in the optimization objective. We devise a training schedule by which we consistently find flatter minima (using both flatness measures), and improve the generalization error for common architectures (e.g. ResNet, EfficientNet)
Deep Networks on Toroids: Removing Symmetries Reveals the Structure of Flat Regions in the Landscape Geometry
No full textWe systematize the approach to the investigation of deep neural network landscapes by basing it on the geometry of the space of implemented functions rather than the space of parameters. Grouping classifiers into equivalence classes, we develop a standardized parameterization in which all symmetries are removed, resulting in a toroidal topology. On this space, we explore the error landscape rather than the loss. This lets us derive a meaningful notion of the flatness of minimizers and of the geodesic paths connecting them. Using different optimization algorithms that sample minimizers with different flatness we study the mode connectivity and relative distances. Testing a variety of state-of-the-art architectures and benchmark datasets, we confirm the correlation between flatness and generalization performance; we further show that in function space flatter minima are closer to each other and that the barriers along the geodesics connecting them are small. We also find that minimizers found by variants of gradient descent can be connected by zero-error paths composed of two straight lines in parameter space, i.e. polygonal chains with a single bend. We observe similar qualitative results in neural networks with binary weights and activations, providing one of the first results concerning the connectivity in this setting. Our results hinge on symmetry removal, and are in remarkable agreement with the rich phenomenology described by some recent analytical studies performed on simple shallow models
Investigations on the temperature dependent performance, degradation, and safety of Li-ion batteries
No full textThe rapid expansion of the use of Li-ion batteries (LIBs) across diverse applications, from portable electronics to battery electric vehicles and large-scale energy storage systems, underscores the need to better understand and to improve their performance, longevity, and safety. An in-depth understanding of LIB degradation processes is essential for developing advanced operating profiles to enlarge their lifetime. This is not only crucial for maximizing sustainability but also for reducing costs simultaneously. Furthermore, ensuring cell safety over the full lifespan even under diverse and extreme operating conditions is critical. This requires a detailed understanding and identification of factors such as deposition of metallic Li that causes rapid battery aging and compromises safety. LIB degradation is driven by numerous factors. They can be grouped into material and design aspects as well as operating and storage conditions. Among these, temperature is one of the most significant influencing factors on LIB degradation. At low temperatures (often < 25 °C), Li plating typically dominates degradation, while at elevated temperatures accelerated parasitic side reactions such as solid-electrolyte-interphase (SEI) growth are dominant. Both mechanisms result in significantly different safety characteristics. Finding the optimal temperature is a trade-off between both aging mechanisms.
This thesis focuses on investigations of temperature dependent aging mechanisms in LIBs using various methods of investigation. Results were reported in the following seven peer-reviewed articles. In the first article, the impact of operating conditions such as temperature, current, and aging state on LIB degradation is examined by using cylindrical 21700-type cells as well as lab-made pouch cells. Arrhenius plots of the cyclic aging rate are used to determine the crossover temperature between Li plating at low temperatures and SEI growth at high temperatures. It is shown that both, aging state and C-rate, shift the crossover temperature. Particularly higher charging C-rates shift the onset of Li plating and thus the crossover temperature towards higher values.
In typical applications, the operating temperature of a battery is not constant but fluctuates in between certain limits. However, most scientific investigations focus on aging processes at one constant temperature only. In the second article, the effect of varying temperatures during cyclic aging is examined. Results reveal that the initial aging temperature can substantially alter subsequent degradation behavior. Especially, aging at high temperatures shifts the onset of Li plating to lower temperatures in subsequent cycles. As a consequence, the optimal aging temperature is reduced and cycle life is extended. Conversely, cycling LIBs under conditions that support Li plating accelerates further Li plating.
The materials choice, particularly on the anode side, further affects the temperature dependent performance. Si-graphite composites are used as anode material in state-of-the-art LIBs. However, Si and graphite have a significantly different electrochemical behavior. The effect of a varying Si content on the temperature dependent performance and degradation is investigated in the third article. A high Si content in the anode is beneficial for a good low temperature performance. However, the high temperature performance is deteriorated by a high Si content due to a strongly increasing cyclic aging rate. Therefore, a balanced Si content in the anode is essential to optimize both, cycle life and performance.
Li deposition in LIBs is problematic, as it correlates with rapid capacity fade and safety concerns. In the fourth article, three workflows addressing whether Li deposition exists in a cell, its location, and the conditions that prevent its formation are presented. Complementary experimental techniques as well as simulation approaches to detect and prevent Li deposition are reviewed. The workflows enable accurate and efficient detection of Li deposition and provide valuable insights for improving battery design, safety, and performance in demanding applications.
Safety of LIBs is significantly influenced by their aging history. In the fifth article, the influence of the dominant aging mechanism and state of health (SOH) on the safety of Li-ion pouch cells is investigated. Accelerating rate calorimetry (ARC) is used to determine safety critical temperatures. In order to gain further insights into thermal decomposition processes and enable early detection of critical events, the ARC method is expanded by multiple low-cost sensors installed on and around the pouch cell. Aging associated with Li plating lowers safety by reducing the onset temperature of self-heating (SH), whereas aging at elevated temperatures raises the onset temperature of SH.
Typically, temperature is measured on the surface of battery cells during operation or safety testing. However, internal temperature can significantly differ from surface temperature. Therefore, a method is presented in the sixth article that uses laser ablation to insert a thermocouple inside a 21700-type cell. ARC experiments of these cells at different state of charge (SOC) provide insights into thermal gradients. The internal temperature changes much faster and to a greater extent than the surface temperature, especially during venting and thermal runaway (TR).
Safety of LIBs is influenced by a number of factors including SOC and aging state. In the seventh article, the safety characteristics of cylindrical 21700-type cells cyclically aged at different temperatures in conjunction with varying SOC are examined by using advanced ARC measurements. The findings indicate that both elevated SOC and cell degradation lower the safety level, both by reducing critical temperatures that necessitate venting and the onset of TR. Furthermore, an increase in destruction level and mass loss of the cell after ARC testing is observed with increasing SOC. Notably, cells exhibiting Li plating show a markedly diminished onset temperature of SH, independent of SOC.
Overall, this work provides for the first time comprehensive insights into temperature dependent LIB degradation, with a particular emphasis on the safety of aged cells. These insights are crucial for a safe and efficient operation and can be used to develop improved next-generation LIBs
Zinc finger proteins and the 3D organization of chromosomes.
No full textZinc finger domains are one of the most common structural motifs in eukaryotic cells, which employ the motif in some of their most important proteins (including TFIIIA, CTCF, and ZiF268). These DNA binding proteins contain up to 37 zinc finger domains connected by flexible linker regions. They have been shown to be important organizers of the 3D structure of chromosomes and as such are called the master weaver of the genome. Using NMR and numerical simulations, much progress has been made during the past few decades in understanding their various functions and their ways of binding to the DNA, but a large knowledge gap remains to be filled. One problem of the hitherto existing theoretical models of zinc finger protein DNA binding in this context is that they are aimed at describing specific binding. Furthermore, they exclusively focus on the microscopic details or approach the problem without considering such details at all. We present the Flexible Linker Model, which aims explicitly at describing nonspecific binding. It takes into account the most important effects of flexible linkers and allows a qualitative investigation of the effects of these linkers on the nonspecific binding affinity of zinc finger proteins to DNA. Our results indicate that the binding affinity is increased by the flexible linkers by several orders of magnitude. Moreover, they show that the binding map for proteins with more than one domain presents interesting structures, which have been neither observed nor described before, and can be interpreted to fit very well with existing theories of facilitated target location. The effect of the increased binding affinity is also in agreement with recent experiments that until now have lacked an explanation. We further explore the class of proteins with flexible linkers, which are unstructured until they bind. We have developed a methodology to characterize these flexible proteins. Employing the concept of barcodes, we propose a measure to compare such flexible proteins in terms of a similarity measure. This measure is validated by a comparison between a geometric similarity measure and the topological similarity measure that takes geometry as well as topology into account. Adv Protein Chem Struct Biol 2013; 90:67-117
Antjie Krog, author and translator: the twain does meet.
No full textPlease help us populate SUNScholar with the post print version of this article. It can be e-mailed to: [email protected] En WysbegeerteAfrikaans En Nederland
Arrhenius plots for Li-ion battery ageing as a function of temperature, C-rate, and ageing state – An experimental study
No full textGints Kucinskis acknowledges Latvian Council of Science project “Cycle life prediction of lithium-ion battery electrodes and cells, utilizing current-voltage response measurements”, project No. LZP-2020/1–0425. Institute of Solid-State Physics, University of Latvia as the Centre of Excellence has received funding from the European Union's Horizon 2020 Framework Program H2020-WIDESPREAD-01-2016-2017-TeamingPhase2 under grant agreement No. 739508, project CAMART2.
ZSW acknowledges funding of the project GradiBatt by the industrial collective research programme (IGF no. 20884 N/2) which was supported by the German Federal Ministry for Economic Affairs and Climate Action (BMWK) through the AiF (German Federation of Industrial Research Associations eV) and of the projects RollBatt (03XP0245A) and CharLiSiKo (03XP0333A) funded by the German Federal Ministry of Education and Research (BMBF).We present an extensive analysis of Li-ion battery ageing via Arrhenius plots. The V-shaped Arrhenius plots show minima at an optimum temperature corresponding to the longest cycle-life. The V-shape of the Arrhenius plots signifies the crossover between the two dominating ageing mechanisms – solid electrolyte interphase (SEI) growth in the high temperature range and lithium deposition in the low temperature range. Subjects of our investigations are commercial 5 Ah high energy 21,700-type cells with LiNi0.90Co0.05Al0.05O2 + LiNiO2 (NCA + LNO) cathode and Si/graphite anode (∼3% Si) and 0.1 Ah lab-made pouch cells with LiNi1/3Mn1/3Co1/3O2 (NMC111) cathode and a graphite anode. The results of the Arrhenius plots are analysed in the context of C-rate, cell ageing, and electrode properties. We find that the crossover between the dominating ageing mechanism and hence the optimum operating temperature of the Li-ion cells depend on C-rate, anode coating thickness/particles sizes, the state of health, and the shape of the capacity fade curve. Considering the change of the dominant ageing mechanism can help designing battery systems with longer service life. Additionally, we show a lifetime estimation for temperature dependent cycling of batteries emphasizing the merit of Arrhenius plots in the context of battery cell ageing. --//-- Gints Kucinskis, Maral Bozorgchenani, Max Feinauer, Michael Kasper, Margret Wohlfahrt-Mehrens, Thomas Waldmann, Arrhenius plots for Li-ion battery ageing as a function of temperature, C-rate, and ageing state – An experimental study, Journal of Power Sources, Volume 549, 2022, 232129, ISSN 0378-7753, https://doi.org/10.1016/j.jpowsour.2022.232129. Published under the CC BY licence.Latvian Council of Science project No. LZP-2020/1–0425; Institute of Solid-State Physics, University of Latvia as the Centre of Excellence has received funding from the European Union's Horizon 2020 Framework Program H2020-WIDESPREAD-01-2016-2017-TeamingPhase2 under grant agreement No. 739508, project CAMART2; project GradiBatt (IGF no. 20884 N/2), RollBatt (03XP0245A) and CharLiSiKo (03XP0333A)
Ideological considerations in the Afrikaans translation of Anne Frank's diary, Het Achterhuis [Ideologiese oorwegings by die vertaling van Anne Frank se dagboek, Het Achterhuis, in Afrikaans]
No full textIdeological considerations in the Afrikaans translation of Anne Frank's diary, Het Achterhuis This article investigates the ideological considerations in the decision making process of the translation of Anne Frank's diary, Het Achterhuis, into Afrikaans, Die Agterhuis. It analyses how the translator succeeds in rendering a translation which is authentic to the source text by remaining faithful to it and by conveying the voice and style of the original author into Afrikaans. In 2008, more than 60 years after Anne Frank's diary, Het Achterhuis, was first published in Dutch (1947), Lina Spies translated it into Afrikaans, with the title, Die Agterhuis. For the first time this famous diary was made available in Afrikaans. This diary has been rewritten and republished in Dutch a number of times as well as translated into over 60 languages. It has also been reworked into different genres, such as musicals and plays. Quite a number of these rewrites and translations rendered versions of the diary that were adapted and manipulated to such an extent that the resulting target texts can be regarded as unfaithful to the original/source text. These manipulations are mostly ideological in nature and mainly entail omitting and/or down toning entries in the diary which are of a strong ideological nature. In his adaptation of the diary, Anne's father, Otto Frank, for example omits entries where Anne depicted her mother in a bad light. Quite a number of the rewrites and translations omit the entries where Anne wrote from a feministic viewpoint about the emancipation of women (Lefevere 1992:64). Especially in German translations the entries where Anne wrote about the severity of the war and the Holocaust, which put the Nazis and Gestapo in a bad light, were rewritten to come across as less severe than they really were. An example of this is Anne's description of the German police on how to kill hostages: "zet de Gestapo doodgewoon een stuk of 5 gijzelaars tegen de muur" (the Gestapo simply puts 5 hostages or so against the wall) which was translated into German as: "dann hat man einen Grund, eine Anzahl dieser Geiseln zu erschiessen" (one then has a reason to shoot a number of these hostages). Lefevere (1992:68) argues that by using the word "man" (one) instead of specifically "Gestapo", the description is less explicit. The link between the severe deed and the German Gestapo is non-existent. Furthermore, the deed is justified in the target text because it is mentioned "das man einen Grund [hat]" to do it. Lefevere (1992:68) also mentions that by using merely "erschiessen" (shoot) instead of conveying the idea of the people being lined up against a wall, makes the deed less shocking. The inhumanity of the act which is conveyed in the Dutch source text is therefore lost in the ideologically manipulated German target text. Because the diary of Anne Frank is an ego document/autobiographical work (in the form of a diary) where the thoughts and feelings of a real person are portrayed, as well as a text which serves as, a type of documentation of the Holocaust, it is of great importance that rewrites and translations of the diary be as faithful as possible to the source text. The manipulated versions of the diary led to the falsification of the voice of the author, Anne Frank, which Lefevere (1992) refers to as "the construction of different Anne Franks". The translator, Spies, as an agent of power in the translation process, believes that the nature of an autobiographical work and a documentation of The Second World War and the Holocaust call for authenticity and that the translator of these genres is therefore compelled to be subservient, to an extent, to the authority of the source text and the source text author. Spies therefore tries to remain as close to the source text as possible, whilst translating it into idiomatic Afrikaans (Spies 2008a:14). According to Spies (2007:10) the purpose of her Afrikaans translation is to serve as a reaction against the manipulated versions of the diary, which led to the falsification thereof and the falsely constructed image of Anne presented to the world. In order to render a target text which is authentic to the source text and reflects the voice and writing style of the source text author, the overall approach Spies followed in the Afrikaans translation in terms of culture and ideology is one of foreignization (Venuti 1995). This approach entails that a target text is produced by purposefully breaking conventions of both the target text language and culture to retain a part of the strangeness of the source text in the target text (Venuti 1995). All pragmatic and contextual references, such as cultural, political and religious references in the source text are therefore preserved in the target text and rendered as faithfully as possible. Translation strategies such as calques and cultural borrowing are applied and the translator makes use of older or Netherlandistic Afrikaans expressions, words and idioms in order to convey the voice and style of the author. The purpose is to make the target text seem as if it were written by the original author, but in Afrikaans. Because the target text is mainly foreignized, Spies frequently made use of paratext (the introduction, footnotes, epilogue and afterword) which accompany the diary entries to provide background information on the source text and author, in order to justify certain decisions made during the translation process and to clarify or explain some references in the target text which the target text readership will most likely not understand or be familiar with (Spies 2011). The combination offoreignization and paratext makes the target text more faithful to the source text, whilst rendering it acceptable, accessible and comprehensible to the target text readership. The ideological considerations in the translation of Het Achterhuis in Afrikaans are mainly aimed at rendering an authentic and faithful translation. By means of the Afrikaans translation of the diary, Anne's original work becomes accessible to the Afrikaans speaking, South African, target text readership. Justice is done to the diary and to the voice of Anne Frank - the face of the six million Jews who died during the Holocaust.Revie
- …
