Search

In recent years, interplanetary exploration has gained significant momentum, leading to a focus on the development of launch vehicles. However, the critical technology of edl mechanisms has not received the same level of attention and remains less mature and capable. To address this gap, we took advantage of the REXUS program to develop a pioneering edl mechanism. We propose an alternative to conventional, parachute based landing vehicles by utilizing autorotation. Our approach enables future additions such as steerability, controllability, and the possibility of a soft landing. To validate the technique and our specific implementation, we conducted a sounding rocket experiment on REXUS29. The systems design is outlined with relevant design decisions and constraints, covering software, mechanics, electronics and control systems. Furthermore, an emphasis will also be the organization and setup of the team entirely made up and executed by students. The flight results on REXUS itself are presented, including the most important outcomes and possible reasons for mission failure. We have not archived an autorotation based landing, but provide a reliable way of building and operating such vehicles. Ultimately, future works and possibilities for improvements are outlined. The research presented in this paper highlights the need for continued exploration and development of edl mechanisms for future interplanetary missions. By discussing our results, we hope to inspire further research in this area and contribute to the advancement of space exploration technology., Comment: 12 pages, 9 figures

Multimodal interleaved datasets featuring free-form interleaved sequences of images and text are crucial for training frontier large multimodal models (LMMs). Despite the rapid progression of open-source LMMs, there remains a pronounced scarcity of large-scale, diverse open-source multimodal interleaved datasets. In response, we introduce MINT-1T, the most extensive and diverse open-source Multimodal INTerleaved dataset to date. MINT-1T comprises one trillion text tokens and 3.4 billion images, a 10x scale-up from existing open-source datasets. Additionally, we include previously untapped sources such as PDFs and ArXiv papers. As scaling multimodal interleaved datasets requires substantial engineering effort, sharing the data curation process and releasing the dataset greatly benefits the community. Our experiments show that LMMs trained on MINT-1T rival the performance of models trained on the previous leading dataset, OBELICS. Our data and code will be released at https://github.com/mlfoundations/MINT-1T.

Tabular data -- structured, heterogeneous, spreadsheet-style data with rows and columns -- is widely used in practice across many domains. However, while recent foundation models have reduced the need for developing task-specific datasets and predictors in domains such as language modeling and computer vision, this transfer learning paradigm has not had similar impact in the tabular domain. In this work, we seek to narrow this gap and present TabuLa-8B, a language model for tabular prediction. We define a process for extracting a large, high-quality training dataset from the TabLib corpus, proposing methods for tabular data filtering and quality control. Using the resulting dataset, which comprises over 1.6B rows from 3.1M unique tables, we fine-tune a Llama 3-8B large language model (LLM) for tabular data prediction (classification and binned regression) using a novel packing and attention scheme for tabular prediction. Through evaluation across a test suite of 329 datasets, we find that TabuLa-8B has zero-shot accuracy on unseen tables that is over 15 percentage points (pp) higher than random guessing, a feat that is not possible with existing state-of-the-art tabular prediction models (e.g. XGBoost, TabPFN). In the few-shot setting (1-32 shots), without any fine-tuning on the target datasets, TabuLa-8B is 5-15 pp more accurate than XGBoost and TabPFN models that are explicitly trained on equal, or even up to 16x more data. We release our model, code, and data along with the publication of this paper.

We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% & 66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation., Comment: Project page: https://www.datacomp.ai/dclm/

There are some global tests for heterogeneity of variance in k-sample one-way layouts, but few consider pairwise comparisons between treatment levels. For experimental designs with a control, comparisons of the variances between the treatment levels and the control are of interest - in analogy to the location parameter with the Dunnett (1955) procedure. Such a many-to-one approach for variances is proposed using the Levene transformation, a kind of residuals. Its properties are characterized with simulation studies and corresponding data examples are evaluated with R code.

Earthquake nowcasting has been proposed as a means of tracking the change in large earthquake potential in a seismically active area. The method was developed using observable seismic data, in which probabilities of future large earthquakes can be computed using Receiver Operating Characteristic (ROC) methods. Furthermore, analysis of the Shannon information content of the earthquake catalogs has been used to show that there is information contained in the catalogs, and that it can vary in time. Here we discuss a new method for earthquake nowcasting that uses an AI-enhanced deep learning model "QuakeGPT" that is based on an attention-based science transformer adapted for time series forecasting. Such dot product attention-based transformers were introduced by Vaswani et al. (2017), and are the basis for the new large language models such as ChatGPT. To use these science transformers, they must first be trained on a large corpus of data. A problem is that the existing history of reliable earthquake catalog data extends back in time only a few decades, which is almost certainly too short to train a model for reliable nowcasting/forecasting. As a result, we turn to earthquake simulations to train the transformer model. Specifically we discuss a simple stochastic earthquake simulation model "ERAS" that has recently been introduced. The ERAS model is similar to the more common "ETAS" models, the difference being that the ERAS model has only 2 novel, adjustable parameters, rather than the 6-8 adjustable parameters that characterize most ETAS models. Using this ERAS model, we then define a transformer model and train it using a long catalog of ERAS simulations, then apply it to an ERAS validation dataset with the transformer model. In this paper, we describe this new method and assess the applicability to observed earthquake catalogs for use in nowcasting/forecasting., Comment: 25 Pages, 10 Figures

Using the transfer matrix method, we numerically investigate the structure of spatial coherences and their fluctuations in the 3d Anderson model in the metallic phase when driven out-of-equilibrium by external leads at zero temperature and in linear response. We find that the stationary state entails non-local non-Gaussian correlations in the longitudinal direction, which are characteristic of diffusive non equilibrium steady states. These correlations are quantitatively matched, at least up to third order, by those analytically derived in the Quantum Symmetric Simple Exclusion Process (QSSEP) which describes diffusive fermions in 1d subject to dynamical disorder. Furthermore, the large deviation scaling and $U(1)$ invariance of these correlations imply a link between the Anderson model and free probability theory. Our findings suggest the existence of a universal structure of correlations in non-interacting diffusive quantum systems that might be captured by QSSEP., Comment: 5 pages main text, 7 pages supplemental material

Millimeter wave (mmWave) cell-free massive MIMO (CF mMIMO) is a promising solution for future wireless communications. However, its optimization is non-trivial due to the challenging channel characteristics. We show that mmWave CF mMIMO optimization is largely an assignment problem between access points (APs) and users due to the high path loss of mmWave channels, the limited output power of the amplifier, and the almost orthogonal channels between users given a large number of AP antennas. The combinatorial nature of the assignment problem, the requirement for scalability, and the distributed implementation of CF mMIMO make this problem difficult. In this work, we propose an unsupervised machine learning (ML) enabled solution. In particular, a graph neural network (GNN) customized for scalability and distributed implementation is introduced. Moreover, the customized GNN architecture is hierarchically permutation-equivariant (HPE), i.e., if the APs or users of an AP are permuted, the output assignment is automatically permuted in the same way. To address the combinatorial problem, we relax it to a continuous problem, and introduce an information entropy-inspired penalty term. The training objective is then formulated using the augmented Lagrangian method (ALM). The test results show that the realized sum-rate outperforms that of the generalized serial dictatorship (GSD) algorithm and is very close to the upper bound in a small network scenario, while the upper bound is impossible to obtain in a large network scenario.

PLATO (PLAnetary Transits and Oscillations of stars) is ESA's M3 mission designed to detect and characterise extrasolar planets and perform asteroseismic monitoring of a large number of stars. PLATO will detect small planets (down to <2 R_(Earth)) around bright stars (<11 mag), including terrestrial planets in the habitable zone of solar-like stars. With the complement of radial velocity observations from the ground, planets will be characterised for their radius, mass, and age with high accuracy (5 %, 10 %, 10 % for an Earth-Sun combination respectively). PLATO will provide us with a large-scale catalogue of well-characterised small planets up to intermediate orbital periods, relevant for a meaningful comparison to planet formation theories and to better understand planet evolution. It will make possible comparative exoplanetology to place our Solar System planets in a broader context. In parallel, PLATO will study (host) stars using asteroseismology, allowing us to determine the stellar properties with high accuracy, substantially enhancing our knowledge of stellar structure and evolution. The payload instrument consists of 26 cameras with 12cm aperture each. For at least four years, the mission will perform high-precision photometric measurements. Here we review the science objectives, present PLATO's target samples and fields, provide an overview of expected core science performance as well as a description of the instrument and the mission profile at the beginning of the serial production of the flight cameras. PLATO is scheduled for a launch date end 2026. This overview therefore provides a summary of the mission to the community in preparation of the upcoming operational phases.

We measure the noise correlations of two-dimensional (2D) Bose gases after free expansion, and use them to characterize the in-situ phase coherence across the Berezinskii-Kosterlitz-Thouless (BKT) transition. The noise-correlation function features a characteristic spatial oscillatory behavior in the superfluid phase, which gives direct access to the superfluid exponent. This oscillatory behavior vanishes above the BKT critical point, as we demonstrate for both single-layer and decoupled bilayer 2D Bose gases. Our work establishes noise interferometry as an important general tool to probe and identify many-body states of quantum gases, extending its application to previously inaccessible correlation properties in multimode systems.

The classical-dS scenario in the type II string theories proposes to search for dS vacua of orientifold flux compactifications in a regime where string corrections to the compactified effective field theory are negligible. We study a minimal extension of this scenario in which the leading string corrections to the O-plane/D-brane actions at the 4-derivative order are included but higher orders as well as string corrections in the bulk are self-consistently neglected. Our proposal is motivated by a recent debate about dS solutions with O8-planes which circumvent a classical no-go theorem due to unusual sources leading to so-called permissive boundary conditions for the 10D supergravity fields. We argue that such sources do not arise in classical supergravity but ask whether including the 4-derivative corrections leads to sources that have a similar effect. However, we find that the 4-derivative corrections do not allow meta-stable dS in a class of models with O8-planes and/or D8-branes we consider. We also study related models which in addition contain O6-planes/D6-branes and find that again no meta-stable dS is allowed, both classically and including the 4-derivative corrections. While some of the arguments in this work require the backreaction of the O-plane/D-brane sources to be small, others are valid including the full backreaction., Comment: 59 pages, 6 figures. v2: minor changes, JHEP version

In the past decade, detection of THz radiation by plasma-wave-assisted frequency mixing in antenna-coupled field-effect transistors (TeraFETs) -- implemented in various semiconductor material systems (Si CMOS, GaN/AlGaN, GaAs/AlGaAs, graphene, etc.) -- has matured and led to a practically applied detector technology. This has been supported by the development of powerful device simulation tools which take into account relevant collective carrier dynamics and mixing processes in various approximations. These tools mostly model carrier transport in 1D and they are usually geared towards continuous-wave illumination of the device and small-signal response. Depending on their implementation, it may not be possible readily to simulate large-signal and pulsed operation. Another approximation which may lead to unsatisfactory results is the 1D restriction to calculate only the longitudinal electric field components. Especially at the edges of the gate electrode, solving of the 2D Poisson equation promises better results. This contribution introduces a stable way to solve the 2D Poisson equation self-consistently with the hydrodynamic transport equations including the numerically challenging convection term. We employ a well-balanced approximate Harten-Lax-van-Leer-Contact Riemann solver. The approach is well suited for a future treatment of transient and large-signal cases as well as the Dyakonov-Shur instability. The 2D treatment also generically extends the model beyond the gradual-channel approximation and allows to calculate the FET's response at high THz frequencies where the gate-to-channel potential acquires a non-local character. Model calculations are performed for the exemplary case of a 65-nm Si CMOS TeraFET in the isothermal approximation., Comment: revised version, 13 pages, 3 figures

Image classification is one of the most fundamental capabilities of machine vision intelligence. In this work, we revisit the image classification task using visually-grounded language models (VLMs) such as GPT-4V and LLaVA. We find that existing proprietary and public VLMs, despite often using CLIP as a vision encoder and having many more parameters, significantly underperform CLIP on standard image classification benchmarks like ImageNet. To understand the reason, we explore several hypotheses concerning the inference algorithms, training objectives, and data processing in VLMs. Our analysis reveals that the primary cause is data-related: critical information for image classification is encoded in the VLM's latent space but can only be effectively decoded with enough training data. Specifically, there is a strong correlation between the frequency of class exposure during VLM training and instruction-tuning and the VLM's performance in those classes; when trained with sufficient data, VLMs can match the accuracy of state-of-the-art classification models. Based on these findings, we enhance a VLM by integrating classification-focused datasets into its training, and demonstrate that the enhanced classification performance of the VLM transfers to its general capabilities, resulting in an improvement of 11.8% on the newly collected ImageWikiQA dataset., Comment: Published at NeurIPS 2024

The lack of interactions between single photons prohibits direct nonlinear operations in quantum optical circuits, representing a central obstacle in photonic quantum technologies. Here, we demonstrate multi-mode nonlinear photonic circuits where both linear and direct nonlinear operations can be programmed with high precision at the single-photon level. Deterministic nonlinear interaction is realized with a tunable quantum dot embedded in a nanophotonic waveguide mediating interactions between individual photons within a temporal linear optical interferometer. We demonstrate the capability to reprogram the nonlinear photonic circuits and implement protocols where strong nonlinearities are required, in particular for quantum simulation of anharmonic molecular dynamics, thereby showcasing the new key functionalities enabled by our technology., Comment: 8 pages, 3 figures

Massive web-crawled image-text datasets lay the foundation for recent progress in multimodal learning. These datasets are designed with the goal of training a model to do well on standard computer vision benchmarks, many of which, however, have been shown to be English-centric (e.g., ImageNet). Consequently, existing data curation techniques gravitate towards using predominantly English image-text pairs and discard many potentially useful non-English samples. Our work questions this practice. Multilingual data is inherently enriching not only because it provides a gateway to learn about culturally salient concepts, but also because it depicts common concepts differently from monolingual data. We thus conduct a systematic study to explore the performance benefits of using more samples of non-English origins with respect to English vision tasks. By translating all multilingual image-text pairs from a raw web crawl to English and re-filtering them, we increase the prevalence of (translated) multilingual data in the resulting training set. Pre-training on this dataset outperforms using English-only or English-dominated datasets on ImageNet, ImageNet distribution shifts, image-English-text retrieval and on average across 38 tasks from the DataComp benchmark. On a geographically diverse task like GeoDE, we also observe improvements across all regions, with the biggest gain coming from Africa. In addition, we quantitatively show that English and non-English data are significantly different in both image and (translated) text space. We hope that our findings motivate future work to be more intentional about including multicultural and multilingual data, not just when non-English or geographically diverse tasks are involved, but to enhance model capabilities at large., Comment: NeurIPS 2024 Spotlight paper

Quantum computers are becoming a reality and numerous quantum computing applications with a near-term perspective (e.g., for finance, chemistry, machine learning, and optimization) and with a long-term perspective (e.g., for cryptography or unstructured search) are currently being investigated. However, designing and realizing potential applications for these devices in a scalable fashion requires automated, efficient, and user-friendly software tools that cater to the needs of end users, engineers, and physicists at every level of the entire quantum software stack. Many of the problems to be tackled in that regard are similar to design problems from the classical realm for which sophisticated design automation tools have been developed in the previous decades. The Munich Quantum Toolkit (MQT) is a collection of software tools for quantum computing developed by the Chair for Design Automation at the Technical University of Munich which explicitly utilizes this design automation expertise. Our overarching objective is to provide solutions for design tasks across the entire quantum software stack. This entails high-level support for end users in realizing their applications, efficient methods for the classical simulation, compilation, and verification of quantum circuits, tools for quantum error correction, support for physical design, and more. These methods are supported by corresponding data structures (such as decision diagrams) and core methods (such as SAT encodings/solvers). All of the developed tools are available as open-source implementations and are hosted on https://github.com/cda-tum., Comment: 8 pages, 4 figures, a live version of this document is available at https://mqt.readthedocs.io

The progression of X-ray technology introduces diverse image styles that need to be adapted to the preferences of radiologists. To support this task, we introduce a novel deep learning-based metric that quantifies style differences of non-matching image pairs. At the heart of our metric is an encoder capable of generating X-ray image style representations. This encoder is trained without any explicit knowledge of style distances by exploiting Simple Siamese learning. During inference, the style representations produced by the encoder are used to calculate a distance metric for non-matching image pairs. Our experiments investigate the proposed concept for a disclosed reproducible and a proprietary image processing pipeline along two dimensions: First, we use a t-distributed stochastic neighbor embedding (t-SNE) analysis to illustrate that the encoder outputs provide meaningful and discriminative style representations. Second, the proposed metric calculated from the encoder outputs is shown to quantify style distances for non-matching pairs in good alignment with the human perception. These results confirm that our proposed method is a promising technique to quantify style differences, which can be used for guided style selection as well as automatic optimization of image pipeline parameters.

Literature is the primary expression of scientific knowledge and an important source of research data. However, scientific knowledge expressed in narrative text documents is not inherently machine reusable. To facilitate knowledge reuse, e.g. for synthesis research, scientific knowledge must be extracted from articles and organized into databases post-publication. The high time costs and inaccuracies associated with completing these activities manually has driven the development of techniques that automate knowledge extraction. Tackling the problem with a different mindset, we propose a pre-publication approach, known as reborn, that ensures scientific knowledge is born reusable, i.e. produced in a machine-reusable format during knowledge production. We implement the approach using the Open Research Knowledge Graph infrastructure for FAIR scientific knowledge organization. We test the approach with three use cases, and discuss the role of publishers and editors in scaling the approach. Our results suggest that the proposed approach is superior compared to classical manual and semi-automated post-publication extraction techniques in terms of knowledge richness and accuracy as well as technological simplicity.

Strength, ductility, and failure properties of metals are tailored by plastic deformation routes. Predicting these properties requires modeling of the structural dynamics and stress evolution taking place on several length scales. Progress has been hampered by a lack of representative 3D experimental data at industrially relevant degrees of deformation. We present an X-ray imaging based 3D mapping of an aluminum polycrystal deformed to the ultimate tensile strength (32% elongation). The extensive dataset reveals significant intra-grain stress variations (36 MPa) up to at least half of the inter-grain variations (76 MPa), which are dominated by grain orientation effects. Local intra-grain stress concentrations are candidates for damage nucleation. Such data are important for models of structure-property relations and damage.

Spin and valley flavor polarization plays a central role in the many-body physics of flat band graphene, with fermi surface reconstructions often accompanied by quantized anomalous Hall and superconducting state observed in a variety of experimental systems. Here we describe an optical technique that sensitively and selectively detects flavor textures via the exciton response of a proximal transition metal dichalcogenide layer. Through a systematic study of rhombohedral and rotationally faulted graphene bilayers and trilayers, we show that when the semiconducting dichalcogenide is in direct contact with the graphene, the exciton response is most sensitive to the large momentum rearrangement of the Fermi surface, providing information that is distinct from and complementary to electrical compressibility measurements. The wide-field imaging capability of optical probes allows us to obtain spatial maps of flavor orders with high throughput, and with broad temperature and device compatibility. Our work paves the way for optical probing and imaging of flavor orders in flat band graphene systems., Comment: 29 pages, 4 figures, with supplementary materials

Recent experimental achievements have demonstrated the potential of neutral atom architectures for fault-tolerant quantum computing. These architectures feature the dynamic rearrangement of atoms during computation, enabling nearly arbitrary two-dimensional rearrangements. Additionally, they employ a zoned layout with dedicated regions for entangling, storage, and readout. This architecture requires design automation software that efficiently compiles quantum circuits to this hardware and takes care that atoms are in the right place at the right time. In this paper, we initiate this line of work by providing, (1) an abstract model of the novel architecture and, (2) an efficient solution to the routing problem of entangling gates. By this, we aim to maximize the parallelism of entangling gates and minimize the overhead caused by the routing of atoms between zones. In addition to that, we keep the realm of fault-tolerant quantum computing in mind and consider logical qubit arrays, each of which encodes one logical qubit. We implemented the proposed idea as a tool called NALAC and demonstrated its effectiveness and efficiency by showing that it can significantly reduce the routing overhead of logical entangling gates compared to the naive approach. As part of the Munich Quantum Toolkit (MQT), NALAC is publicly available as open-source at https://github.com/cda-tum/mqt-qmap., Comment: 12 pages, 19 figures v2: revised manuscript

The integer quantum Hall effect can be observed in a two-dimensional conductor penetrated by a perpendicular magnetic field and with edges connecting the current carrying contacts. Its signature is a state of quantized Hall and simultaneously vanishing longitudinal resistances. A widely accepted model is the Landauer-B\"uttiker picture, which assumes an incompressible, i.e., electrically insulating bulk state surrounded by current carrying one-dimensional edge channels. This single-particle model is challenged by the screening theory. It derives, that electron-electron interaction leads to a fragmentation of the Hall bar into compressible and incompressible strips, where the latter gradually shift from the sample edges into the bulk with increasing magnetic field. This suggests a transition from edge to bulk current if we associate current flow with incompressible strips. We present a direct experimental proof of this transition from edge to bulk current. Our results support the screening theory.

Large Language Models (LLMs) trained on code are revolutionizing the software development process. Increasingly, code LLMs are being integrated into software development environments to improve the productivity of human programmers, and LLM-based agents are beginning to show promise for handling complex tasks autonomously. Realizing the full potential of code LLMs requires a wide range of capabilities, including code generation, fixing bugs, explaining and documenting code, maintaining repositories, and more. In this work, we introduce the Granite series of decoder-only code models for code generative tasks, trained with code written in 116 programming languages. The Granite Code models family consists of models ranging in size from 3 to 34 billion parameters, suitable for applications ranging from complex application modernization tasks to on-device memory-constrained use cases. Evaluation on a comprehensive set of tasks demonstrates that Granite Code models consistently reaches state-of-the-art performance among available open-source code LLMs. The Granite Code model family was optimized for enterprise software development workflows and performs well across a range of coding tasks (e.g. code generation, fixing and explanation), making it a versatile all around code model. We release all our Granite Code models under an Apache 2.0 license for both research and commercial use., Comment: Corresponding Authors: Rameswar Panda, Ruchir Puri; Equal Contributors: Mayank Mishra, Matt Stallone, Gaoyuan Zhang

Generative modeling via stochastic processes has led to remarkable empirical results as well as to recent advances in their theoretical understanding. In principle, both space and time of the processes can be discrete or continuous. In this work, we study time-continuous Markov jump processes on discrete state spaces and investigate their correspondence to state-continuous diffusion processes given by SDEs. In particular, we revisit the $\textit{Ehrenfest process}$, which converges to an Ornstein-Uhlenbeck process in the infinite state space limit. Likewise, we can show that the time-reversal of the Ehrenfest process converges to the time-reversed Ornstein-Uhlenbeck process. This observation bridges discrete and continuous state spaces and allows to carry over methods from one to the respective other setting. Additionally, we suggest an algorithm for training the time-reversal of Markov jump processes which relies on conditional expectations and can thus be directly related to denoising score matching. We demonstrate our methods in multiple convincing numerical experiments.

This paper introduces a new biologically-inspired training method named Continual Learning through Adjustment Suppression and Sparsity Promotion (CLASSP). CLASSP is based on two main principles observed in neuroscience, particularly in the context of synaptic transmission and Long-Term Potentiation (LTP). The first principle is a decay rate over the weight adjustment, which is implemented as a generalization of the AdaGrad optimization algorithm. This means that weights that have received many updates should have lower learning rates as they likely encode important information about previously seen data. However, this principle results in a diffuse distribution of updates throughout the model, as it promotes updates for weights that haven't been previously updated, while a sparse update distribution is preferred to leave weights unassigned for future tasks. Therefore, the second principle introduces a threshold on the loss gradient. This promotes sparse learning by updating a weight only if the loss gradient with respect to that weight is above a certain threshold, i.e. only updating weights with a significant impact on the current loss. Both principles reflect phenomena observed in LTP, where a threshold effect and a gradual saturation of potentiation have been observed. CLASSP is implemented in a Python/PyTorch class, making it applicable to any model. When compared with Elastic Weight Consolidation (EWC) using Computer Vision and sentiment analysis datasets, CLASSP demonstrates superior performance in terms of accuracy and memory footprint., Comment: In this version I included a new experiment in text classification using transformer architecture

This work introduces a novel R package for concise, informative summaries of machine learning models. We take inspiration from the summary function for (generalized) linear models in R, but extend it in several directions: First, our summary function is model-agnostic and provides a unified summary output also for non-parametric machine learning models; Second, the summary output is more extensive and customizable -- it comprises information on the dataset, model performance, model complexity, model's estimated feature importances, feature effects, and fairness metrics; Third, models are evaluated based on resampling strategies for unbiased estimates of model performances, feature importances, etc. Overall, the clear, structured output should help to enhance and expedite the model selection process, making it a helpful tool for practitioners and researchers alike., Comment: 9 pages

Supercontinua are broadband spectra that are essential to optical spectroscopy, sensing, imaging, and metrology. They are generated from ultrashort laser pulses through nonlinear frequency conversion in fibers, bulk media, and chip-integrated waveguides. For any generating platform, balancing the competing criteria of strong nonlinearity, transparency, and absence of multiphoton absorption is a key challenge. Here, we explore supercontinuum generation in integrated gallium nitride (GaN) waveguides, which combine a high Kerr-nonlinearity, mid-infrared transparency, and a large bandgap that prevents two- and three-photon absorption in the technologically important telecom C-band, where compact erbium-based pump lasers exist. Using this type of laser, we demonstrate tunable dispersive waves and gap-free spectra extending to almost 4 micron in wavelength, relevant to functional group chemical sensing. Additionally, leveraging the material's second-order nonlinearity, we implement on-chip f-2f interferometry to detect the pump laser's carrier-envelope offset frequency, which enables precision metrology. These results demonstrate the versatility of GaN-on-sapphire as a new platform for broadband nonlinear photonics.