@inproceedings{18121,
  abstract     = {It is known that sparsity can improve interpretability for deep neural networks. However, existing methods in the area either require networks that are pre-trained with sparsity constraints, or impose sparsity after the fact, altering the network’s general behavior. In this paper, we demonstrate, for the first time, that sparsity can instead be incorporated into the interpretation process itself, as a sample-specific preprocessing step. Unlike previous work, this approach, which we call SPADE, does not place constraints on the trained model and does not affect its behavior during inference on the sample. Given a trained model and a target sample, SPADE uses sample-targeted pruning to provide a "trace" of the network’s execution on the sample, reducing the network to the most important connections prior to computing an interpretation. We demonstrate that preprocessing with SPADE significantly increases the accuracy of image saliency maps across several interpretability methods. Additionally, SPADE improves the usefulness of neuron visualizations, aiding humans in reasoning about network behavior. Our code is available at https://github.com/IST-DASLab/SPADE.},
  author       = {Moakhar, Arshia Soltani and Iofinova, Eugenia B and Frantar, Elias and Alistarh, Dan-Adrian},
  booktitle    = {Proceedings of the 41st International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vienna, Austria},
  pages        = {45955--45987},
  publisher    = {ML Research Press},
  title        = {{SPADE: Sparsity-guided debugging for deep neural networks}},
  volume       = {235},
  year         = {2024},
}

@inproceedings{18976,
  abstract     = {We analyze asynchronous-type algorithms for distributed SGD in the heterogeneous setting, where each worker has its own computation and communication speeds, as well as data distribution. In these algorithms, workers compute possibly stale and stochastic gradients associated with their local data at some iteration back in history and then return those gradients to the server without synchronizing with other workers. We present a unified convergence theory for non-convex smooth functions in the heterogeneous regime. The proposed analysis provides convergence for pure asynchronous SGD and its various modifications. Moreover, our theory explains what affects the convergence rate and what can be done to improve the performance of asynchronous algorithms. In particular, we introduce a novel asynchronous method based on worker shuffling. As a by-product of our analysis, we also demonstrate convergence guarantees for gradient-type algorithms such as SGD with random reshuffling and shuffle-once mini-batch SGD. The derived rates match the best-known results for those algorithms, highlighting the tightness of our approach. Finally, our numerical evaluations support theoretical findings and show the good practical performance of our method.},
  author       = {Islamov, Rustem and Safaryan, Mher and Alistarh, Dan-Adrian},
  booktitle    = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics},
  issn         = {2640-3498},
  location     = {Valencia, Spain},
  pages        = {649--657},
  publisher    = {ML Research Press},
  title        = {{AsGrad: A sharp unified analysis of asynchronous-SGD algorithms}},
  volume       = {238},
  year         = {2024},
}

@inproceedings{18113,
  abstract     = {The emergence of accurate open large language models (LLMs) has led to a race towards performant quantization techniques which can enable their execution on end-user devices. In this paper, we revisit the problem of “extreme” LLM compression—defined as targeting extremely low bit counts, such as 2 to 3 bits per parameter—from the point of view of classic methods in Multi-Codebook Quantization (MCQ). Our algorithm, called AQLM, generalizes the classic Additive Quantization (AQ) approach for information retrieval to advance the state-of-the-art in LLM compression, via two innovations: 1) learned additive quantization of weight matrices in input-adaptive fashion, and 2) joint optimization of codebook parameters across each transformer blocks. Broadly, AQLM is the first scheme that is Pareto optimal in terms of accuracy-vs-model-size when compressing to less than 3 bits per parameter, and significantly improves upon all known schemes in the extreme compression (2bit) regime. In addition, AQLM is practical: we provide fast GPU and CPU implementations of AQLM for token generation, which enable us to match or outperform optimized FP16 implementations for speed, while executing in a much smaller memory footprint.},
  author       = {Egiazarian, Vage and Panferov, Andrei and Kuznedelev, Denis and Frantar, Elias and Babenko, Artem and Alistarh, Dan-Adrian},
  booktitle    = {Proceedings of the 41st International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vienna, Austria},
  pages        = {12284--12303},
  publisher    = {ML Research Press},
  title        = {{Extreme compression of large language models via additive quantization}},
  volume       = {235},
  year         = {2024},
}

@inproceedings{18975,
  abstract     = {Leveraging second-order information about the loss at the scale of deep networks is one of the main lines of approach for improving the performance of current optimizers for deep learning. Yet, existing approaches for accurate full-matrix preconditioning, such as Full-Matrix Adagrad (GGT) or Matrix-Free Approximate Curvature (M-FAC) suffer from massive storage costs when applied even to small-scale models, as they must store a sliding window of gradients, whose memory requirements are multiplicative in the model dimension. In this paper, we address this issue via a novel and efficient error-feedback technique that can be applied to compress preconditioners by up to two orders of magnitude in practice, without loss of convergence. Specifically, our approach compresses the gradient information via sparsification or low-rank compression before it is fed into the preconditioner, feeding the compression error back into future iterations. Extensive experiments on deep neural networks show that this approach can compress full-matrix preconditioners to up to 99% sparsity without accuracy loss, effectively removing the memory overhead of fullmatrix preconditioners such as GGT and M-FAC.},
  author       = {Modoranu, Ionut-Vlad and Kalinov, Aleksei and Kurtic, Eldar and Frantar, Elias and Alistarh, Dan-Adrian},
  booktitle    = {41st International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vienna, Austria},
  pages        = {35910--35933},
  publisher    = {ML Research Press},
  title        = {{Error feedback can accurately compress preconditioners}},
  volume       = {235},
  year         = {2024},
}

@inproceedings{18977,
  abstract     = {Recent advances in large language model (LLM) pretraining have led to high-quality LLMs with impressive abilities. By compressing such LLMs via quantization to 3-4 bits per parameter, they can fit into memory-limited devices such as laptops and mobile phones, enabling personalized use. Quantizing models to 3-4 bits per parameter can lead to moderate to high accuracy losses, especially for smaller models (1-10B parameters), which are suitable for edge deployment. To address this accuracy issue, we introduce the Sparse-Quantized Representation (SpQR), a new compressed format and quantization technique that enables for the first time \emph{near-lossless} compression of LLMs across model scales while reaching similar compression levels to previous methods. SpQR works by identifying and isolating \emph{outlier weights}, which cause particularly large quantization errors, and storing them in higher precision while compressing all other weights to 3-4 bits, and achieves relative accuracy losses of less than 
 in perplexity for highly-accurate LLaMA and Falcon LLMs. This makes it possible to run a 33B parameter LLM on a single 24 GB consumer GPU without performance degradation at 15% speedup, thus making powerful LLMs available to consumers without any downsides. SpQR comes with efficient algorithms for both encoding weights into its format, as well as decoding them efficiently at runtime. Specifically, we provide an efficient GPU inference algorithm for SpQR, which yields faster inference than 16-bit baselines at similar accuracy while enabling memory compression gains of more than 4x.},
  author       = {Dettmers, Tim and Svirschevski, Ruslan A. and Egiazarian, Vage and Kuznedelev, Denis and Frantar, Elias and Ashkboos, Saleh and Borzunov, Alexander and Hoefler, Torsten and Alistarh, Dan-Adrian},
  booktitle    = {12th International Conference on Learning Representations},
  location     = {Vienna, Austria},
  publisher    = {OpenReview},
  title        = {{SpQR: A sparse-quantized representation for near-lossless LLM weight compression}},
  year         = {2024},
}

@inproceedings{17093,
  abstract     = {Federated Learning (FL) enables large-scale distributed training of machine learning models, while still allowing individual nodes to maintain data locally. However, executing FL at scale comes with inherent practical challenges: 1) heterogeneity of the local node data distributions, 2) heterogeneity of node computational speeds (asynchrony), but also 3) constraints in the amount of communication between the clients and the server. In this work, we present the first variant of the classic federated averaging (FedAvg) algorithm which, at the same time, supports data heterogeneity, partial client asynchrony, and communication compression. Our algorithm comes with a novel, rigorous analysis showing that, in spite of these system relaxations, it can provide similar convergence to FedAvg in interesting parameter regimes. Experimental results in the rigorous LEAF benchmark on setups of up to 300 nodes show that our algorithm ensures fast convergence for standard federated tasks, improving upon prior quantized and asynchronous approaches.},
  author       = {Zakerinia, Hossein and Talaei, Shayan and Nadiradze, Giorgi and Alistarh, Dan-Adrian},
  booktitle    = {Proceedings of the 27th International Conference on Artificial Intelligence and Statistics},
  issn         = {2640-3498},
  location     = {Valencia, Spain},
  pages        = {3448--3456},
  publisher    = {ML Research Press},
  title        = {{Communication-efficient federated learning with data and client heterogeneity}},
  volume       = {238},
  year         = {2024},
}

@inproceedings{18117,
  abstract     = {We investigate parameter-efficient fine-tuning (PEFT) methods that can provide good accuracy under limited computational and memory budgets in the context of large language models (LLMs). We present a new PEFT method called Robust Adaptation (RoSA) inspired by robust principal component analysis that jointly trains low-rank
 and highly-sparse components on top of a set of fixed pretrained weights to efficiently approximate the performance of a full-fine-tuning (FFT) solution. Across a series of challenging generative tasks such as grade-school math and SQL query generation, which require fine-tuning for good performance, we show that RoSA outperforms LoRA, pure sparse fine-tuning, and alternative hybrid methods at the same parameter budget, and can even recover the performance of FFT on some tasks. We provide system support for RoSA to complement the training algorithm, specifically in the form of sparse GPU kernels which enable memory- and computationally-efficient training, and show that it is also compatible with low-precision base weights, resulting in the first joint representation combining quantization, low-rank and sparse approximations. Our code is available at https://github.com/IST-DASLab/RoSA.},
  author       = {Nikdan, Mahdi and Tabesh, Soroush and Crncevic, Elvir and Alistarh, Dan-Adrian},
  booktitle    = {Proceedings of the 41st International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vienna, Austria},
  pages        = {38187--38206},
  publisher    = {ML Research Press},
  title        = {{RoSA: Accurate parameter-efficient fine-tuning via robust adaptation}},
  volume       = {235},
  year         = {2024},
}

@article{15315,
  abstract     = {Single and collective cell migration are fundamental processes critical for physiological phenomena ranging from embryonic development and immune response to wound healing and cancer metastasis. To understand cell migration from a physical perspective, a broad variety of models for the underlying physical mechanisms that govern cell motility have been developed. A key challenge in the development of such models is how to connect them to experimental observations, which often exhibit complex stochastic behaviours. In this review, we discuss recent advances in data-driven theoretical approaches that directly connect with experimental data to infer dynamical models of stochastic cell migration. Leveraging advances in nanofabrication, image analysis, and tracking technology, experimental studies now provide unprecedented large datasets on cellular dynamics. In parallel, theoretical efforts have been directed towards integrating such datasets into physical models from the single cell to the tissue scale with the aim of conceptualising the emergent behaviour of cells. We first review how this inference problem has been addressed in both freely migrating and confined cells. Next, we discuss why these dynamics typically take the form of underdamped stochastic equations of motion, and how such equations can be inferred from data. We then review applications of data-driven inference and machine learning approaches to heterogeneity in cell behaviour, subcellular degrees of freedom, and to the collective dynamics of multicellular systems. Across these applications, we emphasise how data-driven methods can be integrated with physical active matter models of migrating cells, and help reveal how underlying molecular mechanisms control cell behaviour. Together, these data-driven approaches are a promising avenue for building physical models of cell migration directly from experimental data, and for providing conceptual links between different length-scales of description.},
  author       = {Brückner, David and Broedersz, Chase P.},
  issn         = {1361-6633},
  journal      = {Reports on Progress in Physics},
  number       = {5},
  publisher    = {IOP Publishing},
  title        = {{Learning dynamical models of single and collective cell migration: a review}},
  doi          = {10.1088/1361-6633/ad36d2},
  volume       = {87},
  year         = {2024},
}

@article{15250,
  abstract     = {Orphan solute carrier (SLC) represents a group of membrane transporters whose exact functions and substrate specificities are not known. Elucidating the function and regulation of orphan SLC transporters is not only crucial for advancing our knowledge of cellular and molecular biology but can potentially lead to the development of new therapeutic strategies. Here, we provide evidence for the biological function of a ubiquitous orphan lysosomal SLC, the Major Facilitator Superfamily Domain-containing Protein 1 (MFSD1), which has remained phylogenetically unassigned. Targeted metabolomics revealed that dipeptides containing either lysine or arginine residues accumulate in lysosomes of cells lacking MFSD1. Whole-cell patch-clamp electrophysiological recordings of HEK293-cells expressing MFSD1 on the cell surface displayed transport affinities for positively charged dipeptides in the lower mM range, while dipeptides that carry a negative net charge were not transported. This was also true for single amino acids and tripeptides, which MFSD1 failed to transport. Our results identify MFSD1 as a highly selective lysosomal lysine/arginine/histidine-containing dipeptide exporter, which functions as a uniporter.},
  author       = {Boytsov, Danila and Madej, Gregor M. and Horn, Georg and Blaha, Nadine and Köcher, Thomas and Sitte, Harald H. and Siekhaus, Daria E and Ziegler, Christine and Sandtner, Walter and Roblek, Marko},
  issn         = {1091-6490},
  journal      = {Proceedings of the National Academy of Sciences of the United States of America},
  number       = {13},
  publisher    = {National Academy of Sciences},
  title        = {{Orphan lysosomal solute carrier MFSD1 facilitates highly selective dipeptide transport}},
  doi          = {10.1073/pnas.2319686121},
  volume       = {121},
  year         = {2024},
}

@article{15024,
  abstract     = {Electrostatic correlations between ions dissolved in water are known to impact their transport properties in numerous ways, from conductivity to ion selectivity. The effects of these correlations on the solvent itself remain, however, much less clear. In particular, the addition of salt has been consistently reported to affect the solution’s viscosity, but most modeling attempts fail to reproduce experimental data even at moderate salt concentrations. Here, we use an approach based on stochastic density functional theory, which accurately captures charge fluctuations and correlations. We derive a simple analytical expression for the viscosity correction in concentrated electrolytes, by directly linking it to the liquid’s structure factor. Our prediction compares quantitatively to experimental data at all temperatures and all salt concentrations up to the saturation limit. This universal link between the microscopic structure and viscosity allows us to shed light on the nanoscale dynamics of water and ions under highly concentrated and correlated conditions.},
  author       = {Robin, Paul},
  issn         = {1089-7690},
  journal      = {Journal of Chemical Physics},
  number       = {6},
  publisher    = {AIP Publishing},
  title        = {{Correlation-induced viscous dissipation in concentrated electrolytes}},
  doi          = {10.1063/5.0188215},
  volume       = {160},
  year         = {2024},
}

@article{17892,
  abstract     = {Enzyme-substrate kinetics form the basis of many biomolecular processes. The interplay between substrate binding and substrate geometry can give rise to long-range interactions between enzyme binding events. Here we study a general model of enzyme-substrate kinetics with restricted long-range interactions described by an exponent −𝛾. We employ a coherent-state path integral and renormalization group approach to calculate the first moment and two-point correlation function of the enzyme-binding profile. We show that starting from an empty substrate the average occupancy follows a power law with an exponent 1/(1−𝛾) over time. The correlation function decays algebraically with two distinct spatial regimes characterized by exponents −𝛾 on short distances and −(2/3)⁢(2−𝛾) on long distances. The crossover between both regimes scales inversely with the average substrate occupancy. Our work allows associating experimental measurements of bound enzyme locations with their binding kinetics and the spatial conformation of the substrate.},
  author       = {Olmeda, Fabrizio and Rulands, Steffen},
  issn         = {2470-0053},
  journal      = {Physical Review E},
  number       = {2},
  publisher    = {American Physical Society},
  title        = {{Field theory of enzyme-substrate systems with restricted long-range interactions}},
  doi          = {10.1103/PhysRevE.110.024404},
  volume       = {110},
  year         = {2024},
}

@article{17104,
  abstract     = {The homeostasis of epithelial tissue relies on a balance between the self-renewal of stem cell populations, cellular differentiation, and loss. Although this balance needs to be tightly regulated to avoid pathologies, such as tumor growth, the regulatory mechanisms, both cell-intrinsic and collective, which ensure tissue steady-state are still poorly understood. Here, we develop a computational model that incorporates basic assumptions of stem cell renewal into distinct populations and mechanical interactions between cells. We find that the model generates unexpected dynamic features: stem cells repel each other in the bulk tissue and are thus found rather isolated, as in a number of in vivo contexts. By mapping the system onto a gas of passive Brownian particles with effective repulsive interactions, that arise from the generated flows of differentiated cells, we show that we can quantitatively describe such stem cell distribution in tissues. The interaction potential between a pair of stem cells decays exponentially with a characteristic length that spans several cell sizes, corresponding to the volume of cells generated per stem cell division. Our findings may help understanding the dynamics of normal and cancerous epithelial tissues.},
  author       = {Krämer, Johannes C. and Hannezo, Edouard B and Gompper, Gerhard and Elgeti, Jens},
  issn         = {2542-4653},
  journal      = {SciPost Physics},
  number       = {4},
  publisher    = {SciPost Foundation},
  title        = {{Mechanically-driven stem cell separation in tissues caused by proliferating daughter cells}},
  doi          = {10.21468/scipostphys.16.4.097},
  volume       = {16},
  year         = {2024},
}

@article{18919,
  abstract     = {The integration of theory and experiment makes possible tracking the slow evolution of a photodoped Mott insulator to a distinct non-equilibrium metallic phase under the influence of electron-lattice coupling.},
  author       = {Baykusheva, Denitsa Rangelova},
  issn         = {1745-2481},
  journal      = {Nature Physics},
  number       = {5},
  pages        = {684--685},
  publisher    = {Springer Nature},
  title        = {{Through the slopes of a light-induced phase transition}},
  doi          = {10.1038/s41567-024-02401-7},
  volume       = {20},
  year         = {2024},
}

@misc{23022,
  abstract     = {Contains animations, data, and source code used for the paper "Mechanically-driven Stem Cell Separation in Tissues caused by Proliferating Daughter Cells" by the same authors. 


This repository contains all the data and analysis scripts to support the manuscript
The file structure is as follows:
- "analysis" contains all the analysis scripts
- "gfx" and "animations" the figures and supporting animations
- "1-SCpair", "2-SCtissue", "3-BDsimulation", "4-tissue-stochdiv", "6-inertia-effects" all the source code and data for the respective models.},
  author       = {Elgeti, Jens and Krämer, Johannes},
  publisher    = {Zenodo},
  title        = {{Data for Mechanically-driven Stem Cell Separation in Tissues caused by Proliferating Daughter Cells}},
  doi          = {10.5281/zenodo.8410956},
  year         = {2024},
}

@article{17123,
  abstract     = {A key feature of many developmental systems is their ability to self-organize spatial patterns of functionally distinct cell fates. To ensure proper biological function, such patterns must be established reproducibly, by controlling and even harnessing intrinsic and extrinsic fluctuations. While the relevant molecular processes are increasingly well understood, we lack a principled framework to quantify the performance of such stochastic self-organizing systems. To that end, we introduce an information-theoretic measure for self-organized fate specification during embryonic development. We show that the proposed measure assesses the total information content of fate patterns and decomposes it into interpretable contributions corresponding to the positional and correlational information. By optimizing the proposed measure, our framework provides a normative theory for developmental circuits, which we demonstrate on lateral inhibition, cell type proportioning, and reaction–diffusion models of self-organization. This paves a way toward a classification of developmental systems based on a common information-theoretic language, thereby organizing the zoo of implicated chemical and mechanical signaling processes.},
  author       = {Brückner, David and Tkačik, Gašper},
  issn         = {1091-6490},
  journal      = {Proceedings of the National Academy of Sciences of the United States of America},
  number       = {23},
  publisher    = {National Academy of Sciences},
  title        = {{Information content and optimization of self-organized developmental systems}},
  doi          = {10.1073/pnas.2322326121},
  volume       = {121},
  year         = {2024},
}

@article{17885,
  abstract     = {The formation of new ribosomes is tightly coordinated with cell growth and proliferation. In eukaryotes, the correct assembly of all ribosomal proteins and RNAs follows an intricate scheme of maturation and rearrangement steps across three cellular compartments: the nucleolus, nucleoplasm, and cytoplasm. We demonstrate that usnic acid, a lichen secondary metabolite, inhibits the maturation of the large ribosomal subunit in yeast. We combine biochemical characterization of pre-ribosomal particles with a quantitative single-particle cryo-EM approach to monitor changes in nucleolar particle populations upon drug treatment. Usnic acid rapidly blocks the transition from nucleolar state B to C of Nsa1-associated pre-ribosomes, depleting key maturation factors such as Dbp10 and hindering pre-rRNA processing. This primary nucleolar block rapidly rebounds on earlier stages of the pathway which highlights the regulatory linkages between different steps. In summary, we provide an in-depth characterization of the effect of usnic acid on ribosome biogenesis, which may have implications for its reported anti-cancer activities.},
  author       = {Kofler, Lisa and Grundmann, Lorenz and Gerhalter, Magdalena and Prattes, Michael and Merl-Pham, Juliane and Zisser, Gertrude and Grishkovskaya, Irina and Hodirnau, Victor-Valentin and Vareka, Martin and Breinbauer, Rolf and Hauck, Stefanie M. and Haselbach, David and Bergler, Helmut},
  issn         = {2041-1723},
  journal      = {Nature Communications},
  publisher    = {Springer Nature},
  title        = {{The novel ribosome biogenesis inhibitor usnic acid blocks nucleolar pre-60S maturation}},
  doi          = {10.1038/s41467-024-51754-3},
  volume       = {15},
  year         = {2024},
}

@misc{23023,
  abstract     = {In this repository, you will find the R-scripts used to generate the figures, along with the data necessary to make the plots. You will also find the Python-scripts used to compute the pair-wise distance between embryos, to estimate the surface, the volume, the α-parameter, to identify the closest rigid packing and to generate the morphomap. Additionally, you have access to the base model used with Plantseg to predict and label the cells at the 8-cell stage. Note that a more accurate and more specific model was generated for each embryo based on the curation of at least 5 timepoints.

Figures
The plots in Figure 1-7 and Figure S1-S7 can be generated with the respective files. The code is written in R and requires RStudio to work out-of-the-box. The R scripts are located in figures/. Figures and supplemental figures are generated with the same file (e.g., Figure 1.R generates all the computed panels of Figure 1 and Figure S1). Data required to generate the plots are located in figures/data/. Note that the code to work with the morphomap and the tracking data is located in local packages figures/fabreges.morphomap and figures/fabreges.tracking. An additional package figures/fabreges.generic contains utility functions.

Python packages
For the needs of this study, we developed a series of package for Python located in python-packages/. The packages can be directly imported within your scripts. Along with the packages, you will find 6 scripts used to compute, extract or estimate the data from the source. The folder is organised as follow:

package interfaces is located in python-packages/interfaces/ and can be tested by running the scripts to extract the angles an estimate the surface of contact (compute-angles.py and estimate-contacts.py respectively) in folder python-packages,
package morphomap is located in python-packages/morphomap/ and can be tested by running the script to compute the geometrical distances between embryos, project the data in 2D and compute the volume of the cells (compute-distances.py, process-distances.py and compute-volumes.py respectively) in folder python-packages,
package splinefit is included in morphomap in python-packages/morphomap/splinefit/, but can function independently,
package rbb is located in python-packages/rbb/ and can be tested by running the script to identify the closest rigid packing of an embryo at the 8-cell stage (identify-packing.py) in folder python-packages.
We provide example of segmented data in python-packages/samples/.

CNN
We provide a convolutional neural network (CNN) used in our study, located in cnn. Please note that we optimised the CNN further for each dataset by doing manual curation of at least 5 time points and by retraining the network with those ground truth. The CNN were trained using pytorch-3dunet. The CNN can be used with Plantseg. Currently (plantseg v.1.6.0 and pytorch-3dunet v.1.6.0), the output model from pytorch-3dunet must currently be modified to work with plantseg:

# python
import torch
for model in ('best_checkpoint.pytorch', 'last_checkpoint.pytorch'):
    state = torch.load(model, map_location='cpu')
    state = state['model_state_dict']
    torch.save(state, model)

Requierements
Figures
The figures were generated with R (v. 4.4.0) and RStudio (v. 2024.04.1+748). The following packages must be installed prior to execution:

alphashape3d (v. 1.3.2)
devtools (v. 2.4.5)
ggplot2 (v. 3.5.1)
ggrepel (v. 0.9.5)
rstudioapi (v. 0.16.0)
To install all these packages, run the following command in the RStudio console:

# R
install.packages(c("alphashape3d@1.3.2", "devtools@2.4.5", "ggplot2@3.5.1", "ggrepel@0.9.5", "rstudioapi@0.16"))

More recent version of R, RStudio and the packages may also work (not tested).

Python packages
Python 3.12.3 was used to analyse the data. The scripts may work with more recent version of Python and with Python 3.8 or earlier (not tested). To run the code, you must first install the following packages:

numpy (v. 1.26.4)
imageio (v. 2.34.1)
scipy (v. 1.13.1)
scikit-image (v. 0.23.2)
scikit-learn (v. 1.5.0)
Alternatively, you can create a conda environment:

# Shell script
conda create -y --name fabreges2024 python=3.12.3
conda activate fabreges2024
python -m pip install numpy==1.26.4 imageio==2.34.1 scipy==1.13.1 scikit-image==0.23.2 scikit-learn==1.5.0

More recent version of the dependencies may work (not tested).

Usage
Figures
Open one of the figure script in RStudio and run the entire code. You may have to create a folder output in figures/ if it does not exist.

The code has local dependencies that will be installed on your computer when encountering them for the first time. They start with fabreges.* and are necessary to generate the figures.

Python packages (example)
Compute angles
# Shell script
# Compute a summary (mean, median, s.d.) of the angles in embryo C1 at time 40
python compute-angles.py -s -r 15 -R 20 samples/C1_t040.tif.gz output/C1_t040_angles.csv
# Compute a detailed list of angle values
python compute-angles.py -r 15 -R 20 samples/C1_t041.tif.gz output/C1_t041_angles.csv

Compute distances
# Shell script
# Compute the distance between two embryos
python compute-distances.py -c 5 -s 5 samples/C1_t040.tif.gz samples/C1_t041.tif.gz output/distances.csv
# Append another distance measurement
python compute-distances.py -a -c 5 -s 5 samples/C1_t040.tif.gz samples/C1_t042.tif.gz output/distances.csv

Compute volumes
# Shell script
# Compute the volume in isotropic data
python compute-volumes.py -r 0.416 0.416 0.416 samples/C1_t040.tif.gz output/volumes.csv

Estimate contacts
# Shell script
# Estimate the contacts and the free surface area
python estimate-contacts.py -r 0.416 0.416 0.416 samples/C1_t040.tif.gz output/contacts.csv

Identify the closest rigid packing
# Shell script
# Assuming the file `output/contacts.csv` exists and contains the output
# from command `estimate-contacts.py`.
# Identify the closest rigid packings
python identify-packing.py output/contacts.csv output/backbones.csv

Generate the seeds for Surface Evolver
# Shell script
# Assuming the file `output/contacts.csv` exists and contains the output
# from command `estimate-contacts.py`, and the file `output/volumes.csv`
# exists and contains the output from command `compute-volumes.py`
# Generate the seeds for Surface Evolver using automatic thresholding
# Automatic threshold will determine the highest cell-cell contact threshold
# that preserves rigidity of the embryo (or 0 if not rigid)
python generate-seed.py output/contacts.csv output/volumes.csv output/seed.fe
# Generate the seeds for Surface Evolver with custom threshold
python generate-seed.py --threshold 0.1 output/contacts.csv output/volumes.csv output/seed.fe

Project in 2D and generate a distance matrix
# Shell script
# Assuming the folder `foo` only contains one or more CSV
# from command `compute-distances.py`.

# Compute the tSNE 2D projection
python process-distances.py -D2 -t output/tsne_2d.csv foo/

# Compute the tSNE 3D projection
python process-distances.py -D3 -t output/tsne_2d.csv foo/

# Generate the distance matrix
python process-distances.py -m output/dismat.csv foo/},
  author       = {Corominas-Murtra, Bernat and   Moghe, Prachiti and Kickuth, Alison and Ichikawa, Takafumi and Iwatani, Chizuru and Tsukiyama, Tomoyuki and Daniel, Nathalie and Gering, Julie and Stokkermans, Anniek and Wolny, Adrian and Kreshuk, Anna and Duranthon, Véronique and Uhlmann, Virginie and Hannezo, Edouard and Hiiragi, Takashi},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Data and software from: Temporal variability and cell mechanics control robustness in mammalian embryogenesis}},
  doi          = {10.5281/zenodo.12819780},
  year         = {2024},
}

@inproceedings{18913,
  abstract     = {With the proliferation of blockchain technology in high-value sectors, consensus protocols are becoming critical infrastructures. The rapid innovation cycle in Byzantine fault tolerant (BFT) consensus protocols has culminated in HotStuff, which provides linear message complexity in the partially synchronous setting. To achieve this, HotStuff leverages a leader that collects, aggregates, and broadcasts the messages of other validators. This paper analyzes the security implications of such approaches in practice, from the perspective of liveness and availability.
By implementing attacks in a globally-distributed testbed, we show that state-of-the-art leader-based protocols are vulnerable to denial-of-service (DoS) attacks on the leader. Our attacks, demonstrated on committees of up to 64 validators, manage to disrupt liveness within seconds, using only a few tens of Mbps of attack bandwidth per validator. Crucially, the cost and effectiveness of the attacks are independent of the committee size. Based on the outcome of these experiments, we then propose and test effective mitigations. Our findings show that advancements in both protocol design and network-layer defenses can greatly improve the practical resilience of BFT consensus protocols.},
  author       = {Giuliari, Giacomo and Sonnino, Alberto and Frei, Marc and Streun, Fabio and Kokoris Kogias, Eleftherios and Perrig, Adrian},
  booktitle    = {Proceedings of the 19th ACM Asia Conference on Computer and Communications Security},
  isbn         = {9798400704826},
  location     = {Singapore, Singapore},
  pages        = {1345--1360},
  publisher    = {ACM},
  title        = {{An empirical study of consensus protocols’ DoS resilience}},
  doi          = {10.1145/3634737.3656997},
  year         = {2024},
}

@inproceedings{18957,
  abstract     = {Sui Lutris is the first smart-contract platform to sustainably achieve sub-second finality. It achieves this significant decrease by employing consensusless agreement not only for simple payments but for a large variety of transactions. Unlike prior work, Sui Lutris neither compromises expressiveness nor throughput and can run perpetually without restarts. Sui Lutris achieves this by safely integrating consensuless agreement with a high-throughput consensus protocol that is invoked out of the critical finality path but ensures that when a transaction is at risk of inconsistent concurrent accesses, its settlement is delayed until the total ordering is resolved. Building such a hybrid architecture is especially delicate during reconfiguration events, where the system needs to preserve the safety of the consensusless path without compromising the long-term liveness of potentially misconfigured clients. We thus develop a novel reconfiguration protocol, the first to provably show the safe and efficient reconfiguration of a consensusless blockchain. Sui Lutris is currently running in production and underpins the Sui smart-contract platform. Combined with the use of Objects instead of accounts it enables the safe execution of smart contracts that expose objects as a first-class resource. In our experiments Sui Lutris achieves latency lower than 0.5 seconds for throughput up to 5,000 certificates per second (150k ops/s with transaction blocks), compared to the state-of-the-art real-world consensus latencies of 3 seconds. Furthermore, it gracefully handles validators crash-recovery and does not suffer visible performance degradation during reconfiguration.},
  author       = {Blackshear, Sam and Chursin, Andrey and Danezis, George and Kichidis, Anastasios and Kokoris Kogias, Eleftherios and Li, Xun and Logan, Mark and Menon, Ashok and Nowacki, Todd and Sonnino, Alberto and Williams, Brandon and Zhang, Lu},
  booktitle    = {Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security},
  isbn         = {9798400706363},
  location     = {Salt Lake City, UT, United States},
  pages        = {2606--2620},
  publisher    = {ACM},
  title        = {{Sui Lutris: A blockchain combining broadcast and consensus}},
  doi          = {10.1145/3658644.3670286},
  year         = {2024},
}

@article{17269,
  abstract     = {The directed migration of epithelial cell collectives through coordinated movements plays a crucial role in various physiological processes and is increasingly understood at the level of large confluent monolayers. However, numerous processes rely on the migration of small groups of polarized epithelial clusters in complex environments, and their responses to external geometries remain poorly understood. To address this, we cultivate primary epithelial keratocyte tissues on adhesive microstripes to create autonomous epithelial clusters with well-defined geometries. We show that their migration efficiency is strongly influenced by the contact geometry and the orientation of cell–cell contacts with respect to the direction of migration. A combination of velocity and polarity alignment with contact regulation of locomotion in an active matter model captures quantitatively the experimental data. Furthermore, we predict that this combination of rules enables efficient navigation in complex geometries, which we confirm experimentally. Altogether, our findings provide a conceptual framework for extracting the interaction rules of active systems from their interaction with physical boundaries, as well as design principles for collective navigation in complex microenvironments.},
  author       = {Vercruysse, Eléonore and Brückner, David and Gómez-González, Manuel and Remson, Alexandre and Luciano, Marine and Kalukula, Yohalie and Rossetti, Leone and Trepat, Xavier and Hannezo, Edouard B and Gabriele, Sylvain},
  issn         = {1745-2481},
  journal      = {Nature Physics},
  pages        = {1492--1500},
  publisher    = {Springer Nature},
  title        = {{Geometry-driven migration efficiency of autonomous epithelial cell clusters}},
  doi          = {10.1038/s41567-024-02532-x},
  volume       = {20},
  year         = {2024},
}