Search

We measure exoplanet occurrence rate as a function of isochrone and gyrochronology ages using confirmed and candidate planets identified in Q1-17 DR25 Kepler data. We employ Kepler's pipeline detection efficiency to correct for the expected number of planets in each age bin. We examine the occurrence rates for planets with radii $0.2 \leq Rp \leq 20$ R$_\oplus$ and orbital periods $0.2 \leq P \leq 100$ days for FGK stars with ages between $1.5-8$ Gyr using the inverse detection efficiency method. We find no significant trend between occurrence rate and stellar ages; a slight, decreasing trend (within $1.5-2.5$ $\sigma$) only emerges for low-mass and metal-rich stars that dominate our sample. We isolate the effects of mass and metallicity on the occurrence rate trend with age, but find the results to be inconclusive due to weak trends and small sample size. Our results hint that the exoplanet occurrence rate may decrease over time due to dynamical instability from planet-planet scattering or planet ejection, but accurate ages and larger sample sizes are needed to resolve a clear relation between occurrence rate and age., Comment: 13 pages, 5 figures, 2 tables. Accepted in AJ

Real-world networks are often characterized by simultaneous interactions between multiple agents that adapt themselves due to feedback from the environment. In this article, we investigate the dynamics of an adaptive multilayer network of Kuramoto oscillators with higher-order interactions. The dynamics of the nodes within the layers are adaptively controlled through the global synchronization order parameter with the adaptations present alongside both pairwise and higher-order interactions. We first explore the dynamics with a linear form of the adaptation function and discover a tiered transition to synchronization, along with continuous and abrupt routes to synchronization. Multiple routes to synchronization are also observed due to the presence of multiple stable states. We investigate the bifurcations behind these routes and illustrate the basin of attraction to attain a deeper understanding of the multistability, that is born as a consequence of the adaptive interactions. When nonlinear adaptation is infused in the system, we observe three different kinds of tiered transition to synchronization, viz., continuous tiered, discontinuous tiered, and tiered transition with a hysteretic region. Our study provides an overview of how inducting order parameter adaptations in higher-order multilayer networks can influence dynamics and alter the route to synchronization in dynamical systems., Comment: 12 pages, 8 figures

Quantized anomalous Hall effects (QAHEs) occur in remarkable electronic states which possess not only quantized Hall signals but in some cases regions of dissipationless electron transport. The initial demonstrations of a QAHE in a magnetically-doped topological insulator (TI) required temperatures below 100 mK, and since then a major focus of the field has been to increase the temperature scale. Here, we report quantized Hall signals up to 10 K (in what is known as the parity anomaly state) in TI/magnet bilayers made by mechanical assembly, rather than by conventional deposition techniques. This is a factor of 100 higher temperature than any previous realization of a QAHE in a proximity-coupled TI/magnet heterostructure made by deposition, and approximately twice the previous record for any QAHE system.

Active Curriculum Language Modeling (ACLM; Hong et al., 2023) is a learner directed approach to training a language model. We proposed the original version of this process in our submission to the BabyLM 2023 task, and now we propose an updated ACLM process for the BabyLM 2024 task. ACLM involves an iteratively- and dynamically-constructed curriculum informed over the training process by a model of uncertainty; other training items that are similarly uncertain to a least certain candidate item are prioritized. Our new process improves the similarity model so that it is more dynamic, and we run ACLM over the most successful model from the BabyLM 2023 task: ELC-BERT (Charpentier and Samuel, 2023). We find that while our models underperform on fine-grained grammatical inferences, they outperform the BabyLM 2024 official base-lines on common-sense and world-knowledge tasks. We make our code available at https: //github.com/asayeed/ActiveBaby.

Synchronization forms the basis of many coordination phenomena in natural systems, enabling them to function cohesively and support their fundamental operations. However, there are scenarios where synchronization disrupts a system's proper functioning, necessitating mechanisms to control or suppress it. While several methods exist for controlling synchronization in non-spatially embedded oscillators, to the best of our knowledge no such strategies have been developed for swarmalators (oscillators that simultaneously move in space and synchronize in time). In this work, we address this gap by introducing a novel control strategy based on Hamiltonian control theory to suppress synchronization in a system of swarmalators confined to a one-dimensional space. The numerical investigations we performed, demonstrate that the proposed control strategy effectively suppresses synchronized dynamics within the swarmalator population. We studied the impact of the number of controlled swarmalators as well as the strength of the control term, in its original form and in a simplified one.

We study the collective dynamics of swarmalators subjected to periodic (sinusoidal) forcing. Although previous research focused on the simplified case of motion in a one-dimensional (1D) periodic domain, we extend this analysis to the more realistic scenario of motion in two and three spatial dimensions with periodic boundary conditions. In doing so, we identify analogues of the 1D states and characterize their dynamics and stability boundaries analytically. Additionally, we investigate the forced swarmalators model with power-law interaction kernels, finding that the analytically tractable model with periodic boundary conditions can reproduce the observed dynamic behaviors of this more complex model., Comment: 11 pages, 8 figures

Here, we present the outcomes from the second Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry, which engaged participants across global hybrid locations, resulting in 34 team submissions. The submissions spanned seven key application areas and demonstrated the diverse utility of LLMs for applications in (1) molecular and material property prediction; (2) molecular and material design; (3) automation and novel interfaces; (4) scientific communication and education; (5) research data management and automation; (6) hypothesis generation and evaluation; and (7) knowledge extraction and reasoning from scientific literature. Each team submission is presented in a summary table with links to the code and as brief papers in the appendix. Beyond team results, we discuss the hackathon event and its hybrid format, which included physical hubs in Toronto, Montreal, San Francisco, Berlin, Lausanne, and Tokyo, alongside a global online hub to enable local and virtual collaboration. Overall, the event highlighted significant improvements in LLM capabilities since the previous year's hackathon, suggesting continued expansion of LLMs for applications in materials science and chemistry research. These outcomes demonstrate the dual utility of LLMs as both multipurpose models for diverse machine learning tasks and platforms for rapid prototyping custom applications in scientific research., Comment: Updating author information, the submission remains largely unchanged. 98 pages total

The financial sector's dependence on digital infrastructure increases its vulnerability to cybersecurity threats, requiring strong IT security protocols with other entities. This collaboration, however, is often identified as the most vulnerable link in the chain of cybersecurity. Adopting both symbolic and substantive measures lessens the impact of IT security spending on decreasing the frequency of data security breaches in the long run. The Protection Motivation Theory clarifies actions triggered by data sharing with other organizations, and the Institutional theory aids in comprehending the intricate relationship between transparency and organizational conduct. We investigate how things like regulatory pressure, teamwork among institutions, and people's motivations to protect themselves influence cybersecurity. By using simple theories to understand these factors, this research aims to provide insights that can help financial institutions make better decisions to protect. We have also included the discussion, conclusion, and future directions in regard to collaboration in financial sector cybersecurity for exploring impact of resource sharing., Comment: 8 pages, 2 figures

Machine learning (ML) has become a ubiquitous tool across various domains of data mining and big data analysis. The efficacy of ML models depends heavily on high-quality datasets, which are often complicated by the presence of missing values. Consequently, the performance and generalization of ML models are at risk in the face of such datasets. This paper aims to examine the nuanced impact of missing values on ML workflows, including their types, causes, and consequences. Our analysis focuses on the challenges posed by missing values, including biased inferences, reduced predictive power, and increased computational burdens. The paper further explores strategies for handling missing values, including imputation techniques and removal strategies, and investigates how missing values affect model evaluation metrics and introduces complexities in cross-validation and model selection. The study employs case studies and real-world examples to illustrate the practical implications of addressing missing values. Finally, the discussion extends to future research directions, emphasizing the need for handling missing values ethically and transparently. The primary goal of this paper is to provide insights into the pervasive impact of missing values on ML models and guide practitioners toward effective strategies for achieving robust and reliable model outcomes.

The existence of high-$\alpha$ stars with inferred ages < 6 Gyr has been confirmed recently with large spectroscopic and photometric surveys. However, stellar mergers or binary interactions can induce properties associated with young ages, such as high mass, rapid rotation, or high activity, even in old populations. Literature studies have confirmed that at least some of these apparently young stars are old merger products. However, none have ruled out the possibility of genuinely young high-$\alpha$ stars. Because cool GKM dwarfs spin down, rapid rotation can be used to indicate youth. In this paper, we provide strong evidence that truly young high-$\alpha$ stars exist by studying high-$\alpha$ rotators in the Kepler and K2 field with abundance measurements from GALAH and APOGEE. After excluding close binaries using radial velocity (RV) measurements from Gaia DR3 and multi-epoch RVs from APOGEE, we find a total of 70 high-$\alpha$ rapid rotators with periods ~10-30 days, 29 of which have lithium measurements from GALAH, indicating that they have not gone through past mass transfer or stellar merger events. We identify 10 young high-$\alpha$ candidates with no signs of merger-induced mixing or close companions. One clear example is a G dwarf with a measurable rotation and an age of 1.98$^{+0.12}_{-0.28}$ Gyr that is likely a single star with multiple RV measurements from APOGEE, has significant lithium detection from GALAH (A(Li) = 1.79), and has no signs of planet engulfment., Comment: 21 pages, 13 figures, submitted to AJ

In today's digitally driven landscape, robust Information Technology (IT) risk assessment practices are essential for safeguarding systems, digital communication, and data. This paper introduces 'AssessITS', an actionable method designed to provide organizations with comprehensive guidelines for conducting IT and cybersecurity risk assessments. Drawing extensively from NIST 800-30 Rev 1, COBIT 5, and ISO 31000, 'AssessITS' bridges the gap between high-level theoretical standards and practical implementation challenges. The paper outlines a step-by-step methodology that organizations can simply adopt to systematically identify, analyze, and mitigate IT risks. By simplifying complex principles into actionable procedures, this framework equips practitioners with the tools needed to perform risk assessments independently, without too much reliance on external vendors. The guidelines are developed to be straightforward, integrating practical evaluation metrics that allow for the precise quantification of asset values, threat levels, vulnerabilities, and impacts on confidentiality, integrity, and availability. This approach ensures that the risk assessment process is not only comprehensive but also accessible, enabling decision-makers to implement effective risk mitigation strategies customized to their unique operational contexts. 'AssessITS' aims to enable organizations to enhance their IT security strength through practical, actionable guidance based on internationally recognized standards., Comment: 25 pages, 8 figures, 7 tables

This paper presents a rapid approach to assessing disaster resilience in Florida, particularly regarding Hurricane Helene (2024). This category four storm made landfall on Florida's Gulf Coast in September 2024. Using the Disaster Resilience Index (DRI) developed in this paper, the preparedness and adaptive capacities of communities across counties in Florida are evaluated, identifying the most resilient areas based on three key variables: population size, average per-person income, and the Social Vulnerability Index (SVI). While the Social Vulnerability Index (SVI) accounts for factors like socioeconomic status, household composition, minority status, and housing conditions-key elements in determining a community's resilience to disasters-incorporating a county's population and per person income provides additional insight. A county's total population is directly linked to the number of individuals impacted by a disaster, while personal income reflects a household's capacity to recover. Spatial analysis was performed on the index to compare the vulnerability and resilience levels across thirty-four counties vulnerable to Hurricane Helene's projected path. The results highlight that counties with high income and lower population densities, such as Monroe and Collier, exhibit greater resilience. In contrast, areas with larger populations and higher social vulnerabilities are at greater risk of damage. This study contributes to disaster management planning by providing a rapid yet comprehensive and reassuring socioeconomic impact assessment, offering actionable insights for anticipatory measures and resource allocation.

Vector embeddings derived from large language models (LLMs) show promise in capturing latent information from the literature. Interestingly, these can be integrated into material embeddings, potentially useful for data-driven predictions of materials properties. We investigate the extent to which LLM-derived vectors capture the desired information and their potential to provide insights into material properties without additional training. Our findings indicate that, although LLMs can be used to generate representations reflecting certain property information, extracting the embeddings requires identifying the optimal contextual clues and appropriate comparators. Despite this restriction, it appears that LLMs still have the potential to be useful in generating meaningful materials-science representations., Comment: 10 pages, 7 figures

We study a simple model of swarmalators subject to periodic forcing and confined to move around a one-dimensional ring. This is a toy model for physical systems with a mix of sync, swarming, and forcing such as colloidal micromotors. We find several emergent macrostates and characterize the phase boundaries between them analytically. The most novel state is a swarmalator chimera, where the population splits into two sync dots, which enclose a `train' of swarmalators that run around a peanut-shaped loop., Comment: 9 pages, 7 figures

Cardiac auscultation, an integral tool in diagnosing cardiovascular diseases (CVDs), often relies on the subjective interpretation of clinicians, presenting a limitation in consistency and accuracy. Addressing this, we introduce the BUET Multi-disease Heart Sound (BMD-HS) dataset - a comprehensive and meticulously curated collection of heart sound recordings. This dataset, encompassing 864 recordings across five distinct classes of common heart sounds, represents a broad spectrum of valvular heart diseases, with a focus on diagnostically challenging cases. The standout feature of the BMD-HS dataset is its innovative multi-label annotation system, which captures a diverse range of diseases and unique disease states. This system significantly enhances the dataset's utility for developing advanced machine learning models in automated heart sound classification and diagnosis. By bridging the gap between traditional auscultation practices and contemporary data-driven diagnostic methods, the BMD-HS dataset is poised to revolutionize CVD diagnosis and management, providing an invaluable resource for the advancement of cardiac health research. The dataset is publicly available at this link: https://github.com/mHealthBuet/BMD-HS-Dataset., Comment: 14 pages, 13 figures

Many natural and human-made complex systems feature group interactions that adapt over time in response to their dynamic states. However, most of the existing adaptive network models fall short of capturing these group dynamics, as they focus solely on pairwise interactions. In this study, we employ adaptive higher-order networks to describe these systems by proposing a general framework incorporating both adaptivity and group interactions. We demonstrate that global synchronization can exist in those complex structures, and we provide the necessary conditions for the emergence of a stable synchronous state. Additionally, we analyzed some relevant settings, and we showed that the necessary condition is strongly related to the master stability equation, allowing to separate the dynamical and structural properties. We illustrate our theoretical findings through examples involving adaptive higher-order networks of coupled generalized Kuramoto oscillators with phase lag. We also show that the interplay of group interactions and adaptive connectivity results in the formation of stability regions that can induce transitions between synchronization and desynchronization

Large Language Models (LLMs), such as GPT models, are increasingly used in software engineering for various tasks, such as code generation, requirements management, and debugging. While automating these tasks has garnered significant attention, a systematic study on the impact of varying hyperparameters on code generation outcomes remains unexplored. This study aims to assess LLMs' code generation performance by exhaustively exploring the impact of various hyperparameters. Hyperparameters for LLMs are adjustable settings that affect the model's behaviour and performance. Specifically, we investigated how changes to the hyperparameters: temperature, top probability (top_p), frequency penalty, and presence penalty affect code generation outcomes. We systematically adjusted all hyperparameters together, exploring every possible combination by making small increments to each hyperparameter at a time. This exhaustive approach was applied to 13 Python code generation tasks, yielding one of four outcomes for each hyperparameter combination: no output from the LLM, non executable code, code that fails unit tests, or correct and functional code. We analysed these outcomes for a total of 14,742 generated Python code segments, focusing on correctness, to determine how the hyperparameters influence the LLM to arrive at each outcome. Using correlation coefficient and regression tree analyses, we ascertained which hyperparameters influence which aspect of the LLM. Our results indicate that optimal performance is achieved with a temperature below 0.5, top probability below 0.75, frequency penalty above -1 and below 1.5, and presence penalty above -1. We make our dataset and results available to facilitate replication.

The roadmap is organized into several thematic sections, outlining current computing challenges, discussing the neuromorphic computing approach, analyzing mature and currently utilized technologies, providing an overview of emerging technologies, addressing material challenges, exploring novel computing concepts, and finally examining the maturity level of emerging technologies while determining the next essential steps for their advancement., Comment: 90 pages, 22 figures, roadmap, neuromorphic

Each new generation of English-oriented Large Language Models (LLMs) exhibits enhanced cross-lingual transfer capabilities and significantly outperforms older LLMs on low-resource languages. This prompts the question: Is there a need for LLMs dedicated to a particular low-resource language? We aim to explore this question for Bengali, a low-to-moderate resource Indo-Aryan language native to the Bengal region of South Asia. We compare the performance of open-weight and closed-source LLMs such as LLaMA-3 and GPT-4 against fine-tuned encoder-decoder models across a diverse set of Bengali downstream tasks, including translation, summarization, paraphrasing, question-answering, and natural language inference. Our findings reveal that while LLMs generally excel in reasoning tasks, their performance in tasks requiring Bengali script generation is inconsistent. Key challenges include inefficient tokenization of Bengali script by existing LLMs, leading to increased computational costs and potential performance degradation. Additionally, we highlight biases in machine-translated datasets commonly used for Bengali NLP tasks. We conclude that there is a significant need for a Bengali-oriented LLM, but the field currently lacks the high-quality pretraining and instruction-tuning datasets necessary to develop a highly effective model.

Networks incorporating higher-order interactions are increasingly recognized for their ability to introduce novel dynamics into various processes, including synchronization. Previous studies on synchronization within multilayer networks have often been limited to specific models, such as the Kuramoto model, or have focused solely on higher-order interactions within individual layers. Here, we present a comprehensive framework for investigating synchronization, particularly global synchronization, in multilayer networks with higher-order interactions. Our framework considers interactions beyond pairwise connections, both within and across layers. We demonstrate the existence of a stable global synchronous state, with a condition resembling the master stability function, contingent on the choice of coupling functions. Our theoretical findings are supported by simulations using Hindmarsh-Rose neuronal and R\"{o}ssler oscillators. These simulations illustrate how synchronization is facilitated by higher-order interactions, both within and across layers, highlighting the advantages over scenarios involving interactions within single layers.

The KEYSTONE project aims to enhance our understanding of solar-like oscillators by delivering a catalogue of global asteroseismic parameters (${\Delta\nu}$ and ${\nu_{\rm max}}$) for 173 stars, comprising mainly dwarfs and subgiants, observed by the K2 mission in its short-cadence mode during campaigns 6-19. We derive atmospheric parameters and luminosities using spectroscopic data from TRES, astrometric data from $\textit{Gaia}$, and the infrared flux method (IRFM) for a comprehensive stellar characterisation. Asteroseismic parameters are robustly extracted using three independent methods, complemented by an iterative refinement of the spectroscopic analyses using seismic ${\log g}$ values to enhance parameter accuracy. Our analysis identifies new detections of solar-like oscillations in 159 stars, providing an important complement to already published results from previous campaigns. The catalogue provides homogeneously derived atmospheric parameters and luminosities for the majority of the sample. Comparison between spectroscopic ${T_{\rm eff}}$ and those obtained from the IRFM demonstrates excellent agreement. The iterative approach to spectroscopic analysis significantly enhances the accuracy of the stellar properties derived., Comment: 22 pages, 18 figures, accepted for publication in Astronomy & Astrophysics (A&A)

In an era where informed decision-making is paramount for sustainable development and effective resource management, the role of Earth observations (EO) in shaping economic landscapes cannot be overstated. EO, facilitated by satellites, sensors, and data analytics, is a cornerstone for evidence-based policymaking, risk mitigation, and resource allocation. SERVIR is a joint initiative of US Agency for International Development and NASA. This paper presents a comprehensive survey of relevant economic impact assessment (EIA) work, summarizes SERVIRs potential interests in EIA, and identifies how and where EIA could improve how SERVIR quantifies and communicates the impact of its services., Comment: 40 pages, 13 Tables and 4 Figures

Surgical action localization is a challenging computer vision problem. While it has promising applications including automated training of surgery procedures, surgical workflow optimization, etc., appropriate model design is pivotal to accomplishing this task. Moreover, the lack of suitable medical datasets adds an additional layer of complexity. To that effect, we introduce a new complex dataset of nephrectomy surgeries called UroSlice. To perform the action localization from these videos, we propose a novel model termed as `ViTALS' (Vision Transformer for Action Localization in Surgical Nephrectomy). Our model incorporates hierarchical dilated temporal convolution layers and inter-layer residual connections to capture the temporal correlations at finer as well as coarser granularities. The proposed approach achieves state-of-the-art performance on Cholec80 and UroSlice datasets (89.8% and 66.1% accuracy, respectively), validating its effectiveness., Comment: Nephrectomy surgery, Surgical Phase Recognition, Surgical Workflow Segmentation, 11 pages, 2 figures, 2 tables

Computer vision, particularly vehicle and pedestrian identification is critical to the evolution of autonomous driving, artificial intelligence, and video surveillance. Current traffic monitoring systems confront major difficulty in recognizing small objects and pedestrians effectively in real-time, posing a serious risk to public safety and contributing to traffic inefficiency. Recognizing these difficulties, our project focuses on the creation and validation of an advanced deep-learning framework capable of processing complex visual input for precise, real-time recognition of cars and people in a variety of environmental situations. On a dataset representing complicated urban settings, we trained and evaluated different versions of the YOLOv8 and RT-DETR models. The YOLOv8 Large version proved to be the most effective, especially in pedestrian recognition, with great precision and robustness. The results, which include Mean Average Precision and recall rates, demonstrate the model's ability to dramatically improve traffic monitoring and safety. This study makes an important addition to real-time, reliable detection in computer vision, establishing new benchmarks for traffic management systems., Comment: 5 pages, 2 figures

Living entities in a group communicate and transfer information to one another for a variety of reasons. It might be for foraging food, migration, or escaping threats and obstacles, etc. They do so by interacting with each other and also with the environment. The tools from statistical mechanics and information theory can be useful to analyze the flow of information among the living entities modelled as active (i.e. self-propelling) particles. Here we consider the active particles confined in a circular trap. The self-organisation of the particles crucially depends on whether the trap boundary is soft or hard. We quench the trap boundary from soft to hard instantaneously. After the mechanical quench, the particles suddenly find themselves in a hard potential. The self-organised cluster of the active particles, which was stable when the boundary was soft, becomes unstable. The cluster undergoes extreme deformation after the quench to find another stable configuration suitable for the hard potential. Together with the structural relaxation, information regarding the quench also flows throughout the deforming cluster. Here, we quantify the flow of information by computing local transfer entropy. We find that the flow spans the whole cluster, propagating ballistically.

Ensembles of coupled nonlinear oscillators are a popular paradigm and an ideal benchmark for analyzing complex collective behaviors. The onset of cluster synchronization is found to be at the core of various technological and biological processes. The current literature has investigated cluster synchronization by focusing mostly on the case of attractive coupling among the oscillators. However, the case of two coexisting competing interactions is of practical interest due to their relevance in diverse natural settings, including neuronal networks consisting of excitatory and inhibitory neurons, the coevolving social model with voters of opposite opinions, ecological plant communities with both facilitation and competition, to name a few. In the present article, we investigate the impact of repulsive spanning trees on cluster formation within a connected network of attractively coupled limit cycle oscillators. We successfully predict which nodes belong to each cluster and the emergent frustration of the connected networks independent of the particular local dynamics at the network nodes. We also determine local asymptotic stability of the cluster states using an approach based on the formulation of a master stability function. We additionally validate the emergence of solitary states and antisynchronization for some specific choices of spanning trees and networks., Comment: 19 pages, 18 figures, Accepted for publication in Physical Review E (2024)

The advent of Vision Language Models (VLM) has allowed researchers to investigate the visual understanding of a neural network using natural language. Beyond object classification and detection, VLMs are capable of visual comprehension and common-sense reasoning. This naturally led to the question: How do VLMs respond when the image itself is inherently unreasonable? To this end, we present IllusionVQA: a diverse dataset of challenging optical illusions and hard-to-interpret scenes to test the capability of VLMs in two distinct multiple-choice VQA tasks - comprehension and soft localization. GPT4V, the best performing VLM, achieves 62.99% accuracy (4-shot) on the comprehension task and 49.7% on the localization task (4-shot and Chain-of-Thought). Human evaluation reveals that humans achieve 91.03% and 100% accuracy in comprehension and localization. We discover that In-Context Learning (ICL) and Chain-of-Thought reasoning substantially degrade the performance of Gemini-Pro in the localization task. Tangentially, we discover a potential weakness in the ICL capabilities of VLMs: they fail to locate optical illusions even when the correct answer is in the context window as a few-shot example.

Automatic speech recognition (ASR) converts the human voice into readily understandable and categorized text or words. Although Bengali is one of the most widely spoken languages in the world, there have been very few studies on Bengali ASR, particularly on Bangladeshi-accented Bengali. In this study, audio recordings of spoken digits (0-9) from university students were used to create a Bengali speech digits dataset that may be employed to train artificial neural networks for voice-based digital input systems. This paper also compares the Bengali digit recognition accuracy of several Convolutional Neural Networks (CNNs) using spectrograms and shows that a test accuracy of 98.23% is achievable using parameter-efficient models such as SqueezeNet on our dataset.