Search

A key process during animal morphogenesis is oriented tissue deformation, which is often driven by internally generated active stresses. Yet, such active oriented materials are prone to well-known instabilities, raising the question of how oriented tissue deformation can be robust during morphogenesis. In a simple scenario, we recently showed that active oriented deformation can be stabilized by the boundary-imposed gradient of a scalar field, which represents, e.g., a morphogen gradient in a developing embryo. Here, we discuss a more realistic scenario, where the morphogen is produced by a localized source region, diffuses across the tissue, and degrades. Consistent with our earlier results, we find that oriented tissue deformation is stable in the gradient-extensile case, i.e. when active stresses act to extend the tissue along the direction of the gradient, but it is unstable in the gradient-contractile case. In addition, we now show that gradient-contractile tissues can not be stabilized even by morphogen diffusion. Finally, we point out the existence of an additional instability, which results from the interplay of tissue shear and morphogen diffusion. Our theoretical results explain the lack of gradient-contractile tissues in the biological literature, suggesting that the active matter instability acts as an evolutionary selection criterion., Comment: 19 pages, 9 figures

Conway's Game of Life (GOL) is an epitome showing how complex dynamical behavior emerges from simple local interactions. Although it has often been found that GOL dynamics lies close to critical behavior, this system has never been studied in the context of a deterministic phase transitions and cluster dynamics. In this work, we study the deterministic critical behavior that emerges in the \textit{logistic} GOL: an extension of Conway's GOL with a parameter that alters the dynamics by expanding the binary state space into a Cantor set, but while maintaining the deterministic nature of the system. Upon tuning the parameter, we find that the logistic GOL comprises at least three types of asymptotic behavior, i.e phases, that are separated by two critical points. One critical point defines the boundary between a sparse-static and a sparse-dynamic asymptotic phase, whereas the other point marks a deterministic percolation transition between the sparse-dynamic and a third, dense-dynamic asymptotic phase. Moreover, we identify distinct power-law distributions of cluster sizes near the critical points, and discuss the underlying mechanisms that give rise to such critical behavior. Overall, our work highlights that scale invariance can emerge even when clusters in a system are generated by a purely deterministic process., Comment: 20 pages,16 figures

Vision-Language Models (VLMs) have shown remarkable capabilities in a large number of downstream tasks. Nonetheless, compositional image understanding remains a rather difficult task due to the object bias present in training data. In this work, we investigate the reasons for such a lack of capability by performing an extensive bench-marking of compositional understanding in VLMs. We compare contrastive models with generative ones and analyze their differences in architecture, pre-training data, and training tasks and losses. Furthermore, we leverage In-Context Learning (ICL) as a way to improve the ability of VLMs to perform more complex reasoning and understanding given an image. Our extensive experiments demonstrate that our proposed approach outperforms baseline models across multiple compositional understanding datasets.

Intrusion Detection Systems (IDSs) have played a significant role in the detection and prevention of cyber-attacks in traditional computing systems. It is not surprising that this technology is now being applied to secure Internet of Things (IoT) networks against cyber threats. However, the limited computational resources available on IoT devices pose a challenge for deploying conventional computing-based IDSs. IDSs designed for IoT environments must demonstrate high classification performance, and utilize low-complexity models. Developing intrusion detection models in the field of IoT has seen significant advancements. However, achieving a balance between high classification performance and reduced complexity remains a challenging endeavor. In this research, we present an effective IDS model that addresses this issue by combining a lightweight Convolutional Neural Network (CNN) with bidirectional Long Short-Term Memory (BiLSTM). Additionally, we employ feature selection techniques to minimize the number of features inputted into the model, thereby reducing its complexity. This approach renders the proposed model highly suitable for resource-constrained IoT devices, ensuring it meets their computation capability requirements. Creating a model that meets the demands of IoT devices and attains enhanced precision is a challenging task. However, our suggested model outperforms previous works in the literature by attaining a remarkable accuracy rate of 97.90% within a prediction time of 1.1 seconds for binary classification. Furthermore, it achieves an accuracy rate of 97.09% within a prediction time of 2.10 seconds for multiclassification., Comment: 2024 11th International Conference on Wireless Networks and Mobile Communications (WINCOM)

Unmanned Aerial Vehicles (UAVs) have become increasingly popular in various applications, especially with the emergence of 6G systems and networks. However, their widespread adoption has also led to concerns regarding security vulnerabilities, making the development of reliable intrusion detection systems (IDS) essential for ensuring UAVs safety and mission success. This paper presents a new IDS for UAV networks. A binary-tuple representation was used for encoding class labels, along with a deep learning-based approach employed for classification. The proposed system enhances the intrusion detection by capturing complex class relationships and temporal network patterns. Moreover, a cross-correlation study between common features of different UAVs was conducted to discard correlated features that might mislead the classification of the proposed IDS. The full study was carried out using the UAV-IDS-2020 dataset, and we assessed the performance of the proposed IDS using different evaluation metrics. The experimental results highlighted the effectiveness of the proposed multiclass classifier model with an accuracy of 95%.

Text-to-video retrieval systems have recently made significant progress by utilizing pre-trained models trained on large-scale image-text pairs. However, most of the latest methods primarily focus on the video modality while disregarding the audio signal for this task. Nevertheless, a recent advancement by ECLIPSE has improved long-range text-to-video retrieval by developing an audiovisual video representation. Nonetheless, the objective of the text-to-video retrieval task is to capture the complementary audio and video information that is pertinent to the text query rather than simply achieving better audio and video alignment. To address this issue, we introduce TEFAL, a TExt-conditioned Feature ALignment method that produces both audio and video representations conditioned on the text query. Instead of using only an audiovisual attention block, which could suppress the audio information relevant to the text query, our approach employs two independent cross-modal attention blocks that enable the text to attend to the audio and video representations separately. Our proposed method's efficacy is demonstrated on four benchmark datasets that include audio: MSR-VTT, LSMDC, VATEX, and Charades, and achieves better than state-of-the-art performance consistently across the four datasets. This is attributed to the additional text-query-conditioned audio representation and the complementary information it adds to the text-query-conditioned video representation., Comment: Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023

HollandPTC is an independent outpatient center for proton therapy, scientific research, and education. Patients with different types of cancer are treated with Intensity Modulated Proton Therapy (IMPT). In addition, the HollandPTC R&D consortium conducts scientific research into the added value and improvements of proton therapy. To this end, HollandPTC created clinical and pre-clinical research facilities including a versatile R&D proton beamline for various types of biologically and technologically oriented research. In this work, we present the characterization of the R&D proton beam line of HollandPTC. Its pencil beam mimics the one used for clinical IMPT, with energy ranging from 70 up to 240 MeV, which has been characterized in terms of shape, size, and intensity. For experiments that need a uniform field in depth and lateral directions, a dual ring passive scattering system has been designed, built, and characterized. With this system, field sizes between 2x2 cm2 and 20x20 cm2 with 98% uniformity are produced with dose rates ranging from 0.5 Gy/min up to 9 Gy/min. The unique and customized support of the dual ring system allows switching between a pencil beam and a large field in a very simple and fast way, making the HollandPTC R&D proton beam able to support a wide range of different applications.

Thompson sampling (TS) is a popular heuristic for action selection, but it requires sampling from a posterior distribution. Unfortunately, this can become computationally intractable in complex environments, such as those modeled using neural networks. Approximate posterior samples can produce effective actions, but only if they reasonably approximate joint predictive distributions of outputs across inputs. Notably, accuracy of marginal predictive distributions does not suffice. Epistemic neural networks (ENNs) are designed to produce accurate joint predictive distributions. We compare a range of ENNs through computational experiments that assess their performance in approximating TS across bandit and reinforcement learning environments. The results indicate that ENNs serve this purpose well and illustrate how the quality of joint predictive distributions drives performance. Further, we demonstrate that the \textit{epinet} -- a small additive network that estimates uncertainty -- matches the performance of large ensembles at orders of magnitude lower computational cost. This enables effective application of TS with computation that scales gracefully to complex environments.

A resource-constrained unmanned aerial vehicle (UAV) can be used as a flying LoRa gateway (GW) to move inside the target area for efficient data collection and LoRa resource management. In this work, we propose deep reinforcement learning (DRL) to optimize the energy efficiency (EE) in wireless LoRa networks composed of LoRa end devices (EDs) and a flying GW to extend the network lifetime. The trained DRL agent can efficiently allocate the spreading factors (SFs) and transmission powers (TPs) to EDs while considering the air-to-ground wireless link and the availability of SFs. In addition, we allow the flying GW to adjust its optimal policy onboard and perform online resource allocation. This is accomplished through retraining the DRL agent using reduced action space. Simulation results demonstrate that our proposed DRL-based online resource allocation scheme can achieve higher EE in LoRa networks over three benchmark schemes.

Transcranial photobiomodulation is an optical method for non-invasive brain stimulation. The method projects red and near-infrared light through the scalp within 600-1100 nm and low energy within the 1-20 J/cm2 range. Recent studies have been optimistic about replacing this method with pharmacotherapy and invasive brain stimulation. However, concerns and ambiguities exist regarding the light penetration depth and possible thermal side effects. While the literature survey indicates that the skin temperature rises after experimental optical brain stimulation, inadequate evidence supports a safe increase in temperature or the amount of light penetration in the cortex. Therefore, we aimed to conduct a comprehensive study on the heat transfer of near-infrared stimulation for the human brain. Our research considers the transcranial photobiomodulation over the human brain model by projecting 810 nm light with 100 mW/cm2 power density to evaluate its thermal and optical effects using bioheat transfer and radiative transfer equation. Our results confirm that the near-infrared light spectrum has a small incremental impact on temperature and approximately penetrates 1 cm, reaching the cortex. A time-variant study of the heat transfer was also carried out to measure the temperature changes during optical stimulation., Comment: The complete geometry proposed in this work is available for download. The proposed geometry is in STL and MPHBIN formats, and the model is surrounded by air. All the tissues are available and assembled, and it is recommended to use the transparency tool to acquire a better observation. Please cite this publication when referencing this material

Here we perform angle and time-resolved photoelectron spectroscopy on the commensurate Charge Density Wave (CDW) phase of 1T-TaS$_2$. Data with different probe pulse polarization are employed to map the dispersion of electronic states below and above the chemical potential. Upon photoexcitation, the fluctuations of CDW order erase the band dispersion near to the chemical potential and halve the charge gap size. This transient phase sets within half a period of the coherent lattice motion and is favored by strong electronic correlations. The experimental results are compared to Density-Functional Theory (DFT) calculations with a self-consistent evaluation of the Coulomb repulsion. Our simulations indicate that the screening of Coulomb repulsion depends on the stacking order of the TaS$_2$ layers. The entanglement of such degrees of freedom suggest that both the structural order and electronic repulsion are locally modified by the photoinduced CDW fluctuations.

Active matter with local polar or nematic order is subject to the well-known Simha-Ramaswamy instability. It is so far unclear how, despite this instability, biological tissues can undergo robust active anisotropic deformation during animal morphogenesis. Here we show that protein concentration gradients (e.g. morphogen gradients), which are known to control large-scale coordination among cells, can stabilize such deformations. To this end, we study a hydrodynamic model of an active polar material. To account for the effect of the protein gradient, the polar field is coupled to the boundary-provided gradient of a scalar field that also advects with material flows. Focusing on the large system size limit, we show in particular: (i) The system can be stable for an effectively extensile coupling between scalar field gradient and active stresses, i.e. gradient-extensile coupling, while it is always unstable for a gradient-contractile coupling. Intriguingly, there are many systems in the biological literature that are gradient-extensile, while we could not find any that are clearly gradient-contractile. (ii) Stability is strongly affected by the way polarity magnitude is controlled. Taken together, our findings, if experimentally confirmed, suggest new developmental principles that are directly rooted in active matter physics., Comment: 20 pages, 7 figures

Time-resolved ARPES gives access to the band structure and ultrafast dynamics of excited electronic states in solids. The orbital character of the bands close to the Fermi level is essential to understand the origin of several exotic phenomena in quantum materials. By performing polarization dependent time- and angle-resolved photoemission spectroscopy and by analyzing the chirality of the photoelectron yield for two different crystal orientations, we identify the orbital character of bands below and above the chemical potential for the Dirac semimetal BaNiS2. Our results illustrate how the control and understanding of matrix elements effects in time-resolved photoemission spectroscopy can be a powerful tool for the study of quantum materials., Comment: 10 pages, 6 figures

Light rare-earth antimonide LaSb2 is a material susceptible to nesting and exhibits a Charge Density Wave (CDW) at 355 K as well as superconductivity below 1.2 K. In the seek for additional CDW transitions, we have studied the temperature-dependent of LaSb2 at 200 and 13 K and infer CDW manifestations by combining angle-resolved photoemission spectroscopy (ARPES) measurements and density functional theory (DFT) calculations. ARPES measurements at 200 K show a metallic system while it appears to be semiconducting at 13 K, at odds with existing resistivity measurements. At 13 K, ARPES shows the band folding of the inner Fermi surface pockets, with considerable spectral weight on the folded band. We find a nesting vector at q = 0.25$\pm$0.02 {\AA} -1. In addition, we observe Umklapps of other bands due to the onset of the new periodicity, together with a semiconducting behavior in the whole reciprocal space. Calculations demonstrate that the folded band is associated with the La-Sb layer and that in-plane distortion is the most probable structural modification in the system, probably affecting the whole unit cell., Comment: In this work, the discussion presents two interpretations to explain the experimental results. We withdraw this work because after further discussions and data analysis we have concluded that only one of the two interpretations is consistent with the experimental results

This work introduces a dataset for large-scale instance-level recognition in the domain of artworks. The proposed benchmark exhibits a number of different challenges such as large inter-class similarity, long tail distribution, and many classes. We rely on the open access collection of The Met museum to form a large training set of about 224k classes, where each class corresponds to a museum exhibit with photos taken under studio conditions. Testing is primarily performed on photos taken by museum guests depicting exhibits, which introduces a distribution shift between training and testing. Testing is additionally performed on a set of images not related to Met exhibits making the task resemble an out-of-distribution detection problem. The proposed benchmark follows the paradigm of other recent datasets for instance-level recognition on different domains to encourage research on domain independent approaches. A number of suitable approaches are evaluated to offer a testbed for future comparisons. Self-supervised and supervised contrastive learning are effectively combined to train the backbone which is used for non-parametric classification that is shown as a promising direction. Dataset webpage: http://cmp.felk.cvut.cz/met/

Contrastive losses have long been a key ingredient of deep metric learning and are now becoming more popular due to the success of self-supervised learning. Recent research has shown the benefit of decomposing such losses into two sub-losses which act in a complementary way when learning the representation network: a positive term and an entropy term. Although the overall loss is thus defined as a combination of two terms, the balance of these two terms is often hidden behind implementation details and is largely ignored and sub-optimal in practice. In this work, we approach the balance of contrastive losses as a hyper-parameter optimization problem, and propose a coordinate descent-based search method that efficiently find the hyper-parameters that optimize evaluation performance. In the process, we extend existing balance analyses to the contrastive margin loss, include batch size in the balance, and explain how to aggregate loss elements from the batch to maintain near-optimal performance over a larger range of batch sizes. Extensive experiments with benchmarks from deep metric learning and self-supervised learning show that optimal hyper-parameters are found faster with our method than with other common search methods., Comment: 15 pages, 10 figures

Learning with noisy labels is an active research area for image classification. However, the effect of noisy labels on image retrieval has been less studied. In this work, we propose a noise-resistant method for image retrieval named Teacher-based Selection of Interactions, T-SINT, which identifies noisy interactions, ie. elements in the distance matrix, and selects correct positive and negative interactions to be considered in the retrieval loss by using a teacher-based training setup which contributes to the stability. As a result, it consistently outperforms state-of-the-art methods on high noise rates across benchmark datasets with synthetic noise and more realistic noise., Comment: Accepted at WACV 2022. 13 pages, 5 figures

Visual Place Recognition (VPR) is generally concerned with localizing outdoor images. However, localizing indoor scenes that contain part of an outdoor scene can be of large value for a wide range of applications. In this paper, we introduce Inside Out Visual Place Recognition (IOVPR), a task aiming to localize images based on outdoor scenes visible through windows. For this task we present the new large-scale dataset Amsterdam-XXXL, with images taken in Amsterdam, that consists of 6.4 million panoramic street-view images and 1000 user-generated indoor queries. Additionally, we introduce a new training protocol Inside Out Data Augmentation to adapt Visual Place Recognition methods for localizing indoor images, demonstrating the potential of Inside Out Visual Place Recognition. We empirically show the benefits of our proposed data augmentation scheme on a smaller scale, whilst demonstrating the difficulty of this large-scale dataset for existing methods. With this new task we aim to encourage development of methods for IOVPR. The dataset and code are available for research purposes at https://github.com/saibr/IOVPR, Comment: Accepted at British Machine Vision Conference (BMVC) 2021

Predictive distributions quantify uncertainties ignored by point estimates. This paper introduces The Neural Testbed: an open-source benchmark for controlled and principled evaluation of agents that generate such predictions. Crucially, the testbed assesses agents not only on the quality of their marginal predictions per input, but also on their joint predictions across many inputs. We evaluate a range of agents using a simple neural network data generating process. Our results indicate that some popular Bayesian deep learning agents do not fare well with joint predictions, even when they can produce accurate marginal predictions. We also show that the quality of joint predictions drives performance in downstream decision tasks. We find these results are robust across choice a wide range of generative models, and highlight the practical importance of joint predictions to the community.

A fundamental challenge for any intelligent system is prediction: given some inputs, can you predict corresponding outcomes? Most work on supervised learning has focused on producing accurate marginal predictions for each input. However, we show that for a broad class of decision problems, accurate joint predictions are required to deliver good performance. In particular, we establish several results pertaining to combinatorial decision problems, sequential predictions, and multi-armed bandits to elucidate the essential role of joint predictive distributions. Our treatment of multi-armed bandits introduces an approximate Thompson sampling algorithm and analytic techniques that lead to a new kind of regret bound.

Intelligence relies on an agent's knowledge of what it does not know. This capability can be assessed based on the quality of joint predictions of labels across multiple inputs. In principle, ensemble-based approaches produce effective joint predictions, but the computational costs of training large ensembles can become prohibitive. We introduce the epinet: an architecture that can supplement any conventional neural network, including large pretrained models, and can be trained with modest incremental computation to estimate uncertainty. With an epinet, conventional neural networks outperform very large ensembles, consisting of hundreds or more particles, with orders of magnitude less computation. The epinet does not fit the traditional framework of Bayesian neural networks. To accommodate development of approaches beyond BNNs, such as the epinet, we introduce the epistemic neural network (ENN) as an interface for models that produce joint predictions.

Reinforcement learning agents have demonstrated remarkable achievements in simulated environments. Data efficiency poses an impediment to carrying this success over to real environments. The design of data-efficient agents calls for a deeper understanding of information acquisition and representation. We discuss concepts and regret analysis that together offer principled guidance. This line of thinking sheds light on questions of what information to seek, how to seek that information, and what information to retain. To illustrate concepts, we design simple agents that build on them and present computational results that highlight data efficiency.

We redefine the transition function of elementary cellular automata (ECA) in terms of discrete operators. The operator representation provides a clear hint about the way systems behave both at the local and the global scale. We show that mirror and complementary symmetric rules are connected to each other via simple operator transformations. It is possible to decouple the representation into two pairs of operators which are used to construct a periodic table of ECA that maps all unique rules in such a way that rules having similar behavior are clustered together. Finally, the operator representation is used to implement a generalized logistic extension to ECA. Here a single tuning parameter scales the pace with which operators iterate the rules. We show that, as this parameter is tuned, many rules of ECA undergo multiple phase transitions between periodic, locally chaotic, chaotic and complex (Class 4) behavior.

We study the use of hypermodels to represent epistemic uncertainty and guide exploration. This generalizes and extends the use of ensembles to approximate Thompson sampling. The computational cost of training an ensemble grows with its size, and as such, prior work has typically been limited to ensembles with tens of elements. We show that alternative hypermodels can enjoy dramatic efficiency gains, enabling behavior that would otherwise require hundreds or thousands of elements, and even succeed in situations where ensemble methods fail to learn regardless of size. This allows more accurate approximation of Thompson sampling as well as use of more sophisticated exploration schemes. In particular, we consider an approximate form of information-directed sampling and demonstrate performance gains relative to Thompson sampling. As alternatives to ensembles, we consider linear and neural network hypermodels, also known as hypernetworks. We prove that, with neural network base models, a linear hypermodel can represent essentially any distribution over functions, and as such, hypernetworks are no more expressive., Comment: Published as a conference paper at ICLR 2020

Artificial, CNN-generated images are now of such high quality that humans have trouble distinguishing them from real images. Several algorithmic detection methods have been proposed, but these appear to generalize poorly to data from unknown sources, making them infeasible for real-world scenarios. In this work, we present a framework for evaluating detection methods under real-world conditions, consisting of cross-model, cross-data, and post-processing evaluation, and we evaluate state-of-the-art detection methods using the proposed framework. Furthermore, we examine the usefulness of commonly used image pre-processing methods. Lastly, we evaluate human performance on detecting CNN-generated images, along with factors that influence this performance, by conducting an online survey. Our results suggest that CNN-based detection methods are not yet robust enough to be used in real-world scenarios., Comment: Accepted to the workshop on Media Forensics at CVPR 2020

In recent years, Generative Adversarial Networks (GANs) have improved steadily towards generating increasingly impressive real-world images. It is useful to steer the image generation process for purposes such as content creation. This can be done by conditioning the model on additional information. However, when conditioning on additional information, there still exists a large set of images that agree with a particular conditioning. This makes it unlikely that the generated image is exactly as envisioned by a user, which is problematic for practical content creation scenarios such as generating facial composites or stock photos. To solve this problem, we propose a single pipeline for text-to-image generation and manipulation. In the first part of our pipeline we introduce textStyleGAN, a model that is conditioned on text. In the second part of our pipeline we make use of the pre-trained weights of textStyleGAN to perform semantic facial image manipulation. The approach works by finding semantic directions in latent space. We show that this method can be used to manipulate facial images for a wide range of attributes. Finally, we introduce the CelebTD-HQ dataset, an extension to CelebA-HQ, consisting of faces and corresponding textual descriptions., Comment: Accepted to the AI for content creation workshop at CVPR 2020