University at Albany, State University of New York
University at Albany, State University of New York (SUNY): Scholars ArchiveNot a member yet
6197 research outputs found
Sort by
Datum–Wise Learning and Inference for Supervised Classification
No full textTraditional supervised classification typically involves assigning a label to a single variable, considering a common subset of features for all the instances, and employing a single classifier. However, in many real–world applications like behavioral analysis or insurance recommendation, a data instance is described by a set of related variables such as physical activity and emotion, or driving quality and accident severity. At the same time, these variables are not directly observable but can be inferred via noisy but costly features. Specifically, access to all features is prohibitive due to cost, invasiveness, or limited resources. Finally, while using a single classifier may be suitable for some instances, others may require the use of one or more simple or advanced pre–trained models to be correctly predicted. To address the above challenges, this thesis presents mathematical frameworks and algorithms for datum–wise learning and inference in the context of supervised classification.
First, a framework is proposed to classify related and unobservable variables while keeping the feature acquisition cost to a minimum. Specifically, the relationships between variables are modeled by a known Bayesian network structure. The proposed framework dynamically acquires features in a sequential manner and reaches a classification decision for each variable using a subset of the acquired features. Classification decisions are propagated through the network by exploiting the Bayesian network properties. The proposed framework outperforms existing classification and feature acquisition methods in terms of accuracy and the average number of acquired features. Unfortunately, in many cases, the structure of the Bayesian network is unknown. To address this challenge, a method is proposed that sequentially evaluates variable relationships until it makes a specific decision about the underlying Bayesian network structure. The proposed method speeds–up variable relationship identification without compromising accuracy.
Second, an instance–wise feature acquisition and classifier selection mathematical framework is presented to further enhance the accuracy. The proposed framework sequentially acquires features until it determines that additional features will not improve label assignment. At that stage, easy–to–classify instances are handled by a simple classifier, while difficult–to–classify instances are forwarded to one out of a number of advanced classifiers. This improves overall accuracy while reducing the average number of acquired features. Further, the proposed framework is extended to classify variables related via a known Bayesian network structure.
Finally, inspired by the success of ensemble learning, a mathematical framework that considers the more general problem of determining both the features and the expert decisions to be acquired in a datum–wise fashion is proposed. The framework initially acquires features one by one, until the feature acquisition process terminates. Then it either assigns a label or switches to acquire expert decisions followed by the label assignment. The proposed framework further enhances accuracy compared to existing methods acquiring fewer features and expert decisions on average
The Contextual Process of Bystander Intervention in Bias-Motivated Violent Victimization: An Experimental Approach
Get PDFHate crimes against Asian Americans and LGBT individuals have been on the rise within the U.S. in the past few years (UCR, 2024). Although these offenses tend to occur in public, numerous stories in the media highlight that few bystanders choose to intervene. To unveil why bystanders choose not to intervene, this dissertation explored the extent to which 1,001 U.S. adults engaged with the five-step situational model for bystander intervention proposed by Latané and Darley (1970). Respondents were recruited through Qualtrics and randomly assigned to a hate scenario involving one of six victims: an Asian American man, an Asian American woman, a gay man, a lesbian, a trans man, and a trans woman. The hate scenario escalates such that it begins with a microaggression that then leads to a hate incident before culminating in a hate crime. Study 1 examined whether there was a relationship between increased incident severity as the hate scenario progresses and increased respondent engagement with the situational model. Bystander heterogeneity was also explored through a latent class analysis of bystander progression through each phase of the hate scenario. The latent classes of bystanders then served as one of the dependent variables for the next two studies. Study 2 examined whether bystander progress through the situational model differed based upon the victim’s identity. Study three tested whether bystander characteristics such as empathy, bystander efficacy, and regard for others predict bystander latent class.
Several key findings emerged from these three studies. First, bystander progress through the situational model indeed increases as the hate event escalates. This relationship was observed through examining whether the bystander saw a problem, whether the bystander intervened, and the number of steps the bystander proceeded through in the situational model. Second, three latent classes of bystanders were found in this sample: always interveners, those who progressed further through the model as the event escalated, and those who never intervened. Third, situational model progress significantly differed based on the identity of the victim. This most prominently occurred with subjects being significantly more likely to see a problem and intervene for the Asian woman victim than the sexual minority and gender minority victims. Lastly, the bystander’s favorability towards the victim’s group was consistently the strongest predictor of whether the bystander saw a problem or intervened followed by empathy, bystander behavior, and bystander efficacy. These results together will help to inform the applicability of the situational model to hate as well as the cognitive mechanisms that can be leveraged to enhance the probability of intervention
An Investigation of Food System Localization Efforts in New York Municipalities: Projects, Practices and Policies, Survey Report
Get PDFThis document is a component of a research project funded by the New York State Health Foundation from January 2022 to May 2023. The overall project goal was to better understand how we design and sustain resilient local food systems in New York from the perspective of elected officials. The project used three different data collection methods, interviews, a survey and spatial analysis. This document reports on the Survey component of the project
Implications of 5-Methylcytosine (m5C) and N4-Actylcytidine (ac4C) in Epithelial to Mesenchymal Transition (EMT) in Human Breast Cancer
Get PDFBreast cancer is a complex disease that continues to haunt women and their loved ones around the world. It is currently the leading cause of cancer-related mortality among women. While most breast cancer patients can catch the disease before progression, others are less fortunate. The prognosis of those whose disease has advanced to metastatic or invasive breast cancer (IBC) is grim. Metastasis occurs when cancer cells undergo epithelial to mesenchymal transition (EMT), migrate to distant body tissues, then form new tumors. It is imperative to better understand the process of EMT to develop superior ways to detect breast cancer before becoming metastatic. Recently, post-transcriptional modifications have been implicated in the regulation of EMT in different cancer types. Previous studies lead us to believe that N4-actylcytidine may be involved in the process of EMT in human breast cancer. N4-actylcytidine (ac4C) is produced by the enzyme N-acetyltransferase 10 (NAT10). To establish an understanding of ac4C’s role, we conducted several experiments while altering the expression of NAT10 in Human Mammary Epithelial (HMLE) cells. In this study, we show that the overexpression of NAT10 in HMLE cells induces certain characteristics of the mesenchymal phenotype. We also confirm a proposed method of ac4C-seq by confirming one N4-acetylcytidine site (C-1842) on the 18S ribosome of HMLE cells. These results help us to better understand the role of ac4C in the process of EMT in human breast cancer which may help to pave the way for better preventative interventions and treatments for patients with metastatic breast cancer
Advancements in Forensic Analysis: Development of Mass Spectrometric and Chemometric Approaches for the Identification of Synthetic New Psychoactive Substances and Plant Materials
Get PDFWhile forensic science is a well-established discipline, a number of federal agencies have highlighted challenges that continue to plague the field and the extent to which these challenges remain unaddressed. Examples are the illegal trade of wildlife timber and the drug epidemic. Characterization of these materials requires nuanced method development for compound determination, matrix material-specific protocols, and heavy use of expensive consumables. The application of a technique such as direct analysis in real time – high-resolution mass spectrometry (DART-HRMS) provides the opportunity to circumvent many of the challenges presented by conventional methods. In general, little to no sample preparation is required and a consistent sample analysis approach can be applied to most samples. This work explored the development and application of DART-HRMS through the investigation of the identification of New Psychoactive Substances (NPSs), psychoactive plants, and trade-regulated timber. The procedure for rapid structure determination of NPSs combines neutral loss mass spectral information from DART-HRMS data acquired at multiple voltages under collision-induced dissociation (CID) conditions, thereby resulting in varying levels of molecule fragmentation. This approach falls under the umbrella of “data fusion”, which is a strategy that combines the output from multiple data sets in order to improve the accuracy of the results. A second focus on psychoactive substances is the development of the Database of Psychoactive Plants (DoPP). This tool is designed to be user-friendly and includes an architecture for identifying plant unknowns. The application is based on the observation that plants display specific chemical signatures that are detectable by DART-HRMS. The subsequent automated machine learning processing of libraries of these spectra enabled the rapid discrimination and identification of species, resulting in a chemical signature database containing 57 available plant species. Another focus of this work is the development of an analysis approach to be used in a wildlife forensics context. Depending on the species, trade in timber can be totally or heavily restricted. A current technique used by law enforcement to differentiate species of wood is DART-HRMS, coupled with multivariate statistical analysis. Although this method is useful in a laboratory setting, it is impractical in field applications (such as for the determination of timber species identity in shipping containers at ports). The added dimension of wood headspace analysis by solid phase microextraction (SPME) was used to generate data to complement that acquired using the conventional wood analysis technique to facilitate the development of “stand-off” approaches for the differentiation of wood species based on their volatiles profiles
Equipping Middle School Teachers with Culturally Responsive Pedagogy in Teaching Computer Science through Professional Learning: A Longitudinal Case Study
Get PDFCulturally responsive pedagogy (CRP) is an effective way to foster equitable and meaningful computer science (CS) learning experience for historically marginalized students. This dissertation study employed longitudinal data to investigate five middle school teachers’ development of CRP understandings and enactments over many years participating in the CS Pathways Research Practitioner Partnership (RPP) project. Grounded in conceptual change theory (knowledge-in-pieces perspective), the study explored the five teachers’ CRP conceptual change trajectories as indicated by their definitions and enactments of the pedagogy over time. From the epistemological perspective, the findings presented how teachers reflected, refined, and revised their naïve knowledge of CRP into potentially expert-like CRP knowledge to teach CS under their context-dependent environment. In addition, the study examined teachers’ perceptions of the project’s various support to their CRP conceptual development. The results documented each teacher’s knowledge growth trajectory and implementation pathways of CRP to teach CS. As teachers progressed through the program, they all developed awareness of racial disparities within CS education. Furthermore, some teachers exhibited deficit model of beliefs indicating the socio-economic impacts on students’ CS learning. However, their expanded awareness did not correspond to any direct curriculum enactment changes. The study also concluded that teachers perceived the project professional development support as a positive impact on their knowledge growth and classroom implementation. This dissertation study provided empirical evidence and understanding on how teachers’ equity-focused pedagogical knowledge developed. It can inform the design of future professional learning programs to enhance teachers’ understanding and practices of CRP to teach CS
Decoding Molecular Recognition Mechanism Of Small Molecule Ligand-Rna Binding Pocket Systems Using Molecular Dynamics And Metadynamics
Get PDFFrom my PhD research, under Amber99-Chen-Garcia force field, umbrella sampling and MD simulation helped us found that the 2’-5’ phosphodiester RNA backbone linkage modification on in vitro selected neomycin RNA aptamer can transfer it to ‘amikacin RNA aptamer’ by changing the conformation and backbone rigidity of the RNA binding pocket, which further influenced the binding mode and binding pathway of amikacin. Furthermore, under the same force field, metadynamics was found more powerful for simulating the binding pathway, binding mechanism and predicting binding modes of small molecule ligand recognizing its RNA target by filling up all energy basins on free energy surface of the binding pathway with extra Gaussian potential, which was well-trained on five small molecule/RNA binding pocket systems. This metadynamics prediction method training study also discovered a theory of ‘conformational-selection model of small molecule binding with different degree of induced-fit effect on RNA target based on the rigidity of small molecule ligand’ for the binding model of small molecule ligand/RNA binding pocket recognition system. Furthermore, this well-trained metadynamics simulation prediction method was firstly applied to help us found that 2’-5’ linkage modifications on FMN RNA aptamer narrowed the width of the frontside opening of RNA binding pocket, at the same time, the backside opening became wider, which changed the binding pathway and binding mode of FMN to ‘upside down’. Secondly, this well-trained metadynamics simulation prediction method helped us refining the NMR binding mode of small molecule splicing modifier SMN-C5/RNA duplex and further predicted a new mechanism of splicing modifier SMN-C5 promoting SMN2 exon 7 inclusion for treating spinal muscular atrophy (SMA). In sum, Amber99-Chen-Garcia force field showed its power and potential of capturing RNA-small molecule ligand interactions correctly. Umbrella sampling showed its power and potential of simulating unbinding events and calculating ΔG bind/unbind for RNA-small molecule ligand complex correctly. With a good force field of Amber99-Chen-Garcia and good CVs selection, metadynamics showed its power and potential of predicting binding modes, binding pathway and binding mechanism of RNA-small molecule ligand recognition correctly
Comprehensive Examination Of The Effect Of A Randomized Controlled Brief Mindfulness-Based Intervention On Autonomic Nervous System Functioning
Get PDFBrief mindfulness-based interventions are comparable to traditional mindfulness-based interventions in alleviating stress and improving well-being. To explore the underlying biological mechanisms of these practices, researchers have been exploring the effects of mindfulness-based interventions through the autonomic nervous system (ANS), a major bodily stress response system. Nevertheless, most of these studies relied on one stress response system or a specific biomarker. This study aims to comprehensively examine the effectiveness of two sessions of brief mindfulness-based interventions on ANS functioning. Participants were assigned to the brief mindfulness intervention condition (n = 44) or control condition (n = 47). Participants’ physiological data, including respiratory sinus arrhythmia (RSA), skin conductance levels (SCL), blood pressure, and total peripheral resistance (TPR), were collected in response to the assigned condition and a psychosocial stressor. Compared to the control group, the mindfulness group displayed greater RSA augmentation and lower SCL activity in response to the mindfulness intervention, greater decline in blood pressure from preintervention to post-mindfulness intervention, and higher blood pressure immediately after psychosocial stress. Moreover, the assigned condition significantly influenced ANS coordination. In face of psychosocial stress, participants in the mindfulness condition displayed reciprocal ANS activation patterns, whereas control participants exhibited non-reciprocal ANS coordination. Greater RSA augmentation, reduced SCL and blood pressure, and reciprocal ANS activation are considered to confer adaptive stress response and have been shown to support health and socioemotional outcomes. The results of this study highlight the promise of two sessions of brief mindfulness-based interventions in promoting optimal physiological functioning
The Complexities Of Adverse Childhood Experiences: A Study Of Health, Behavioral Health, And Quality Of Life Among Women With And Without Disabilities
Get PDFBackground and Purpose: The negative effects of adverse childhood experiences (ACEs) on behavioral and physical health and well-being long after exposure are a global and national concern. For over two decades, ACEs have been associated with numerous medical conditions, health risk behaviors, and poor mental and physical health outcomes. However, few studies have examined the effects of ACEs on the health outcomes of persons with disabilities (PWDs). The main objective of this study was to examine the impact of ACEs on women with and without disabilities, their health and well-being