A novel approach to learning is presented, combining features of on-line and off-line methods to achieve considerable performance in the task of learning a backgammon value function in a process that exploits the processing power of parallel supercomputers. The off-line methods comprise a set of techniques for parallelizing neural network training and $TD(λ)$ reinforcement learning; here Monte-Carlo ``Rollouts'' are introduced as a massively parallel on-line policy improvement technique which applies resources to the decision points encountered during the search of the game tree to further augment the learned value function estimate. A level of play roughly as good as, or possibly better than, the current champion human and computer backgammon players has been achieved in a short period of learning.
Dongseok Kim, Hyoungsun Choi, Mohamed Jismy Aashik Rasool
et al.
The success of federated learning (FL) ultimately depends on how strategic participants behave under partial observability, yet most formulations still treat FL as a static optimization problem. We instead view FL deployments as governed strategic systems and develop an analytical framework that separates welfare-improving behavior from metric gaming. Within this framework, we introduce indices that quantify manipulability, the price of gaming, and the price of cooperation, and we use them to study how rules, information disclosure, evaluation metrics, and aggregator-switching policies reshape incentives and cooperation patterns. We derive threshold conditions for deterring harmful gaming while preserving benign cooperation, and for triggering auto-switch rules when early-warning indicators become critical. Building on these results, we construct a design toolkit including a governance checklist and a simple audit-budget allocation algorithm with a provable performance guarantee. Simulations across diverse stylized environments and a federated learning case study consistently match the qualitative and quantitative patterns predicted by our framework. Taken together, our results provide design principles and operational guidelines for reducing metric gaming while sustaining stable, high-welfare cooperation in FL platforms.
This study aims to analyze the influence of the discourse of international organizations on university social responsibility (USR), from international organizations, in the policy and management frameworks of the State University System in Colombia. Then, we reviewed institutional documents of global and regional organizations, as well as educational policy documents from higher education institutions. Using Atlas Ti. 23 software, our findings indicate that universities adopt the models proposed by international organizations with different approaches. The policy and management frameworks, aligned with François Vallaeys’“impact” perspective, emphasize the importance of tangible results. Seven of the universities studied articulate a well-defined USR policy, with most integrating it into extension and outreach processes. We conclude that USR should be related to education, initiative integration, management, research, organizational culture, and management indicators. We recommend further studies on how USR policies are integrated in higher education institutions.
History of scholarship and learning. The humanities, Social Sciences
Shahid Mahmood, Asifa Iqbal, Amel Ali Alhussan
et al.
Abstract This study investigates the pivotal role of political stability, good governance, and institutional support in achieving Sustainable Development Goals (SDGs) 7, 11, and 12 in Pakistan, an emerging economy. SDG 7 deals with affordable, reliable, sustainable and modern energy for all, SDG 11 deals with sustainable cities and communities and SDGs 12 promotes sustainable consumption and production patterns. The data was collected using a quantitative method from various sources, including the Ministry of Planning, Development and Special Initiatives, the Ministry of Climate Change, the Pakistan Institute of Development Economics, and various NGOs involved in issues such as renewable energy, sustainable cities, and responsible consumption. The collected data was analyzed by using SMART PLS. This study concludes that political stability serves as the foundation for achieving sustainable development goals. It has been observed that the implementation of good governance, which includes principles of transparency, accountability, and public participation, significantly enhances the effectiveness of policies aimed at achieving these SDGs. More specifically, sufficient financial resources and appropriate physical infrastructure are identified as key factors in addressing challenges related to Pakistan’s sustainable development goals. Policymakers should also prioritize investments in sustainable infrastructure projects that align with SDGs 7, 11, and 12, such as investing in renewable energy sources, sustainable urban development, and promoting responsible consumption and production practices.
History of scholarship and learning. The humanities, Social Sciences
While deep reinforcement learning (RL) has fueled multiple high-profile successes in machine learning, it is held back from more widespread adoption by its often poor data efficiency and the limited generality of the policies it produces. A promising approach for alleviating these limitations is to cast the development of better RL algorithms as a machine learning problem itself in a process called meta-RL. Meta-RL is most commonly studied in a problem setting where, given a distribution of tasks, the goal is to learn a policy that is capable of adapting to any new task from the task distribution with as little data as possible. In this survey, we describe the meta-RL problem setting in detail as well as its major variations. We discuss how, at a high level, meta-RL research can be clustered based on the presence of a task distribution and the learning budget available for each individual task. Using these clusters, we then survey meta-RL algorithms and applications. We conclude by presenting the open problems on the path to making meta-RL part of the standard toolbox for a deep RL practitioner.
Practical Imitation Learning (IL) systems rely on large human demonstration datasets for successful policy learning. However, challenges lie in maintaining the quality of collected data and addressing the suboptimal nature of some demonstrations, which can compromise the overall dataset quality and hence the learning outcome. Furthermore, the intrinsic heterogeneity in human behavior can produce equally successful but disparate demonstrations, further exacerbating the challenge of discerning demonstration quality. To address these challenges, this paper introduces Learning to Discern (L2D), an offline imitation learning framework for learning from demonstrations with diverse quality and style. Given a small batch of demonstrations with sparse quality labels, we learn a latent representation for temporally embedded trajectory segments. Preference learning in this latent space trains a quality evaluator that generalizes to new demonstrators exhibiting different styles. Empirically, we show that L2D can effectively assess and learn from varying demonstrations, thereby leading to improved policy performance across a range of tasks in both simulations and on a physical robot.
Solving complex classification tasks using deep neural networks typically requires large amounts of annotated data. However, corresponding class labels are noisy when provided by error-prone annotators, e.g., crowdworkers. Training standard deep neural networks leads to subpar performances in such multi-annotator supervised learning settings. We address this issue by presenting a probabilistic training framework named multi-annotator deep learning (MaDL). A downstream ground truth and an annotator performance model are jointly trained in an end-to-end learning approach. The ground truth model learns to predict instances' true class labels, while the annotator performance model infers probabilistic estimates of annotators' performances. A modular network architecture enables us to make varying assumptions regarding annotators' performances, e.g., an optional class or instance dependency. Further, we learn annotator embeddings to estimate annotators' densities within a latent space as proxies of their potentially correlated annotations. Together with a weighted loss function, we improve the learning from correlated annotation patterns. In a comprehensive evaluation, we examine three research questions about multi-annotator supervised learning. Our findings show MaDL's state-of-the-art performance and robustness against many correlated, spamming annotators.
Abstract This study aimed to examine the influence of emotional media information on information-processing mechanisms in the current post-truth era. A cognitive conflict monitoring and evaluation (CCME) model was proposed to explore news audiences’ attention and implicit attitudes. The study had a 2 (information type, emotional vs. neutral) × 2 (condition, compatible vs. incompatible) × 3 (electrode position: Fz vs. Cz vs. Pz) design, and an implicit association test (IAT) was administered, with event-related potential (ERP) data collected. The results revealed that emotional information evoked different information-processing mechanisms than neutral information. First, in the early conflict-monitoring stage, emotional information altered arousal, and more attentional resources were allocated to semantic processing. Second, in the late evaluation stage, the lack of attentional resources (due to prior allocation) reduced the late-stage evaluation of the target stimuli by participants. Thus, in this post-truth era, attentional resources may be exhausted by processing emotional information in unnecessary media cues irrelevant to facts, inducing early cognitive conflict and prolonged late-stage evaluation of news articles.
History of scholarship and learning. The humanities, Social Sciences
Abstract Much has been made in recent years of the transformative potential of digital resources and historical data for historical research. Historians seem to be flooded with retro-digitized and born-digital materials and tend to take these for granted, grateful for the opportunities they afford. In a research environment that increasingly privileges what is available online, the questions of why, where, and how we can access what we can access, and how it affects historical research have become ever more urgent. This article proposes a framework through which to contextualize the politics of (digital) heritage preservation, and a model to analyse its most important political dimensions, drawing upon literature from the digital humanities and history as well as archival, library, and information science. The first part will outline the global dimensions of the politics of digital cultural heritage, focusing on developments between and within the Global North and South, framed within the broader context of the politics of heritage and its preservation. The second part surveys the history and current state of digitization and offers a structured analysis of the process of digitization and its political dimensions. Choices and decisions about selection for digitization, how to catalogue, classify, and what metadata to add are all political in nature and have political consequences, and the same is true for access. The article concludes with several recommendations and a plea to acknowledge the importance of digital cataloguing in enabling access to the global human record.
Learning curves are a concept from social sciences that has been adopted in the context of machine learning to assess the performance of a learning algorithm with respect to a certain resource, e.g., the number of training examples or the number of training iterations. Learning curves have important applications in several machine learning contexts, most notably in data acquisition, early stopping of model training, and model selection. For instance, learning curves can be used to model the performance of the combination of an algorithm and its hyperparameter configuration, providing insights into their potential suitability at an early stage and often expediting the algorithm selection process. Various learning curve models have been proposed to use learning curves for decision making. Some of these models answer the binary decision question of whether a given algorithm at a certain budget will outperform a certain reference performance, whereas more complex models predict the entire learning curve of an algorithm. We contribute a framework that categorises learning curve approaches using three criteria: the decision-making situation they address, the intrinsic learning curve question they answer and the type of resources they use. We survey papers from the literature and classify them into this framework.
Phosphorus (P) is considered to be one of the key elements for life, making it an important element to look for in the abundance analysis of spectra of stellar systems. Yet, there exists only a handful of spectroscopic studies to estimate the P abundances and investigate its trend across a range of metallicities. We have observed full HK band spectra at a spectral resolving power of R=45,000 with IGRINS instrument. Abundances are determined using SME in combination with 1D MARCS stellar atmosphere models. The investigated sample of stars have reliable stellar parameters estimated using optical FIES spectra (GILD; Jönsson et al. in prep.). In order to determine the P abundances from the 16482.92 Angstrom P line, we take special care of the CO($ν=7-4$) blend. We determine the C, N, O abundances from atomic carbon and a range of non-blended molecular lines (CO, CN, OH) which are aplenty in the H band region of K giant stars, assuring an appropriate modelling of the blending CO($ν=7-4$) line. We present [P/Fe] vs [Fe/H] trend for 38 K giant stars in the metallicity range of -1.2 dex $<$ [Fe/H] $<$ 0.4 dex. We find that our trend matches well with the compiled literature sample of prominently dwarf stars and limited number of giant stars. Our trend is found to be higher by $\sim$ 0.05 - 0.1 dex compared to the theoretical chemical evolution trend in Cescutti et al. 2012 resulting from core collapse supernova (type II) of massive stars with the P yields from Kobayashi et al. (2006) arbitrarily increased by a factor of 2.75. Thus the enhancement factor might need to be $\sim$ 0.05 - 0.1 dex higher to match our trend. We also find an empirically determined primary behaviour for phosphorus. Furthermore, the phosphorus abundance is found to be elevated by $\sim$ 0.6 - 0.9 dex in two metal poor s-enriched stars compared to the theoretical chemical evolution trend.
The advent of federated learning has facilitated large-scale data exchange amongst machine learning models while maintaining privacy. Despite its brief history, federated learning is rapidly evolving to make wider use more practical. One of the most significant advancements in this domain is the incorporation of transfer learning into federated learning, which overcomes fundamental constraints of primary federated learning, particularly in terms of security. This chapter performs a comprehensive survey on the intersection of federated and transfer learning from a security point of view. The main goal of this study is to uncover potential vulnerabilities and defense mechanisms that might compromise the privacy and performance of systems that use federated and transfer learning.
یدور هذا البحث حول (الوعی بتاریخ الیونان القدیم فی الشعر الجاهلی –ذو القرنین أنموذجاً-) وتکمن أهمیة هذا البحث فی إبراز جوانب من فکر الإِنسان الجاهلی وإظهار مدى وعی الشعراء الجاهلیین ومعرفتهم بتاریخ من جاورهم من أوائل الأمم القدیمة، وتقصی المرجعیات والموروثات التی أفادوا منها فی التعبیر عن أفکارهم وأغراضهم الشعریة بتضمینها تلک المرجعیات والموروثات بصیغ فنیة وجمالیة، وذلک من خلال إحصاء الإشارات التی أشار إلیه شعراء الجاهلیة إلى شخصیة ذی القرنین فی أشعارهم وذلک عن طریق استقراء عدد غیر قلیل من دواوین الشعر الجاهلی، وإظهار کیف تمکن أُولئک الشعراء بعبقریتهم الفنیة من توظیف ما عرفوه وورثوه سواء أکانت تلک المعرفة دینیة أم تاریخیة، فضلاً عما یتداخل فیها من قصص وأساطیر تدخل فی عالم الخیال.
History of scholarship and learning. The humanities
Open Ended Learning Team, Adam Stooke, Anuj Mahajan
et al.
In this work we create agents that can perform well beyond a single, individual task, that exhibit much wider generalisation of behaviour to a massive, rich space of challenges. We define a universe of tasks within an environment domain and demonstrate the ability to train agents that are generally capable across this vast space and beyond. The environment is natively multi-agent, spanning the continuum of competitive, cooperative, and independent games, which are situated within procedurally generated physical 3D worlds. The resulting space is exceptionally diverse in terms of the challenges posed to agents, and as such, even measuring the learning progress of an agent is an open research problem. We propose an iterative notion of improvement between successive generations of agents, rather than seeking to maximise a singular objective, allowing us to quantify progress despite tasks being incomparable in terms of achievable rewards. We show that through constructing an open-ended learning process, which dynamically changes the training task distributions and training objectives such that the agent never stops learning, we achieve consistent learning of new behaviours. The resulting agent is able to score reward in every one of our humanly solvable evaluation levels, with behaviour generalising to many held-out points in the universe of tasks. Examples of this zero-shot generalisation include good performance on Hide and Seek, Capture the Flag, and Tag. Through analysis and hand-authored probe tasks we characterise the behaviour of our agent, and find interesting emergent heuristic behaviours such as trial-and-error experimentation, simple tool use, option switching, and cooperation. Finally, we demonstrate that the general capabilities of this agent could unlock larger scale transfer of behaviour through cheap finetuning.
In South and Central Kalimantan (southern Borneo) there are some unusual linguistic features shared among languages which are adjacent but do not belong to the same genetic linguistic subgroups. These languages are predominantly Banjar Malay (a Malayic language), Ngaju (a West Barito language), and Ma’anyan (a Southeast Barito language). The same features also appear to some degree in Malagasy, a Southeast Barito language in East Africa. The shared linguistic features are the following ones: a grammaticalized form of the originally Malay noun buah ‘fruit’ expressing affectedness, nasal spreading in which N- not only nasalizes the onset of the first syllable but also a *y in the next syllable, a non-volitional marker derived from the Banjar Malay prefix combination ta-pa- (related to Indonesian tər- + pər-), and the change from Proto Malayo-Polynesian *s to h (or Malagasy Ø). These features have their origins in the various members of the language configuration outlined above and form a Sprachbund or “Linguistic Area”. The concept of Linguistic Area is weak and difficult to define. Lyle Campbell (2002) considers it little else than borrowing or diffusion and writes it off as “no more than [a] post hoc attempt [...] to impose geographical order on varied conglomerations of [...] borrowings”. While mindful of its shortcomings, the current author still uses the concept as a useful tool to distinguish betweeninherited and borrowed commonalities. In the configuration of languages currently under discussion it also provides a better understanding of the linguistic situation in South Borneo at a time prior to the Malagasy migrations to East Africa (some thirteen centuries ago).
History of scholarship and learning. The humanities