Search

Classification is a ubiquitous and fundamental problem in artificial intelligence and machine learning, with extensive efforts dedicated to developing more powerful classifiers and larger datasets. However, the classification task is ultimately constrained by the intrinsic properties of datasets, independently of computational power or model complexity. In this work, we introduce a formal entropy-based measure of classificability, which quantifies the inherent difficulty of a classification problem by assessing the uncertainty in class assignments given feature representations. This measure captures the degree of class overlap and aligns with human intuition, serving as an upper bound on classification performance for classification problems. Our results establish a theoretical limit beyond which no classifier can improve the classification accuracy, regardless of the architecture or amount of data, in a given problem. Our approach provides a principled framework for understanding when classification is inherently fallible and fundamentally ambiguous.

This paper investigates how teachers learn about race in the school context, with a particular focus on teachers' development of racial competency. Using in-depth, semi-structured interviews we find that teachers learn through three sources: from their peers, from years of experience, and from teacher preparation and in-service experiences. Furthermore, we find that learning occurs both informally and formally and that these sources of learning are moderated by three contextual factors: career status, school culture, and out-of-school factors We find that teachers rely most on informal avenues and encounters to develop racial competency.

Local teacher recruitment through "grow-your-own" programs is a prominent strategy to address workforce shortages and ensure that incoming teachers resemble, understand, and have strong connections to their communities. We exploit the staggered rollout of the Teacher Academy of Maryland career and technical education certificate program across public high schools, finding that exposed students were more likely to become teachers by 0.6 percentage points (pp), or 47%. Effects are concentrated among White girls (1.4pp/39%) and Black girls (0.7pp/80%). We also identify positive impacts on wages (5% on average/18% for Black girls), countering a prevailing narrative that teaching leaves one worse off financially relative to other labor market opportunities.

Mounting evidence supporting the advantages of a diverse teacher workforce prompts policymakers to scrutinize existing recruitment pathways. Following four cohorts of Maryland public high-school students over 12 years reveals several insights. Early barriers require timely interventions, aiding students of color in achieving educational milestones that are prerequisites for teacher candidacy (high school graduation, college enrollment). While alternative pathways that bypass traditional undergraduate teacher preparation may help, current approaches still show persistent racial disparities. Data simulations underscore the need for race-conscious policies specifically targeting or differentially benefiting students of color, as race-neutral strategies have minimal impact. Ultimately, multiple race-conscious policy solutions addressing various educational milestones must demonstrate significant effects--approximately 30% increases--to reshape the teacher workforce to align with student body demographics.

The United Nations 2030 Agenda for Sustainable Development outlines 17 goals to address global challenges. However, progress has been slower than expected and, consequently, there is a need to investigate the reasons behind this fact. In this study, we used a novel data-driven methodology to analyze data from 107 countries (2000$-$2022) using unsupervised machine learning techniques. Our analysis reveals strong positive and negative correlations between certain SDGs. The findings show that progress toward the SDGs is heavily influenced by geographical, cultural and socioeconomic factors, with no country on track to achieve all goals by 2030. This highlights the need for a region specific, systemic approach to sustainable development that acknowledges the complex interdependencies of the goals and the diverse capacities of nations. Our approach provides a robust framework for developing efficient and data-informed strategies, to promote cooperative and targeted initiatives for sustainable progress.

In recent years, several models using Quaternion-Valued Convolutional Neural Networks (QCNNs) for different problems have been proposed. Although the definition of the quaternion convolution layer is the same, there are different adaptations of other atomic components to the quaternion domain, e.g., pooling layers, activation functions, fully connected layers, etc. However, the effect of selecting a specific type of these components and the way in which their interactions affect the performance of the model still unclear. Understanding the impact of these choices on model performance is vital for effectively utilizing QCNNs. This paper presents a statistical analysis carried out on experimental data to compare the performance of existing components for the image classification problem. In addition, we introduce a novel Fully Quaternion ReLU activation function, which exploits the unique properties of quaternion algebra to improve model performance., Comment: 17 pages, 6 figures

Understanding the emergence and evolution of information handling is essential for unraveling the origins of life. Traditional genetic-first and metabolism-first models fall short in explaining how instructional information control systems naturally arise from molecular dynamics. To address this gap, we adopt an information-first approach, integrating Hofmeyr's (F, A)-systems -- an extension of Rosen's (M, R)-systems -- with temporal parametrization and multiscale causality. These models, which embody closure to efficient causation while remaining open to formal causation, provide a robust framework for primitive autopoiesis, anticipation, and adaptation. We establish a formal equivalence between extended (F, A)-systems and communicating X-machines, resolving self-referential challenges and demonstrating the hypercomputational nature of life processes. Our stepwise model traces the evolution of information handling from simple reaction networks that recognize regular languages to self-replicating chemical systems with memory and anticipatory capabilities. This transition from analog to digital architectures enhances evolutionary robustness and aligns with experimental evidence suggesting that chemical computation does not require life-specific chemistry. Furthermore, we incorporate open-ended evolutionary dynamics driven by computational undecidability and irreducibility, reinforcing the necessity of unconventional computing frameworks. This computational enactivist perspective provides a cohesive theoretical basis for a recently proposed trialectic between autopoiesis, anticipation and adaptation in order to solve the problem of relevance. By highlighting the critical role of hypercomputational processes in life's emergence and evolution, our framework offers new insights into the fundamental principles underlying biological information processing., Comment: 12 pages

The US teaching force remains disproportionately white while the student body grows more diverse. It is therefore important to understand how and under what conditions white teachers learn racial competency. This study applies a mixed-methods approach to investigate the hypothesis that Black peers improve white teachers' effectiveness when teaching Black students. The quantitative portion of this study relies on longitudinal data from North Carolina to show that having a Black same-grade peer significantly improves the achievement and reduces the suspension rates of white teachers' Black students. These effects are persistent over time and largest for novice teachers. Qualitative evidence from open-ended interviews of North Carolina public school teachers reaffirms these findings. Broadly, our findings suggest that the positive impact of Black teachers' ability to successfully teach Black students is not limited to their direct interaction with Black students but is augmented by spillover effects on early-career white teachers, likely through peer learning.

Now that the most acute phase of the COVID-19 crisis is over, public conversation has turned to the millions of students who are still struggling academically and emotionally--and how the nation's schools ought to respond. Decisions that education leaders make right now will determine whether this generation of students recovers or continues to lose ground. That's why policies that lower expectations--such as rescinding "third-grade reading guarantees" or keeping pandemic-era no-grading policies in effect--are so troubling. This study uses nationally representative survey data to explore how teacher expectations differ by sector, and how they affect achievement, attainment, and other outcomes. [Foreword by Amber M. Northern and David Griffith.]

The performance of various superconducting devices operating at ultra-low temperatures is impaired by the presence of non-equilibrium quasiparticles. Inelastic quasiparticle (QP) tunneling across Josephson junctions in superconducting qubits results in decoherence and spurious excitations and, notably, can trigger correlated errors that severely impede quantum error correction. In this work, we use "gap engineering" to suppress the tunneling of low-energy quasiparticles in Al-based transmon qubits, a leading building block for superconducting quantum processors. By implementing potential barriers for QP, we strongly suppress QP tunneling across the junction and preserve charge parity for over $10^3$ seconds. The suppression of QP tunneling also results in a reduction in the qubit energy relaxation rates. The demonstrated approach to gap engineering can be easily implemented in all Al-based circuits with Josephson junctions., Comment: 10 pages, 6 figures, 2 tables; Added references for the supplementary information and updated an author's preferred initials; Revised version in response to referee comments

Since their first applications, Convolutional Neural Networks (CNNs) have solved problems that have advanced the state-of-the-art in several domains. CNNs represent information using real numbers. Despite encouraging results, theoretical analysis shows that representations such as hyper-complex numbers can achieve richer representational capacities than real numbers, and that Hamilton products can capture intrinsic interchannel relationships. Moreover, in the last few years, experimental research has shown that Quaternion-Valued CNNs (QCNNs) can achieve similar performance with fewer parameters than their real-valued counterparts. This paper condenses research in the development of QCNNs from its very beginnings. We propose a conceptual organization of current trends and analyze the main building blocks used in the design of QCNN models. Based on this conceptual organization, we propose future directions of research.

A recent increase in data availability has allowed the possibility to perform different statistical linguistic studies. Here we use the Google Books Ngram dataset to analyze word flow among English, French, German, Italian, and Spanish. We study what we define as ``migrant words'', a type of loanwords that do not change their spelling. We quantify migrant words from one language to another for different decades, and notice that most migrant words can be aggregated in semantic fields and associated to historic events. We also study the statistical properties of accumulated migrant words and their rank dynamics. We propose a measure of use of migrant words that could be used as a proxy of cultural influence. Our methodology is not exempt of caveats, but our results are encouraging to promote further studies in this direction., Comment: 13 pages

Most models of complex systems have been homogeneous, i.e., all elements have the same properties (spatial, temporal, structural, functional). However, most natural systems are heterogeneous: few elements are more relevant, larger, stronger, or faster than others. In homogeneous systems, criticality -- a balance between change and stability, order and chaos -- is usually found for a very narrow region in the parameter space, close to a phase transition. Using random Boolean networks -- a general model of discrete dynamical systems -- we show that heterogeneity -- in time, structure, and function -- can broaden additively the parameter region where criticality is found. Moreover, parameter regions where antifragility is found are also increased with heterogeneity. However, maximum antifragility is found for particular parameters in homogeneous networks. Our work suggests that the "optimal" balance between homogeneity and heterogeneity is non-trivial, context-dependent, and in some cases, dynamic.

Statistical linguistics has advanced considerably in recent decades as data has become available. This has allowed researchers to study how statistical properties of languages change over time. In this work, we use data from Twitter to explore English and Spanish considering the rank diversity at different scales: temporal (from 3 to 96 hour intervals), spatial (from 3km to 3000+km radii), and grammatical (from monograms to pentagrams). We find that all three scales are relevant. However, the greatest changes come from variations in the grammatical scale. At the lowest grammatical scale (monograms), the rank diversity curves are most similar, independently on the values of other scales, languages, and countries. As the grammatical scale grows, the rank diversity curves vary more depending on the temporal and spatial scales, as well as on the language and country. We also study the statistics of Twitter-specific tokens: emojis, hashtags, and user mentions. These particular type of tokens show a sigmoid kind of behaviour as a rank diversity function. Our results are helpful to quantify aspects of language statistics that seem universal and what may lead to variations.

Criticality has been proposed as a mechanism for the emergence of complexity, life, and computation, as it exhibits a balance between robustness and adaptability. In classic models of complex systems where structure and dynamics are considered homogeneous, criticality is restricted to phase transitions, leading either to robust (ordered) or adaptive (chaotic) phases in most of the parameter space. Many real-world complex systems, however, are not homogeneous. Some elements change in time faster than others, with slower elements (usually the most relevant) providing robustness, and faster ones being adaptive. Structural patterns of connectivity are also typically heterogeneous, characterized by few elements with many interactions and most elements with only a few. Here we take a few traditionally homogeneous dynamical models and explore their heterogeneous versions, finding evidence that heterogeneity extends criticality. Thus, parameter fine-tuning is not necessary to reach a phase transition and obtain the benefits of (homogeneous) criticality. Simply adding heterogeneity can extend criticality, making the search/evolution of complex systems faster and more reliable. Our results add theoretical support for the ubiquitous presence of heterogeneity in physical, social, and technological systems, as natural selection can exploit heterogeneity to evolve complexity "for free". In artificial systems and biological design, heterogeneity may also be used to extend the parameter range that allows for criticality., Comment: 17 pages, 3 figures