Search

Your search keyword '"Kramar A"' showing total 7,447 results
Start Over Author "Kramar A"
7,447 results on '"Kramar A"'

1. Dynamic cost allocation allows network-forming forager to switch between search strategies

Author

Schick, Lisa, Kramar, Mirna, and Alim, Karen

Subjects
Physics - Biological Physics, Physics - Fluid Dynamics
Abstract

Network-forming organisms, like fungi and slime molds, dynamically reorganize their networks during foraging. The resulting re-routing of resource flows within the organism's network can significantly impact local ecosystems. In current analysis limitations stem from a focus on single-time-point morphology, hindering understanding of continuous dynamics and underlying constraints. Here, we study ongoing network reorganization in the foraging slime mold \textit{Physarum~polycephalum}, identifying three distinct states with varying morphology and migration velocity. We estimate the energetic cost of each state and find a trade-off between building and transport costs within the morphological variability, facilitating different search strategies. Adaptation of state population to the environment suggests that diverse network morphologies support varied foraging strategies, though constrained by associated costs. Our findings provide insights for evaluating the impact of resource flow re-routing in changing ecosystems.

Published
2024
Full Text
View/download PDF

2. Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2

Author

Lieberum, Tom, Rajamanoharan, Senthooran, Conmy, Arthur, Smith, Lewis, Sonnerat, Nicolas, Varma, Vikrant, Kramár, János, Dragan, Anca, Shah, Rohin, and Nanda, Neel

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computation and Language
Abstract

Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features. Despite recent excitement about their potential, research applications outside of industry are limited by the high cost of training a comprehensive suite of SAEs. In this work, we introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models. We primarily train SAEs on the Gemma 2 pre-trained models, but additionally release SAEs trained on instruction-tuned Gemma 2 9B for comparison. We evaluate the quality of each SAE on standard metrics and release these results. We hope that by releasing these SAE weights, we can help make more ambitious safety and interpretability research easier for the community. Weights and a tutorial can be found at https://huggingface.co/google/gemma-scope and an interactive demo can be found at https://www.neuronpedia.org/gemma-scope, Comment: 12 main text pages, and 14 pages of acknowledgements, references and appendices

Published
2024

3. Ukrainian PhD Students' Attitudes toward AI Language Processing Tools in the Context of English for Academic Purposes

Author

Natalie Kramar, Yaroslava Bedrych, and Zinaida Shelkovnikova

Abstract

Mastering academic writing skills in English is essential for future researchers. At present, AI language processing tools provide high-quality, accessible, and fast assistance for translation, editing, and stylistic enhancement of scientific texts. However, their use within English for Academic Purposes (EAP) courses generates mixed reactions among educators and raises ethical concerns. Our study aimed to explore the predominant perceptions of AI language processing tools by PhD students of the National Academy of Sciences of Ukraine (NASU) from the viewpoint of their integration into the academic English course taught in the first year of their PhD studies. The study involved 52 PhD students from various NASU institutes. They completed a survey with both closed-ended and open-ended questions regarding their previous and expected use of online translators, writing enhancement tools, and ChatGPT for research writing purposes. The results of the survey show that NASU PhD students have extensive experience with online translators, but are less familiar with writing enhancement tools and less certain about their potential use in the future. Almost a third of the respondents expressed reservations about using ChatGPT for academic writing due to academic integrity concerns. Most of the respondents (66%) agree that the basics of ethical use of AI writing and editing tools should be incorporated into EAP courses. One subgroup of the participants (n = 11) took part in a small-scale additional intervention focused on writing enhancement tools. They were asked to apply Grammarly, QuillBot, and ChatGPT to edit their course projects (presentations of the current state of their dissertation research) and compare these tools according to various criteria. The feedback provided by this subgroup indicates that they were most satisfied with the quality of editing provided by ChatGPT but found Grammarly and QuillBot easier to use and more suitable for superficial grammar checks. We found out that the AI tools helped participants achieve improvements primarily in such aspects as the use of articles, punctuation, use of prepositions, and elimination of redundancy. The study has significant pedagogical implications, promoting the wider use of AI tools in the context of teaching English for Academic Purposes and addressing appropriate teaching techniques and methods.

Published
2024

4. Obtaining of medical textiles based on viscose and chitosan with simultaneously improved sorption and antibacterial properties by using dielectric barrier discharge

Author

Korica Matea, Kramar Ana, Peršin Zdenka, Obradović Bratislav, Kuraica Milorad, Fras-Zemljič Lidija, and Kostić Mirjana

Subjects
viscose, chitosan, dielectric barrier discharge, sorption properties, antibacterial properties, medical textiles, Textile bleaching, dyeing, printing, etc., TP890-933
Abstract

This study aimed to obtain textile based on viscose and chitosan with simultaneously improved sorption and antibacterial properties. For that reason, before functionalization with chitosan, the viscose fabric was modified by nonthermal gas plasma of dielectric barrier discharge (DBD) to introduce into its structure functional groups suitable for improved adsorption of chitosan. Functional groups' content of unmodified and DBD modified viscose fabric was determined by the calcium acetate method, while their influence on the surface charge, as well as the success of the functionalization of viscose fabrics with chitosan, was evaluated by measuring the zeta potential. Morphological, sorption and antibacterial properties of unmodified and DBD modified viscose fabrics, before and after functionalization with chitosan, were characterized by scanning electron microscopy, capillary rise method, and standard test method ASTME 2149-01: 2001, respectively. The obtained results showed that the treatment with DBD contributed to the simultaneous improvement of sorption and antibacterial properties of viscose fabric functionalized with chitosan, making it suitable for the production of high value-added medical textiles.

Published
2021
Full Text
View/download PDF

5. Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Author

Rajamanoharan, Senthooran, Lieberum, Tom, Sonnerat, Nicolas, Conmy, Arthur, Varma, Vikrant, Kramár, János, and Nanda, Neel

Subjects
Computer Science - Machine Learning
Abstract

Sparse autoencoders (SAEs) are a promising unsupervised approach for identifying causally relevant and interpretable linear features in a language model's (LM) activations. To be useful for downstream tasks, SAEs need to decompose LM activations faithfully; yet to be interpretable the decomposition must be sparse -- two objectives that are in tension. In this paper, we introduce JumpReLU SAEs, which achieve state-of-the-art reconstruction fidelity at a given sparsity level on Gemma 2 9B activations, compared to other recent advances such as Gated and TopK SAEs. We also show that this improvement does not come at the cost of interpretability through manual and automated interpretability studies. JumpReLU SAEs are a simple modification of vanilla (ReLU) SAEs -- where we replace the ReLU with a discontinuous JumpReLU activation function -- and are similarly efficient to train and run. By utilising straight-through-estimators (STEs) in a principled manner, we show how it is possible to train JumpReLU SAEs effectively despite the discontinuous JumpReLU function introduced in the SAE's forward pass. Similarly, we use STEs to directly train L0 to be sparse, instead of training on proxies such as L1, avoiding problems like shrinkage., Comment: v2: new appendix H comparing kernel functions & bug-fixes to pseudo-code in Appendix J v3: further bug-fix to pseudo-code in Appendix J

Published
2024

6. On scalable oversight with weak LLMs judging strong LLMs

Author

Kenton, Zachary, Siegel, Noah Y., Kramár, János, Brown-Cohen, Jonah, Albanie, Samuel, Bulian, Jannis, Agarwal, Rishabh, Lindner, David, Tang, Yunhao, Goodman, Noah D., and Shah, Rohin

Subjects
Computer Science - Machine Learning
Abstract

Scalable oversight protocols aim to enable humans to accurately supervise superhuman AI. In this paper we study debate, where two AI's compete to convince a judge; consultancy, where a single AI tries to convince a judge that asks questions; and compare to a baseline of direct question-answering, where the judge just answers outright without the AI. We use large language models (LLMs) as both AI agents and as stand-ins for human judges, taking the judge models to be weaker than agent models. We benchmark on a diverse range of asymmetries between judges and agents, extending previous work on a single extractive QA task with information asymmetry, to also include mathematics, coding, logic and multimodal reasoning asymmetries. We find that debate outperforms consultancy across all tasks when the consultant is randomly assigned to argue for the correct/incorrect answer. Comparing debate to direct question answering, the results depend on the type of task: in extractive QA tasks with information asymmetry debate outperforms direct question answering, but in other tasks without information asymmetry the results are mixed. Previous work assigned debaters/consultants an answer to argue for. When we allow them to instead choose which answer to argue for, we find judges are less frequently convinced by the wrong answer in debate than in consultancy. Further, we find that stronger debater models increase judge accuracy, though more modestly than in previous studies., Comment: 15 pages (53 including appendices). V2: minor correction to Figure 3; add Figure A.9 comparing open vs assigned consultancy; add a reference

Published
2024

7. Dirac operators on the half-line: stability of spectrum and non-relativistic limit

Author

Kramar, David and Krejcirik, David

Subjects
Mathematics - Spectral Theory, Mathematical Physics, Mathematics - Analysis of PDEs, Quantum Physics
Abstract

We consider Dirac operators on the half-line, subject to generalised infinite-mass boundary conditions. We derive sufficient conditions which guarantee the stability of the spectrum against possibly non-self-adjoint potential perturbations and study the optimality of the obtained results. Finally, we establish a non-relativistic limit which makes a relationship of the present model to the Robin Laplacian on the half-line., Comment: 17 pages, 1 figure

Published
2024

8. Improving Dictionary Learning with Gated Sparse Autoencoders

Author

Rajamanoharan, Senthooran, Conmy, Arthur, Smith, Lewis, Lieberum, Tom, Varma, Vikrant, Kramár, János, Shah, Rohin, and Nanda, Neel

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity., Comment: 15 main text pages, 22 appendix pages

Published
2024

9. AtP*: An efficient and scalable method for localizing LLM behaviour to components

Author

Kramár, János, Lieberum, Tom, Shah, Rohin, and Nanda, Neel

Subjects
Computer Science - Machine Learning, Computer Science - Computation and Language
Abstract

Activation Patching is a method of directly computing causal attributions of behavior to model components. However, applying it exhaustively requires a sweep with cost scaling linearly in the number of model components, which can be prohibitively expensive for SoTA Large Language Models (LLMs). We investigate Attribution Patching (AtP), a fast gradient-based approximation to Activation Patching and find two classes of failure modes of AtP which lead to significant false negatives. We propose a variant of AtP called AtP*, with two changes to address these failure modes while retaining scalability. We present the first systematic study of AtP and alternative methods for faster activation patching and show that AtP significantly outperforms all other investigated methods, with AtP* providing further significant improvement. Finally, we provide a method to bound the probability of remaining false negatives of AtP* estimates.

Published
2024

12. C5aR1 signaling promotes region‐ and age‐dependent synaptic pruning in models of Alzheimer's disease

Author

Gomez‐Arboledas, Angela, Fonseca, Maria I, Kramar, Enikö, Chu, Shu‐Hui, Schartz, Nicole D, Selvan, Purnika, Wood, Marcelo A, and Tenner, Andrea J

Subjects
Biomedical and Clinical Sciences, Neurosciences, Clinical Sciences, Dementia, Acquired Cognitive Impairment, Alzheimer's Disease including Alzheimer's Disease Related Dementias (AD/ADRD), Genetics, Aging, Brain Disorders, Alzheimer's Disease, Neurodegenerative, 2.1 Biological and endogenous factors, Aetiology, Neurological, Mice, Animals, Humans, Alzheimer Disease, Synapses, Long-Term Potentiation, Disease Models, Animal, Alzheimer's disease, C1q, C5aR1, LTP, microglia, synaptic pruning, Geriatrics, Clinical sciences, Biological psychology
Abstract

IntroductionSynaptic loss is a hallmark of Alzheimer's disease (AD) that correlates with cognitive decline in AD patients. Complement-mediated synaptic pruning has been associated with this excessive loss of synapses in AD. Here, we investigated the effect of C5aR1 inhibition on microglial and astroglial synaptic pruning in two mouse models of AD.MethodsA combination of super-resolution and confocal and tridimensional image reconstruction was used to assess the effect of genetic ablation or pharmacological inhibition of C5aR1 on the Arctic48 and Tg2576 models of AD.ResultsGenetic ablation or pharmacological inhibition of C5aR1 partially rescues excessive pre-synaptic pruning and synaptic loss in an age and region-dependent fashion in two mouse models of AD, which correlates with improved long-term potentiation (LTP).DiscussionReduction of excessive synaptic pruning is an additional beneficial outcome of the suppression of C5a-C5aR1 signaling, further supporting its potential as an effective targeted therapy to treat AD.HighlightsC5aR1 ablation restores long-term potentiation in the Arctic model of AD. C5aR1 ablation rescues region specific excessive pre-synaptic loss. C5aR1 antagonist, PMX205, rescues VGlut1 loss in the Tg2576 model of AD. C1q tagging is not sufficient to induce VGlut1 microglial ingestion. Astrocytes contribute to excessive pre-synaptic loss at late stages of the disease.

Published
2024

13. Explaining grokking through circuit efficiency

Author

Varma, Vikrant, Shah, Rohin, Kenton, Zachary, Kramár, János, and Kumar, Ramana

Subjects
Computer Science - Machine Learning
Abstract

One of the most surprising puzzles in neural network generalisation is grokking: a network with perfect training accuracy but poor generalisation will, upon further training, transition to perfect generalisation. We propose that grokking occurs when the task admits a generalising solution and a memorising solution, where the generalising solution is slower to learn but more efficient, producing larger logits with the same parameter norm. We hypothesise that memorising circuits become more inefficient with larger training datasets while generalising circuits do not, suggesting there is a critical dataset size at which memorisation and generalisation are equally efficient. We make and confirm four novel predictions about grokking, providing significant evidence in favour of our explanation. Most strikingly, we demonstrate two novel and surprising behaviours: ungrokking, in which a network regresses from perfect to low test accuracy, and semi-grokking, in which a network shows delayed generalisation to partial rather than perfect test accuracy.

Published
2023

14. The Hydra Effect: Emergent Self-repair in Language Model Computations

Author

McGrath, Thomas, Rahtz, Matthew, Kramar, Janos, Mikulik, Vladimir, and Legg, Shane

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computation and Language
Abstract

We investigate the internal structure of language model computations using causal analysis and demonstrate two motifs: (1) a form of adaptive computation where ablations of one attention layer of a language model cause another layer to compensate (which we term the Hydra effect) and (2) a counterbalancing function of late MLP layers that act to downregulate the maximum-likelihood token. Our ablation studies demonstrate that language model layers are typically relatively loosely coupled (ablations to one layer only affect a small number of downstream layers). Surprisingly, these effects occur even in language models trained without any form of dropout. We analyse these effects in the context of factual recall and consider their implications for circuit-level attribution in language models.

Published
2023

15. Intruder in a two-dimensional granular system: statics and dynamics of force networks in an experimental system experiencing stick-slip dynamics

Author

Basak, R., Kozlowski, R., Pugnaloni, L. A., Kramar, M., Socolar, E. S., Carlevaro, C. M., and Kondic, L.

Subjects
Condensed Matter - Soft Condensed Matter
Abstract

In quasi-two-dimensional experiments with photoelastic particles confined to an annular region, an intruder constrained to move in a circular path halfway between the annular walls experiences stick-slip dynamics. We discuss the response of the granular medium to the driven intruder, focusing on the evolution of the force network during sticking periods. Because the available experimental data does not include precise information about individual contact forces, we use an approach developed in our previous work (Basak et al, J. Eng. Mechanics (2021)) based on networks constructed from measurements of the integrated strain magnitude on each particle. These networks are analyzed using topological measures based on persistence diagrams, revealing that force networks evolve smoothly but in a nontrivial manner throughout each sticking period, even though the intruder and granular particles are stationary. Characteristic features of persistence diagrams show identifiable changes as a slip is approaching, indicating the existence of slip precursors. Key features of the dynamics are similar for granular materials composed of disks or pentagons, but some details are consistently different. In particular, we find significantly larger fluctuations of the measures computed based on persistence diagrams, and therefore of the underlying networks, for systems of pentagonal particles.

Published
2023

16. Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla

Author

Lieberum, Tom, Rahtz, Matthew, Kramár, János, Nanda, Neel, Irving, Geoffrey, Shah, Rohin, and Mikulik, Vladimir

Subjects
Computer Science - Machine Learning
Abstract

\emph{Circuit analysis} is a promising technique for understanding the internal mechanisms of language models. However, existing analyses are done in small models far from the state of the art. To address this, we present a case study of circuit analysis in the 70B Chinchilla model, aiming to test the scalability of circuit analysis. In particular, we study multiple-choice question answering, and investigate Chinchilla's capability to identify the correct answer \emph{label} given knowledge of the correct answer \emph{text}. We find that the existing techniques of logit attribution, attention pattern visualization, and activation patching naturally scale to Chinchilla, allowing us to identify and categorize a small set of `output nodes' (attention heads and MLPs). We further study the `correct letter' category of attention heads aiming to understand the semantics of their features, with mixed results. For normal multiple-choice question answers, we significantly compress the query, key and value subspaces of the head without loss of performance when operating on the answer labels for multiple-choice questions, and we show that the query and key subspaces represent an `Nth item in an enumeration' feature to at least some extent. However, when we attempt to use this explanation to understand the heads' behaviour on a more general distribution including randomized answer labels, we find that it is only a partial explanation, suggesting there is more to learn about the operation of `correct letter' heads on multiple choice question answering.

Published
2023

17. The physiological interactome of TCR-like antibody therapeutics in human tissues

Author

Marrer-Berger, Estelle, Nicastri, Annalisa, Augustin, Angelique, Kramar, Vesna, Liao, Hanqing, Hanisch, Lydia Jasmin, Carpy, Alejandro, Weinzierl, Tina, Durr, Evelyne, Schaub, Nathalie, Nudischer, Ramona, Ortiz-Franyuti, Daniela, Breous-Nystrom, Ekaterina, Stucki, Janick, Hobi, Nina, Raggi, Giulia, Cabon, Lauriane, Lezan, Emmanuelle, Umaña, Pablo, Woodhouse, Isaac, Bujotzek, Alexander, Klein, Christian, and Ternette, Nicola

Published
2024
Full Text
View/download PDF

19. Combining venetoclax and azacytidine with T-cell bispecific antibodies for treatment of acute myeloid leukemia: a preclinical assessment

Author

Hänel, Gerulf, Schönle, Anne, Neumann, Anne-Sophie, Nixdorf, Daniel, Philipp, Nora, Sponheimer, Monika, Leutbecher, Alexandra, Emhardt, Alica-Joana, Magno, Giulia, Bücklein, Veit, Eckmann, Jan, Dunshee, Diana, Kramar, Vesna, Korfi, Koorosh, Colombetti, Sara, Umaña, Pablo, Klein, Christian, and Subklewe, Marion

Published
2024
Full Text
View/download PDF

20. Power-seeking can be probable and predictive for trained agents

Author

Krakovna, Victoria and Kramar, Janos

Subjects
Computer Science - Artificial Intelligence
Abstract

Power-seeking behavior is a key source of risk from advanced AI, but our theoretical understanding of this phenomenon is relatively limited. Building on existing theoretical results demonstrating power-seeking incentives for most reward functions, we investigate how the training process affects power-seeking incentives and show that they are still likely to hold for trained agents under some simplifying assumptions. We formally define the training-compatible goal set (the set of goals consistent with the training rewards) and assume that the trained agent learns a goal from this set. In a setting where the trained agent faces a choice to shut down or avoid shutdown in a new situation, we prove that the agent is likely to avoid shutdown. Thus, we show that power-seeking incentives can be probable (likely to arise for trained agents) and predictive (allowing us to predict undesirable behavior in new situations).

Published
2023

21. Best Practices Application in the Development of Industry in Russia

Author

Demidenko Alexander, Kramar Anastasia, and Demidenko Igor

Subjects
Engineering (General). Civil engineering (General), TA1-2040
Abstract

This article is devoted to the analysis of the problems of the Russian economy in modern conditions, in particular, the technical and technological backwardness of Russia from the developed countries. The analysis of the reasons for the low level of competitiveness of the national economy has been carried out, and ways of its possible improvement have been investigated. The necessity of changing the export-raw model of economic development, based on the accelerated increase of fuel and raw materials exports, on the model based on the production of high-tech competitive products is shown. A study of the world’s best practices in the development of industry and the agro-industrial complex has been conducted, and statistics of economic indicators of other countries is given. This article analyses the results obtained by these countries. Also, the most effective principles of economic management, which could be applied in Russia are identified. It is substantiated that the Russian Federation needs to intensify innovation. It is revealed that the strategy of innovative development of the agri-food and industrial complex should be determined on the basis of the state innovation policy, the main task of which remains to mobilize the potential of the scientific and technological potential of the industry for the technical and technological renewal of domestic industry and agro-industrial complex.

Published
2019
Full Text
View/download PDF

22. ENERGY CONSUMPTION ANALYSIS AND PARAMETER OPTIMIZATION IN HIGH-FEED MILLING OPERATION

Author

Branislav Sredanovic, Stevo Borojevic, Djordje Cica, Sasa Tesic, Dajana Jokic, and Davorin Kramar

Subjects
high-feed milling, energy consumption, modelling, optimization, Engineering (General). Civil engineering (General), TA1-2040
Abstract

Improving the performance, efficiency and productivity of the machining, reducing energy consumption and establishing control over the machining process are the main goals of the production technologies development. This paper provides the basis for establishing of a process management and determining of optimal conditions for performing high-feed milling, as a special machining method. An experimental study of the analysis of the influence of process parameters, namely depth of cut and feed rate, on the output parameters of high-feed milling were carried out. As output indicators of the performance of the high-feed milling, namely the energy efficiency of the machine tool and the productivity of the machining process were analysed. Based on experimental data, analysis and modelling of the material removal rate per unit of time and electrical energy consumption were carried out. A workpiece made from aluminium was machined on a three-axis machining centre. By using the boundaries of the domain of process parameters and the optimization objective function, the optimization of the values of the process parameters was carried out. As an optimal solution, the following process parameters were obtained: cutting depth ap = 2 mm and feed rate vf = 1500 mm/min.

Published
2024
Full Text
View/download PDF

24. Reporting Verbs Argue, Claim and Believe in Linguistics Research Articles: A Corpus-Based Study

Author

Ilchenko, Olga and Kramar, Natalie

Abstract

No research article is imaginable without reporting verbs. They help the writer voice his/her own views against the backdrop of other community voices, while also serving to project his/her authority and expertise in the field. The vast variety of reporting verbs in academic discourse and their disciplinary specificity account for a huge challenge this topic can pose for EAL academic writers. In this study, we analyzed the rhetorical and discursive functions of strong reporting verbs "argue," "claim" and "believe" in a corpus of 40 articles from leading Linguistics journals, with close attention to their Self/Other references. It has been revealed that "argue" and "claim" are mostly used in this realm to refer to other authors, while "believe" is slightly more common with reference to the writer himself/herself. While it tends to be perceived as interchangeable with "argue," "claim" is almost exclusively used to introduce the opinion of others; besides, it is often an opinion that the writer does not fully support. Also, we have shown that it is typical for "argue" to play a discourse-organizing role, being featured in retrospective endophoric markers. We believe it is important to help novice academic writers understand the subtle nuances in the semantics of these reporting verbs and their discourse functions.

Published
2022

25. Tracr: Compiled Transformers as a Laboratory for Interpretability

Author

Lindner, David, Kramár, János, Farquhar, Sebastian, Rahtz, Matthew, McGrath, Thomas, and Mikulik, Vladimir

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Statistics - Machine Learning
Abstract

We show how to "compile" human-readable programs into standard decoder-only transformer models. Our compiler, Tracr, generates models with known structure. This structure can be used to design experiments. For example, we use it to study "superposition" in transformers that execute multi-step algorithms. Additionally, the known structure of Tracr-compiled models can serve as ground-truth for evaluating interpretability methods. Commonly, because the "programs" learned by transformers are unknown it is unclear whether an interpretation succeeded. We demonstrate our approach by implementing and examining programs including computing token frequencies, sorting, and parenthesis checking. We provide an open-source implementation of Tracr at https://github.com/google-deepmind/tracr., Comment: Presented at NeurIPS 2023 (Spotlight)

Published
2023

27. Functionalization of silk with actinomycins from Streptomyces anulatus BV365 for biomedical applications

Author

Tatjana Ilic-Tomic, Ana Kramar, Mirjana Kostic, Sandra Vojnovic, Jelena Milovanovic, Milos Petkovic, Paul M. D’Agostino, Tobias A. M. Gulder, and Jasmina Nikodinovic-Runic

Subjects
silk, antimicrobial, Streptomyces, actinomycins, nonactin, functional biomaterials, Biotechnology, TP248.13-248.65
Abstract

Silk, traditionally acclaimed as the “queen of fiber,” has been widely used thanks to its brilliant performance such as gentleness, smoothness and comfortableness. Owing to its mechanical characteristics and biocompatibility silk has a definitive role in biomedical applications, both as fibroin and fabric. In this work, the simultaneous dyeing and functionalization of silk fabric with pigments from Streptomyces anulatus BV365 were investigated. This strain produced high amounts of orange extracellular pigments on mannitol-soy flour agar, identified as actinomycin D, C2 and C3. The application of purified actinomycins in the dyeing of multifiber fabric was assessed. Actinomycins exhibited a high affinity towards protein fibers (silk and wool), but washing durability was maintained only with silk. Acidic condition (pH5) and high temperature (65°C) facilitated the silk dyeing. The morphologies and chemical components of the treated silk fabrics were analyzed using scanning electron microscopy and Fourier transform infrared spectroscopy. The results showed the pigments bind to the silk through interaction with the carbonyl group in silk fibroin rendering the functionalized, yet surface that does not cause skin irritation. The treated silk exhibited a remarkable antibacterial effect, while the biocompatibility test performed with 3D-reconstructed human epidermis model indicated safe biological properties, paving the way for future application of this material in medicine.

Published
2024
Full Text
View/download PDF

31. Uncovering the protective neurological mechanisms of hypofractionated FLASH radiotherapy

Author

Alaghband, Yasaman, Allen, Barrett D, Kramar, Eniko A, Zhang, Richard, Drayson, Olivia GG, Ru, Ning, Petit, Benoit, Almeida, Aymeric, Doan, Ngoc-Lien, Wood, Marcelo, Baulch, Janet E, Ballesteros-Zebadua, Paola, Vozenin, Marie-Catherine, and Limoli, Charles L

Subjects
Biological Psychology, Biomedical and Clinical Sciences, Oncology and Carcinogenesis, Psychology, Neurosciences, Behavioral and Social Science, Cancer, Neurological, Male, Mice, Female, Animals, Neuroinflammatory Diseases, Long-Term Potentiation, Neuronal Plasticity, Radiation Dose Hypofractionation
Abstract

Implementation of ultra-high dose-rate FLASH radiotherapy (FLASH-RT) is rapidly gaining traction as a unique cancer treatment modality able to dramatically minimize normal tissue toxicity while maintaining antitumor efficacy compared with standard-of-care radiotherapy at conventional dose rate (CONV-RT). The resultant improvements in the therapeutic index have sparked intense investigations in pursuit of the underlying mechanisms. As a preamble to clinical translation, we exposed non-tumor-bearing male and female mice to hypofractionated (3 × 10 Gy) whole brain FLASH- and CONV-RT to evaluate differential neurologic responses using a comprehensive panel of functional and molecular outcomes over a 6-month follow-up. In each instance, extensive and rigorous behavioral testing showed FLASH-RT to preserve cognitive indices of learning and memory that corresponded to a similar protection of synaptic plasticity as measured by long-term potentiation (LTP). These beneficial functional outcomes were not found after CONV-RT and were linked to a preservation of synaptic integrity at the molecular (synaptophysin) level and to reductions in neuroinflammation (CD68+ microglia) throughout specific brain regions known to be engaged by our selected cognitive tasks (hippocampus, medial prefrontal cortex). Ultrastructural changes in presynaptic/postsynaptic bouton (Bassoon/Homer-1 puncta) within these same regions of the brain were not found to differ in response to dose rate. With this clinically relevant dosing regimen, we provide a mechanistic blueprint from synapse to cognition detailing how FLASH-RT reduces normal tissue complications in the irradiated brain.SignificanceFunctional preservation of cognition and LTP after hypofractionated FLASH-RT are linked to a protection of synaptic integrity and a reduction in neuroinflammation over protracted after irradiation times.

Published
2023

32. Stochastic methods for slip prediction in a sheared granular system

Author

Bretz, Philip, Kondic, Lou, and Kramar, Miro

Subjects
Condensed Matter - Soft Condensed Matter
Abstract

We consider a sheared granular system experiencing intermittent dynamics of stick-slip type via discrete element simulations. The considered setup consists of a two-dimensional system of soft frictional particles sandwiched between solid walls, one of which is exposed to a shearing force. The slip events are detected using stochastic state space models applied to various measures describing the system. We show that the measures describing the forces between the particles provide earlier detection of an upcoming slip event than the measures based solely on the wall movement. By comparing the detection times obtained from the considered measures, we observe that a typical slip event starts with a local change in the force network. However, some local changes do not spread globally over the force network. For the changes that become global, we find a sharp critical value for their size. If the size of a global change exceeds the critical value, then it triggers a slip event; if it does not, then a much weaker micro-slip follows. Quantification of the changes in the force network is made possible by formulating clear and precise measures describing their static and dynamic properties., Comment: 14 pages, 13 figures

Published
2022
Full Text
View/download PDF

33. Development of Integral Field Spectrographs to Revolutionize Spectroscopic Observations of Solar Flares and other Energetic Solar Eruptions

Author

Lin, Haosheng, Anan, Tetsu, Cauzzi, Gianna, Fletcher, Lyndsay, Huang, Pei, Kowalski, Adam, Kramar, Maxim, Qiu, Jiong, Samra, Jenna, Spittler, Constance, Sukegawa, Takashi, and Wirth, Gregory

Subjects
Astrophysics - Instrumentation and Methods for Astrophysics, Astrophysics - Solar and Stellar Astrophysics
Abstract

The Sun's proximity offers us a unique opportunity to study in detail the physical processes on a star's surface; however, the highly dynamic nature of the stellar surface -- in particular, energetic eruptions such as flares and coronal mass ejections -- presents tremendous observational challenges. Spectroscopy probes the physical state of the solar atmosphere, but conventional scanning spectrographs and spectrometers are unable to capture the full evolutionary history of these dynamic events with a sufficiently wide field of view and high spatial, spectral, and temporal resolution. Resolving the physics of the dynamic sun requires gathering simultaneous spectra across a contiguous area over the full duration of these events, a goal now tantalizingly close to achievable with continued investment in developing powerful new Integral Field Spectrographs to serve as the foundation of both future ground- and space-based missions. This technology promises to revolutionize our ability to study solar flares and CMEs, addressing NASA's strategic objective to "understand the Sun, solar system, and universe." Since such events generate electromagnetic radiation and high-energy particles that disrupt terrestrial electric infrastructure, this investment not only advances humanity's scientific endeavors but also enhances our space weather forecasting capability to protect against threats to our technology-reliant civilization., Comment: 14 pages, 4 figures, White Paper submitted for 2022 Solar and Space Physics (Heliophysics) Decadal Survey

Published
2022
Full Text
View/download PDF

34. Justification of the use of Stokes-V Cryo-NIRSP/DKIST observations for the 3D Reconstruction of the Coronal Magnetic Field

Author

Kramar, Maxim and Lin, Haosheng

Subjects
Astrophysics - Solar and Stellar Astrophysics
Abstract

This study presents a justification of the use of Stokes-V Cryo-NIRSP/DKIST observations for the 3D Reconstruction of the Coronal Magnetic Field. A magnetohydrodynamic (MHD) model of the solar corona during solar minimum generated by Predictive Science Inc. was used to synthesize spectropolarimetric measurements of the Fe XIII 1075 nm coronal emission line as observed from the Earth. Stokes-Q,U data is "taken" for half a solar rotation period (about two weeks) by Upgraded Coronal Multichannel Polarimeter (UCoMP) with field of view (FOV) up to $2\ R_\odot$. Stokes-V data is "taken" once a day for two days by Cryo-NIRSP/DKIST with a total FOV equal to the two Cryo-NIRSP FOVs. We demonstrated that even this amount Stokes-V observations with Cryo-NIRSP FOV coverage can remarkably improve the 3D coronal magnetic field reconstruction over an active region compared with tomography reconstruction based only on UCoMP linear polarization data.

Published
2022

35. The physiological interactome of TCR-like antibody therapeutics in human tissues

Author

Estelle Marrer-Berger, Annalisa Nicastri, Angelique Augustin, Vesna Kramar, Hanqing Liao, Lydia Jasmin Hanisch, Alejandro Carpy, Tina Weinzierl, Evelyne Durr, Nathalie Schaub, Ramona Nudischer, Daniela Ortiz-Franyuti, Ekaterina Breous-Nystrom, Janick Stucki, Nina Hobi, Giulia Raggi, Lauriane Cabon, Emmanuelle Lezan, Pablo Umaña, Isaac Woodhouse, Alexander Bujotzek, Christian Klein, and Nicola Ternette

Subjects
Science
Abstract

Abstract Selective binding of TCR-like antibodies that target a single tumour-specific peptide antigen presented by human leukocyte antigens (HLA) is the absolute prerequisite for their therapeutic suitability and patient safety. To date, selectivity assessment has been limited to peptide library screening and predictive modeling. We developed an experimental platform to de novo identify interactomes of TCR-like antibodies directly in human tissues using mass spectrometry. As proof of concept, we confirm the target epitope of a MAGE-A4-specific TCR-like antibody. We further determine cross-reactive peptide sequences for ESK1, a TCR-like antibody with known off-target activity, in human liver tissue. We confirm off-target-induced T cell activation and ESK1-mediated liver spheroid killing. Off-target sequences feature an amino acid motif that allows a structural groove-coordination mimicking that of the target peptide, therefore allowing the interaction with the engager molecule. We conclude that our strategy offers an accurate, scalable route for evaluating the non-clinical safety profile of TCR-like antibody therapeutics prior to first-in-human clinical application.

Published
2024
Full Text
View/download PDF

36. Current situation and transformation ways of housing policy in Ukraine

Author

V. Teremetskyi, O. Avramova, Kh. Maikut, O. Tserkovna, and R. Kramar

Subjects
public administration, normative and legal regulation of housing relations, housing restoration, housing, housing facilities stock, right to housing, fundamental human rights, Law
Abstract

Housing policy is one of the urgent scientific and practical areas within legal research because it has not been updated since 1995, and actual housing relations have already received a new meaning. The purpose of the study is to identify the features of the current situation and propose transformational approaches to housing policy. The study employs formal legal, statistical, and general scientific research methods, considering dialectical, synergistic, and axiological approaches. The main issue with Ukrainian housing policy is the absence of a modern concept guiding housing development and legal regulation. It is observed that the intensification of efforts to develop and adopt the principles of future state housing policy was conducted at the level of state authoritiesʼ activities. Based on the analysis of regulations in housing relations, it was determined that housing policy is part of a broader state policy – social policy. This is because it aims to meet the social interest – the need for housing. It is emphasised that housing policy is divided into national and regional housing policies. Particular attention is paid to the difficulty of restoring the destroyed housing facilities stock as a result of the war; it is suggested that a system of action be implemented to overcome the complicated situation in the housing sector. The study supports the idea of highlighting a specific area of modern housing policy in Ukraine, namely, ensuring the housing rights of vulnerable individuals, particularly those who have been forcibly evicted due to war. The study identifies specific areas of housing policy, including: the restoration of housing facilities (such as constructing new housing, compensating for destroyed housing, and conducting repair and construction works); devising an investment strategy for the housing sector; safeguarding citizensʼ housing rights; and revising housing legislation. These recommendations can be considered by the legislative and executive authorities when formulating a strategy and policy for the reconstruction of the housing stock of the state

Published
2024
Full Text
View/download PDF

37. Comparative Analysis of Electrokinetic Properties of Periodate- and TEMPO-Oxidized Regenerated Cellulose Fabric Functionalized with Chitosan

Author

Ana Kramar, Matea Korica, and Mirjana Kostić

Subjects
oxidation, cellulose, viscose, chitosan, zeta potential, periodate, Textile bleaching, dyeing, printing, etc., TP890-933
Abstract

The electrokinetic properties of materials give useful insight into the behavior of surfaces in contact with liquids and other compounds and their quantification is a powerful tool to predict their behavior during further processing and application, especially in textile materials. In this work, we perform a comparative analysis of influence of the two most common selective oxidative protocols for viscose (regenerated cellulose) fabrics on subsequent functionalization with chitosan, and cellulose fabrics’ electrokinetic properties, zeta potential in a pH range of approx. 3–10, and isoelectric point (IEP). For oxidation before deposition of chitosan, sodium periodate and 2,2,6,6-tetramethylpiperidine-1-oxy radical (TEMPO) were used. The content of functional groups in oxidized cellulose fabric (carboxyl and carbonyl groups) was determined by titration methods, while amino functional groups’ availability in samples with chitosan was determined using the CI acid orange 7 dye absorption method. This study reveals that the periodate oxidation (PO) of cellulose is more effective for binding chitosan onto material, which gave rise to higher availability of amino groups onto cellulose/chitosan material, which also influenced the shift in zeta potential curve towards positive values at a pH below 5. Analysis of a relationship between zeta potential increase at pH 4.4 and amino groups’ amount measured using absorption of CI acid orange 7 dye at pH 4.4 revealed dependency that can be fitted linearly or exponentially, with the latter providing the better fit (R2 = 0.75).

Published
2024
Full Text
View/download PDF

38. Well-posedness of non-autonomous transport equation on metric graphs

Author

Budde, Christian and Fijavž, Marjeta Kramar

Subjects
Mathematics - Analysis of PDEs, 37B55, 47D06, 35R02, 35F46
Abstract

We consider transport processes on metric graphs with time-dependent velocities and show that, under continuity assumption of the velocity coefficients, the corresponding non-autonomous abstract Cauchy problem is well-posed by means of evolution families and evolution semigroups., Comment: 11 pages

Published
2022

39. Contributors

Author

Abrar, Shazia, primary, Alves, Cátia, additional, Azeem, Musaddaq, additional, Barani, Hossein, additional, Basyigit, Zeynep Omerogullari, additional, Constantin, Cătălin, additional, Dinca, Laurențiu, additional, Fernandes, Marta, additional, Fernandes, Rui D.V., additional, Gong, R. Hugh, additional, Grujić, Dragana, additional, Haji, Aminoddin, additional, Hes, Lubos, additional, Iordache, Ovidiu, additional, Iqbal, Sarosh, additional, Jelil, R. Abd, additional, Kesici Güler, Hülya, additional, King, Martin W., additional, Kiran, Shumaila, additional, Kostić, Mirjana M., additional, Kramar, Ana D., additional, Kuraica, Milorad M., additional, Ladhari, N., additional, Mamun, Al, additional, Masin, Ivan, additional, Melro, Liliana, additional, Milanović, Jovana, additional, Mitu, Bogdana, additional, Morari, Cristian, additional, Munir, Bushra, additional, Nambi Gowri, Karuna, additional, Nourbakhsh, Shirin, additional, Obradović, Bratislav M., additional, Padrão, Jorge, additional, Petru, Michal, additional, Rasool, Atizaz, additional, Ribeiro, Ana I., additional, Rodrigues, Rui, additional, Rădulescu, Ion Răzvan, additional, Sabantina, Lilia, additional, Savić, Aleksandar, additional, Scarlat, Răzvan, additional, Sătulu, Veronica, additional, Surdu, Lilioara, additional, Talebian, Aazam, additional, Topalić-Trivunović, Ljiljana, additional, Visileanu, Emilia, additional, and Zille, Andrea, additional

Published
2024
Full Text
View/download PDF

41. Retrospective comparative analysis of two medical evacuation systems for Ukrainian patients affected by war

Author

Adyrov, Mykhaylo V, Alanbousi, Inna, Alexander, Sarah Weeks, Apel, Anna, Avula, Meghana, Bal, Wioletta Anna, Balwierz, Walentyna Aniela, Basset-Salom, Luisa, Bastardo Blanco, Daniel, Bauer, Karolina Jadwiga, Bayazitov, Ildar T, Berlanga, Pablo, Bhakta, Nickhill Hitesh, Bieniek, Katarzyna Anna, Bien, Ewa Iwona, Blackwood, Christopher Andrew, Blair, Sally Jane, Bodak, Khrystyna Ihorivna, Bordeianu, Irina, Bouffet, Eric Eric, Braganca, Joao Maria, Bucurenci, Mihaela Silvia, Budny, Elżbieta Beata, Budzyn, Andrii, Bumgardner, Christopher Carl, Burditt, Raina Nichole, Burnside Clapp, Victoria, Bykov, Viacheslav Valeriyovych, Cañete, Adela, Carnelli, Monica, Cela, Elena, Cepowska, Zuzanna, Chaber, Radoslaw, Cherner-Drieux, Anna, Chubata, Mariya, Clough, Heidi M, Czauderna, Piotr Stefan, Czernicka - Siwecka, Jolanta, Czyzewski, Krzysztof, Dalle, Jean-Hugues, Dashchakovska, Olha, de Koning, Linda A, Dembowska-Baginska, Bozenna Malgorzata, Derwich, Katarzyna, Dirksen, Uta, Dommett, Rachel, Dorosh, Olha Ihorivna, dos Reis Farinha, Nuno Jorge, Drabko, Katarzyna Anna, Dragomir, Monica Desiree, Dutkiewicz, Malgorzata, Dworzak, Michael, Dyma, Sergii Vitaliiovych, Earl, Julian, Eggert, Angelika, English, Martin William, Farren, Becky S, Fedyk, Nataliia Yuriina, Fernández-Teijeiro, Ana, Ferneza, Severyn Romanovych, Foster, Whitney Baer, Fox Irwin, Leeanna Elizabeth, Gałązkowski, Robert Maciej, Ganieva, Galyna, Garanzha, Vasylyna Andriivna, Gelman, Marina S, Godzinski, Jan Krzysztof, Goeres, Anne Françoise, Golban, Rodica, Graetz, Dylan Elizabeth, Greiner, Jeanette, Griksaitis, Michael J, Gupta, Sumit, Hampel, Michal Andrzej, Hastings, Sara Grace, Heenen, Delphine Liliane, Hill, Marcela C, Holiuk, Ihor, Holter, Wolfgang, Hough, Rachael Emma, Hutnik, Lukasz Marek, Irga-Jaworska, Ninela, Istomin, Oleksandr Andriyovych, Ignatova, Anna, Janczar, Szymon Lech, Kacharian, Arman, Kalwak, Krzysztof, Karolczyk, Grażyna Małgorzata, Karpenko, Nataliia Mikolaivna, Katsubo, Halyna Oleksandrivna, Kattamis, Antonis, Kazanowska, Bernarda Jadwiga, Kentsis, Alex, Ketteler, Petra, Kienesberger, Anita, Kiselev, Roman, Kizyma, Zoryana Petrivna, Kliuchkivska, Khrystyna, Klymniuk, Hryhorii Ivanovych, Kolenova, Alexandra, Kolodrubiec, Julia, Kostiuk, Yuliia, Kowalik, Tomasz, Kozlova, Olena Igorivna, Kozubenko, Vladyslav, Kraal, Kathelijne, Kramar, Tetyana Oleksandrivna, Krawczuk-Rybak, Maryna Maryna, Kulemzina, Irina, Kurkowska, Paulina, Kuzyk, Andriy S., Ladenstein, Ruth Lydia, Laguna, Pawel Jozef, Lassaletta, Alvaro, Lehmberg, Kai, Leontieva, Oksana, Liashenko, Serhii, Loizou, Loizos G., Lucchetta, Sonia Anna, Lupo, Matthew William, Lysytsia, Lesya, Lysytsia, Oleksandr, Machnik, Katarzyna Anna, Massimino, Maura, Mainland, Jeff A, Matczak, Katarzyna, Matysiak, Michal Jacek, Mayeur, Pierre, Miller, Beth Anne, Minervina, Anastasia A, Mishkova, Volha, Mizia-Malarz, Agnieszka Joanna, Morales La Madrid, Andres, Moreira, Daniel C, Moreno, Lucas, Moskvin, Vadim P, Mukkada, Sheena Teresa, Muszyńska-Rosłan, Katarzyna Maria, Mykychak, Iryna Volodymyrivna, Niemeyer, Charlotte, Nelson, Akoya Janae', Nogovitsyna, Yuliya, Ociepa, Tomasz, Oltolini, Stefano, Onipko, Nataliia, Pappas, Andrew, Patel, Amit B, Patrahau, Alina Alina, Pauley, Jennifer L., Pavlenko, Yehor Mikhailovich, Pavlovych, Andrij Oleksandrovych, Peregud-Pogorzelski, Jarosław Władyslaw, Perek-Polnik, Marta, Pérez, Vanesa, Pérez-Martínez, Antonio, Pikman, Yana, Pitozzi, Graziano Pitozzi, Portugal, Rui Gentil, Posternak, Victoria Vita, Pleshkan, Viktoriya, Prete, Arcangelo, Pritchard-Jones, Kathy, Raciborska, Anna, Radaelli, Alessandra, Reeves, Tegan Jemma, Reinhardt, Dirk, Reshetnyak, Andrey V, Rider, Andrew Jacob, Rizzari, Carmelo, Rizzi, Damiano Damiano, Rodriguez Hermosillo, Karen Gabriela, Ronenko, Olena Volodymyrivna, Rostkowska, Aneta Olga, Rudko, Liudmyla Yaroslavivna, Sakaan, Firas Mohamed, Sakhar, Nadezhda Aleksandrovna, Salman, Zeena S, Savva, Natallia N., Scaccaglia, Davide, Schaeffer, Elizabeth Hawthorne, Schneider, Carina Ursula, Scobie, Nicole, Semeniuk, Olena Volodymyrivna, Shevchyk, Roksoliana, Shuler, Ana I., Shvets, Stanislav, Sniderman, Liz, Skoczen, Szymon Pawel, Smeal, William John, Sokolowski, Igor, Sonkin, Anna Alexandra, Spolinyak, Andriy, Spota, Andrea, Sramkova, Lucie, Stepanjuk, Alla Ivanivna, Sterba, Jaroslav, Strahm, Brigitte, Styczynski, Jan, Svintsova, Olha, Synyuta, Andriy V, Szczepanski, Tomasz, Szczucinski, Pawel Kukiz, Szmyd, Bartosz Miroslaw, Tasso Cereceda, Maria, Teliuk, Alina, Tomanek, Iwona, Topping, Phoebe, Torrent, Montserrat, Trelińska, Joanna, Troyanovska, Olha Orestivna, Tsurkan, Lyudmila G., Tsymbalyuk-Voloshyn, Iryna, Tyupa, Sergiy Ihorovych, Urasinski, Tomasz Franciszek, Urbanek Dądela, Agnieszka, Vasilieva, Nataliia Jroslavivna, Vasilyeva, Aksana, Verdú-Amorós, Jaime, Vilcu-Bajurean, Natalia, Vinitsky, Leo, Vivtcharenko, Victoria, Vovk, Nelia, Volpe, Giovanni, Vorobel, Oksana Ivanivna, Wachowiak, Jacek Tadeusz, Wasiak, Marcin Slawomir, Wiedower, Lance Allan, Wobst, Natalia, Wuenschel, Lena Isolde, Wysocki, Mariusz Stanislaw, Yurieva, Marina, Zagurska, Anastasiia, Zakharenko, Stanislav S, Zakharenko, Aelita V, Zapotochna, Khrystyna, Zawitkowska, Joanna Emilia, Zecca, Marco, Mueller, Alexandra, Salek, Marta, Oszer, Aleksandra, Evseev, Dmitry, Yakimkova, Taisiya, Wlodarski, Marcin, Vinitsky, Anna, Kizyma, Roman, Pogorelyy, Mikhail, Zuber, Maria, Escalante, Juan, Lipska, Elzbieta, Fendler, Wojciech, Nowicka, Zuzanna, Szyszka, Adam, Rodriguez-Galindo, Carlos, Wise, Paul H., Agulnik, Asya, and Mlynarski, Wojciech

Published
2024
Full Text
View/download PDF

43. Predictive modeling of turning operations under different cooling/lubricating conditions for sustainable manufacturing with machine learning techniques

Author

Djordje Cica, Branislav Sredanovic, Sasa Tesic, and Davorin Kramar

Subjects
Machine learning, Sustainable machining, Machining force, Cutting power, Cutting pressure, Information technology, T58.5-58.64
Abstract

Sustainable manufacturing is one of the most important and most challenging issues in present industrial scenario. With the intention of diminish negative effects associated with cutting fluids, the machining industries are continuously developing technologies and systems for cooling/lubricating of the cutting zone while maintaining machining efficiency. In the present study, three regression based machine learning techniques, namely, polynomial regression (PR), support vector regression (SVR) and Gaussian process regression (GPR) were developed to predict machining force, cutting power and cutting pressure in the turning of AISI 1045. In the development of predictive models, machining parameters of cutting speed, depth of cut and feed rate were considered as control factors. Since cooling/lubricating techniques significantly affects the machining performance, prediction model development of quality characteristics was performed under minimum quantity lubrication (MQL) and high-pressure coolant (HPC) cutting conditions. The prediction accuracy of developed models was evaluated by statistical error analyzing methods. Results of regressions based machine learning techniques were also compared with probably one of the most frequently used machine learning method, namely artificial neural networks (ANN). Finally, a metaheuristic approach based on a neural network algorithm was utilized to perform an efficient multi-objective optimization of process parameters for both cutting environment.

Published
2024
Full Text
View/download PDF

46. Antioxidant properties of some plant extracts and effect of their addition on the oxidation stability of biodiesel

Author

Pavlo Kuzema, Iryna Laguta, Oksana Stavinskaya, Viktor Anishchenko, Anastasiia Kramar, Natalia Smirnova, Tetiana Fesenko, Roman Ivannikov, and Oksana Linnik

Subjects
plant extract, phenolic compound, antioxidant, biodiesel, oxidation stability, Chemistry, QD1-999, General. Including alchemy, QD1-65
Abstract

The extracts from the leaves of Deschampsia antarctica É. Desv., Camelina sativa (L.) Crantz, and Camellia japonica L. plants, as well as from defatted Camelina sativa and Silybum marianum seedcakes were investigated as potential additives for improvement of biodiesel stability against oxidation. Composition of the extracts was studied by means of HPLC, and antioxidant properties were evaluated using the Folin-Ciocalteu assay and the DPPH test. The oxidation of biodiesel was monitored during the accelerated procedure at 43C, with the changes in the acid number of biodiesel samples being the criteria of this process. In spite of significant distinctions in the content of various phenolic compounds, all the extracts were found to possess high antioxidant activity and decelerate biodiesel oxidation by 9-26%. The data did not reveal a directly proportional relationship between the antioxidants content in the extract, on the one hand, and the enhancement in biodiesel stability, on the other hand; various extracts had different influence on the behaviour of biodiesel from rape and Camelina seed oils. The results obtained are consistent with the assumption that there is no universal stabilizer for different types of biodiesel and indicate the prospects on searching for novel antioxidants of natural origin to inhibit oxidative processes.

Published
2023
Full Text
View/download PDF

47. On intermittency in sheared granular systems

Author

Kramar, Miroslav, Cheng, Chao, Basak, Rituparna, and Kondic, Lou

Subjects
Condensed Matter - Soft Condensed Matter
Abstract

We consider a system of granular particles, modeled by two dimensional frictional elastic disks, that is exposed to externally applied time-dependent shear stress in a planar Couette geometry. We concentrate on the external forcing that produces intermittent dynamics of stick-slip type. In this regime, the top wall remains almost at rest until the applied stress becomes sufficiently large, and then it slips. We focus on the evolution of the system as it approaches a slip event. Our main finding is that there are two distinct groups of measures describing system behavior before a slip event. The first group consists of global measures defined as system-wide averages at a fixed time. Typical examples of measures in this group are averages of the normal or tangent forces acting between the particles, system size and number of contacts between the particles. These measures do not seem to be sensitive to an approaching slip event. On average, they tend to increase linearly with the force pulling the spring. The second group consists of the time-dependent measures that quantify the evolution of the system on a micro (particle) or mesoscale. Measures in this group first quantify the temporal differences between two states and only then aggregate them to a single number. For example, Wasserstein distance quantitatively measures the changes of the force network as it evolves in time while the number of broken contacts quantifies the evolution of the contact network. The behavior of the measures in the second group changes dramatically before a slip event starts. They increase rapidly as a slip event approaches, indicating a significant increase in fluctuations of the system before a slip event is triggered.

Published
2021

48. Flows on Metric Graphs with General Boundary Conditions

Author

Engel, Klaus-Jochen and Fijavž, Marjeta Kramar

Subjects
Mathematics - Analysis of PDEs, Mathematics - Functional Analysis, 47D06, 34G10, 35L04, 35R02
Abstract

In this note we study the generation of $C_0$-semigroups by first order differential operators on $\mathrm{L}^p (\mathbb{R}_+,\mathbb{C}^{\ell})\times \mathrm{L}^p ([0,1],\mathbb{C}^{m})$ with general boundary conditions. In many cases we are able to characterize the generation property in terms of the invertibility of a matrix associated to the boundary conditions. The abstract results are used to study well-posedness of transport equations on non-compact metric graphs.

Published
2021

50. Evaluation of experimental error in accelerometer monitoring: Variation among individual animals versus variation among devices

Author

Colin Tobin, Derek Bailey, Caroline Wade, Ly Ly Trieu, Kelsey Nelson, Cory Oltjen, Huiping Cao, Tran Cao Son, Victor Flores, Briza Castro, Jennifer Hernandez Gifford, Mark Trotter, and David Kramar

Subjects
Accelerometer, Movement intensity, Movement variation, Random forest, Agriculture (General), S1-972, Agricultural industries, HD9000-9495
Abstract

Application of sensors is becoming prevalent in research and production settings. With increasing battery power, improved case and component durability, and consistent data connectivity, precision technologies, such as accelerometers, can help identify changes in livestock behavior. The objective of this study is to identify the variation among individual animals and among different accelerometer devices. A repeated 4 × 4 Latin-square design was utilized to identify differences between accelerometer, animal, and week. Twelve ewes separated into three age groups were randomly assigned to 4 different accelerometers deployed as an ear tag weekly over the course of 4 weeks. Manual behavior observations were paired to calculated accelerometer metrics and were used for training and validation dataset to predict animal behavior using random forest machine learning. Movement variation had the greatest importance in predicting behaviors. Across the four week study, differences were found for animal and week through each of the calculated metrics. Differences in accelerometers were detected for 80 % of the calculated metrics. This study shows the importance to account for variation among individual animals and accelerometer devices in experimental designs.

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results