Search for: All records

Online reinforcement learning (RL) enhances policies through direct interactions with the environment, but faces challenges related to sample efficiency. In contrast, offline RL leverages extensive pre-collected data to learn policies, but often produces suboptimal results due to limited data coverage. Recent efforts integrate offline and online RL in order to harness the advantages of both approaches. However, effectively combining online and offline RL remains challenging due to issues that include catastrophic forgetting, lack of robustness to data quality and limited sample efficiency in data utilization. In an effort to address these challenges, we introduce A3RL, which incorporates a novel confidence aware Active Advantage Aligned (A3) sampling strategy that dynamically prioritizes data aligned with the policy's evolving needs from both online and offline sources, optimizing policy improvement. Moreover, we provide theoretical insights into the effectiveness of our active sampling strategy and conduct diverse empirical experiments and ablation studies, demonstrating that our method outperforms competing online RL techniques that leverage offline data. Our code will be publicly available at:this https URL. 

While training models and labeling data are resource-intensive, a wealth of pre-trained models and unlabeled data exists. To effectively utilize these resources, we present an approach to actively select pre-trained models while minimizing labeling costs. We frame this as an online contextual active model selection problem: At each round, the learner receives an unlabeled data point as a context. The objective is to adaptively select the best model to make a prediction while limiting label requests. To tackle this problem, we propose CAMS, a contextual active model selection algorithm that relies on two novel components: (1) a contextual model selection mechanism, which leverages context information to make informed decisions about which model is likely to perform best for a given context, and (2) an active query component, which strategically chooses when to request labels for data points, minimizing the overall labeling cost. We provide rigorous theoretical analysis for the regret and query complexity under both adversarial and stochastic settings. Furthermore, we demonstrate the effectiveness of our algorithm on a diverse collection of benchmark classification tasks. Notably, CAMS requires substantially less labeling effort (less than 10%) compared to existing methods on CIFAR10 and DRIFT benchmarks, while achieving similar or better accuracy. 

Identification of cryptic species often relies on invasive techniques such as comparison of cranial morphology or generation of DNA sequences. Myotis lucifugus and M. septentrionalis recently have been reported to occur near the Texas border in Oklahoma and Louisiana, respectively, and due to similarity of appearance, both species easily could be mistaken for M. austroriparius, a common inhabitant of East Texas. All three species co-occur across much of the southeastern United States. Myotis septentrionalis recently was listed under the Endangered Species Act as Endangered, and M. lucifugus has seen drastic reductions in abundance due to white-nose syndrome. Therefore, special care is needed when capturing any of these species due to the cryptic nature of their external morphology and the potential for misidentification in the field. The objective of this study was to determine if wing measurements obtained in the field could be used to differentiate among these three species. Measurements of 13 wing elements from 45 museum specimens were compared using univariate and multivariate statistics. Significant multivariate differences among species were detected, indicating that some wing characteristics may be effective for differentiation. These wing characteristics were compiled into a dichotomous key that researchers can use to easily identify species in the field. Using this technique, non-target species can be released quickly without harm, whereas individuals of species of interest can be confidently collected for scientific research. 

Introduction: Morphological abnormalities and their underlying causes are well documented and understood in humans and some domesticated animals but are less often reported within wild populations. This is likely because many abnormalities lead to the early death of individuals and typically are only encountered within large samples of specimens. Here, we report an adult female great fruit-eating bat Artibeus lituratus (Chiroptera, Phyllostomidae) collected from Paraguay with notable skull and post-cranial abnormalities. Methods: Specimens, including the atypical A. lituratus, were collected from eastern Paraguay and prepared as skin and skull museum voucher specimens and deposited in the Natural Science Research Laboratory of the Museum of Texas Tech University. Sequences of Cytochrome b (Cyt b) were generated and uploaded to BLAST to confirm species identification. We then quantitatively compared the atypical individual with 15 typical females collected from the same locality using 13 wing measurements and 13 skull measurements. Results: The Cyt b sequence of the atypical specimen was a 100 % match to A. lituratus. The atypical specimen was much smaller from the perspective of overall body size and wing morphology. The skull was also qualitatively different, much smaller and less robust than other female A. lituratus from this site. Mastoids and the sagittal crest were greatly reduced, and the frontal shield was absent in the atypical individual. Discussion and conclusions: We encourage reports of morphological abnormalities to be made as determining rates of abnormalities within populations may indicate their overall health. Key words: Morphological deformity; morphometrics; New World bats; skeletal deformity; skull morphology; wing morphology. 

Continuous phenotypic variation reflecting geographic clines can be difficult to distinguish from subspecific discontinuities when specimens are few and heterogeneously distributed. Nonetheless, increases in the holdings of museum collections over the last decades contribute to our ability to resolve more fine-scaled phenotypic gradients for many species. Although the Big-eared woolly bat is not commonly encountered and thus poorly represented in museum collections, sufficient numbers have accumulated to allow an assessment of sub-specific, sexual, spatial and environmental components of phenotypic variation. I examined 123 specimens from across the distribution of Chrotopterus auritus and characterized phenotype based on external, cranial and mandibular characteristics and decomposed variation into components based on univariate and multivariate statistical analyses. All components accounted for significant phenotypic variation. Nonetheless, when examined together and after accounting for correlated variation among components, only sexual, spatial and environmental components accounted for significant unique variation. This, combined with the observation that all qualitative characteristics used to define subspecies of C. auritus can be observed throughout its geographic range, suggest that phenotypic variation is clinal and not characterized by discontinuities reflective of subspecies. Clinal variation was most related to temperature and its seasonality highlighting the important role that these climatic characteristics play in many aspects of the biology of Phyllostomidae. 

Guidelines identifying best practices for harvesting tissues that lead to optimal DNA preservation are few but are important curatorial concerns for genetic resource collections. We conducted a temporal study to establish rate of DNA degradation of tissue samples extracted from field-caught museum specimens. Five individuals of Sigmodon hispidus were collected and their liver and muscle tissues were harvested. Each tissue type was sectioned into 15 subsamples, and each was preserved in liquid nitrogen at different time intervals (2, 4, 8, 16 and 32 minutes; 1, 2, 4, 8 and 16 hours; and 1, 2, 4, 8 and 16 days) following death. DNA was extracted using an automated robotic instrument and molecular mass profiles were determined fluorometrically. Post-mortem DNA degradation was continuous and dependent on time, but also was significantly affected by differences among individual cotton rats. DNA fragments of ≥10,000 base pair in length were present in muscle samples across all time intervals, whereas DNA fragments of this size in liver samples were no longer present after 8 to 16 hours post-mortem. DNA molecular mass profiles showed that muscle samples retained 80% of their longest fragments (≥10,000 bp) until 1 day post-mortem, whereas liver samples retained the same percentage only until 8 minutes after death. Although rates of decay were measured from samples in a laboratory (not field) setting, rates of decay presented here can guide field and museum workers in best practices. Results suggest that opportunistic samples, such as those from roadkill specimens, are more likely to be of use for a variety of molecular methods when muscle is preserved. Considerations of differences in rates of degradation may also guide selection of tissue types housed in genetic resource collections, especially under space-limited circumstances. 

Phenotypic variability is ubiquitous. This is especially true in bats where families such as Phyllostomidae encompass as much phenotypic variability as some entire orders of mammals. Typically, phenotypic variability is characterized based on cranial morphology with studies of other functionally important aspects of the phenotype such as legs, feet and wings less frequent. We examined patterns of secondary-sexual dimorphism and allometry of wing elements of the fringed fruit-eating bat (Artibeus fimbriatus) as well as examined for the first time modularity of bat wings. Patterns were based on 13 wing measurements taken from 21 female and 15 males from eastern Paraguay. From a multivariate perspective A. fimbriatus exhibited significant secondary-sexual dimorphism. Females were larger than males for all 13 wing characteristics with significant differences involving the last phalanx of the 4th and 5th digits. Female wings were also relatively larger than male wings from a multivariate perspective as well as the last phalanx of the 4th and 5th digit, after adjusting for wing size based on forearm length. Wing elements were highly variable regarding allometric relationships with some exhibiting no allometric patterns, and others exhibiting isometry or hyperallometry depending on the element. Wings exhibited significant modularity with metacarpals, proximal phalanges and distal phalanges each representing a discrete module. Wings of A. fimbriatus exhibit substantive patterns of dimorphism, allometry and modularity. While the Big Mother Hypothesis is a strong theoretical construct to explain wing dimorphism, there is yet no sound theoretical basis to patterns of allometry and modularity of the wing. Indeed, trying to understand the determinants of variation in wing morphology is ripe for future investigation. 

We report a case drawn from 6 months of ethnographic and interview research in a STEAM Lab. Free to choose what to work on, students acquired difficult skills and demonstrated what school leaders had identified as valued capacities: collaboration, creativity, and persistence. The teacher struggled, however, to apply conventional grading practices to recognize this learning. We analyze the case of Kira, a 7th grader who learned to design and 3D print an original fidget spinner and began to imagine herself as a designer and entrepreneur. Kira’s story came to a surprising and sad end, as she failed to produce the “evidence of learning” required for a high grade. The case highlights the dilemmas of recognizing and assessing consequential learning experiences in project-based environments. 

The problem of sustaining and spreading educational innovations is one that has vexed many researchers. The flipside of this question, equally important, is what leads to the ‘death’ of educational innovations? Here, to shed light on this question, we provide an autopsy on the death of one local implementation of an otherwise successful STEAM exploration program called FUSE.