6609 research outputs found
Sort by
(To Appear) Towards Traceability of Git Issues and Requirements Through Clustering: Supplemental Information
No full textDes objets nous accompagnent : (ou l\u27inverse)
No full textC’est un livre écrit sur le motif. Parce que tout, autour de nous, nous interroge. Le monde est là énigmatique, à la fois étrange et familier. Un livre de poèmes est-il une réponse ? Il prolonge plutôt l’interrogation, il devient geste du vivre (inquiétude, émerveillements) qui sait trop bien qu’il ne fait qu’errer autour de l’énigme.https://scholarworks.smith.edu/frn_books/1022/thumbnail.jp
Standardizing Canine Breed Data in Veterinary Records Is Challenging, but Computer Vision Offers an Alternative Perspective on Breed Assignment
Full text linkDog breed is fundamental health information, especially in the context of breed-linked diseases. The standard-ization of breed terminology across health records is necessary to leverage the big data revolution for veterinary research. Breed can also inform clinical decision making. However, client-reported breeds vary in their reliability depending on how breed was determined. Surprisingly, research in computer science reports that AI can assign breed to dogs with over 90% accuracy from a photograph. Here, we explore the extent to which current research in AI is relevant to breed assignment or validation in veterinary contexts. This review provides a primer on approaches used in dog breed identification and the datasets used to train models to identify breed. Closely examining these datasets reveals that AI research uses unreliable definitions of breed and therefore does not currently generate predictions relevant in veterinary contexts. We identify issues with the curation of the datasets used to develop these models, which are also likely to depress model performance as evaluated within the field of AI. Therefore, expert curation of datasets that can be used alongside existing algorithms is likely to improve research on this topic in both fields. Such advances will only be possible through collaboration between veterinary experts and computer scientists
The Binding of Cu2+ to Lipid Membranes Is Not Substantially Influenced by Electrostatic Screening
Full text linkGouy-Chapman theory predicts that salt screening and modulating the interfacial charge density should strongly influence the apparent dissociation constant, Kd,app, between Cu2+ and negatively charged phosphatidylserine (PS) lipids in supported lipid bilayers (SLBs). Specifically, Kd,app would be expected to increase (weaken binding) by a factor of 40 when 100 mM NaCl is introduced into the solution because of electrostatic screening between the membrane and Cu2+ cations. Surprisingly, however, fluorescence quenching measurements demonstrate that Kd,app increases by less than a factor of 2 when increasing the salt concentration in the presence of standard buffers, such as tris(hydroxymethyl)aminomethane (Tris). Moreover, increasing the negative surface charge density by a factor of 4 would be predicted to decrease (strengthen binding) Kd,app by 3 orders of magnitude. Instead, Kd,app increases slightly when 15 mol % phosphatidylglycerol (PG), a negatively charged lipid, is introduced into SLBs already containing 5 mol % PS. Such findings indicate that electrostatic double layer theory is not a useful approach for predicting the binding behavior of transition metal cations to negatively charged interfaces. The problem lies with the fact that standard buffers, such as Tris, form a wide variety of coordination complexes in bulk solution with transition metal cations like Cu2+. Typically, dozens of complexes are present simultaneously at any given pH value and the net charge on them ranges from positive to neutral to negative. Such variations in charge on the complexes result in electrostatic screening and interfacial potential effects that are substantially diminished or nonexistent. These results should generally apply to the binding behavior of first row transition metal ions, when the cations predominantly reside in complexes rather than as free ions. This includes in vivo conditions, where the concentration of free transition metal ions is often very low
Rethinking Large-Scale Phylogenomics with EukPhylo v.1.0, a Flexible Toolkit to Enable Phylogeny-Informed Data Curation and Analyses of Diverse Eukaryotic Lineages
Full text linkEukaryotic diversity is largely microbial, with macroscopic lineages (plants, animals, and fungi) nesting among a plethora of diverse protists. Our understanding of the evolutionary relationships among eukaryotes is rapidly advancing through ’omics analyses, but phylogenomic analyses are challenging for microeukaryotes, particularly uncultivable lineages, as single-cell sequencing approaches generate a mixture of sequences from hosts, associated microbiomes, and contaminants. Moreover, many analyses of eukaryotic gene families and phylogenies rely on boutique data sets and methods that are challenging for other research groups to replicate. To address these challenges, we present EukPhylo v.1.0, a modular, user-friendly pipeline that enables effective data curation through phylogeny-informed contamination removal, estimation of homologous gene families (GFs), and generation of both multisequence alignments and gene trees. For the GF assignment, we provide the “Hook Database” of ~15,000 ancient GFs, which users can easily replace with a set of gene families of interest. We demonstrate the power of EukPhylo, including a suite of stand-alone utilities, through phylogenomic analyses of 500 conserved GFs sampled from 1,000 diverse species of eukaryotes, bacteria, and archaea. We show improvements in estimates of the eukaryotic tree of life, recovering clades that are well established in the literature, through successive rounds of curation using the EukPhylo contamination loop. The final trees corroborate numerous hypotheses in the literature (e.g., Opisthokonta, Rhizaria, Amoebozoa) while challenging others (e.g., CRuMs, Obazoa, Diaphoretickes). The flexibility and transparency of EukPhylo set new standards for curation of ’omics data for future studies
The Adoption of the Braided River Model Toward an Inclusive Stem Workforce for All
Full text linkDespite decades of interventions aiming to transform the science, technology, engineering, and mathematics (STEM) workforce to be more inclusive and diverse, little progress has been made in creating long-lasting, sustainable change. For a long period of time, the STEM workforce has been described as a leaky pipeline. While there has been some utility to thinking about the STEM workforce in this way, in this article, we discuss how characterizing the STEM workforce as a leaky pipeline can impede the design of innovative interventions that contribute to sustainable change toward a more inclusive scientific enterprise. As an alternative, we join others in proposing the braided river ecosystem model, related social sciences and career development theories as more inclusive ways to think about the STEM workforce and how a target group or an individual navigates their career choices and development as a scientist. New models and paradigms to understand the STEM workforce and individuals’ careers in science may open the door to finding novel strategies to make careers in STEM accessible to all. We present case studies demonstrating the practical applications of these inclusive models
Integrating AI, Art, and History in Language Learning
No full textIntegrating AI and applying culture “the 3Ps”, (product, practice, and perspective) in language teaching
Episode 15: Collins Aerospace DC1819
Full text linkThis episode features four alums from the Class of 2019: Sarah Chu, Jackie Granillo, Gaea Ridenhour, and Shuying Zhen. Their Design Clinic project with Collins Aerospace focused on design and process improvements for the ram fan shaft heat exchangers in an airplane
Augmenting, Not Replacing: The Role of LLMs in Human-Centric Formal RE: Supplemental Material
No full textThis repository contains the supplemental information for the RE\u2725 paper entitled Augmenting, Not Replacing: The Role of LLMs in Human-Centric Formal RE and the Smith College Departmental Honors Thesis entitled The LTL Whisperer: Prompting AI to Explain Temporal Logic: Supplemental Information . This work investigates how and to what extent generative AI with large language models (LLMs) can assist practitioners and novices in interpreting formal requirements expressed in Linear Temporal Logic (LTL)
Implementing a Contributions Approach to Complement Decarbonization at Smith College
No full textSmith College has expressed interest in exploring alternatives to carbon offsetting to address its remaining greenhouse gas (GHG) emissions. However, a distinct gap remains between the recommendation of such alternatives and their actual implementation. This memo outlines the climate contributions approach as a strategy that aligns with the college’s values and decarbonization goals. It assesses the feasibility of this approach and identifies key questions that must be addressed for its successful execution. The shift toward a contributions model is motivated by growing concerns about the integrity of the carbon offset market. Carbon offsets have been widely criticized by researchers and policy experts for issues including non-additionality, double counting, and impermanence. The variability in carbon credit pricing also makes it easy for institutions to purchase cheap, low-quality credits tied to projects that may not deliver real emissions reductions. In light of these challenges, we recommend that Smith account for its remaining Scope 1 and 3 emissions—after the completion of its geothermal transition—through a climate contributions framework rather than by purchasing offsets. Unlike offsets, which require the purchase of carbon credits on a ton-for-ton basis, the contributions model operates on a money-per-ton framework. Contributions are directed toward climate adaptation, local decarbonization, and environmental justice initiatives—projects which are often excluded from traditional offset schemes. This approach allows Smith to support a broader array of meaningful climate actions while reinforcing its institutional values of justice, sustainability, and community partnership. If implemented, Smith would be the first known higher education institution to both fully adopt and explicitly name a climate contributions model—setting a precedent for peer institutions. This memo recommends structuring the contributions fund through an equitable governance model, such as Participatory Grantmaking or Trust-Based Philanthropy, which share decision-making power between community stakeholders and institutional representatives. A grant advisory committee, selected by the Committee on Sustainability, would oversee project selection and ensure that this process reflects Smith’s values. Financially, the contributions fund would be calculated by multiplying Smith’s remaining emissions by an internal carbon price. Our analysis provides a cost range using a lower-bound price of 115 per ton (Smith’s projected 2030 proxy price). These are illustrative estimates, and any final pricing decision should be flexible while still reflecting a commitment to high-impact, justice-centered climate action. In conclusion, we recommend that Smith College implement a climate contributions approach to address its remaining emissions. This model avoids the pitfalls of the offset market and provides an opportunity for Smith to lead with integrity—prioritizing justice, equity, and sustainability in its path to decarbonization