The study of infectious disease is more than how pathogens transmit between hosts and cause illness and death. Infectious diseases are social, political, environmental, and infrastructural phenomena that humans have limited but equal power to eradicate or to exacerbate. Using modern disease modeling methods, researchers have the ability to test our assumptions about disease dynamics and to explore ways in which our behavior may help or hurt, and to imagine a better world with effective interventions. In this dissertation I focus on three applications of such models to two diseases: an evaluation of vaccine distribution and long-term seasonality of SARS-CoV-2, the pathogen causing the COVID-19 pandemic, and Chagas Disease, a parasitic infection in Brazil that despite its relative infrequency, has the potential to cause debilitating and fatal complications. In all studies, I focus on the heterogeneous risk of transmission and infection present in the population with an eye towards disease-averting interventions.In Chapter 1, I find that distributing vaccines for COVID-19 in decreasing priority by age is a more effective strategy to limit deaths than prioritizing by high-contact employment status, consistent with the CDC’s recommendations in early 2021. In the second chapter, I continue the analysis of SARS-CoV-2 using an Agent-Based Network Simulation, finding that recurrent outbreaks of the disease are likely to persist but, depending on the epidemiology of the pathogen, may be mitigated by annual booster doses distributed at certain times of year. Finally, I conclude in chapter 3 by finding that cases of Acute Chagas Disease may be expected to increase over the ensuing decade, driven by an increasing population in affected areas and changing climate.

Social science discourses on migration and development emphasize migrants' altruism or self-interest as the determinants of their remitting practices, which is countered by alternative explanations focusing on social culture and structure. Due to the dichotomy of structure/agency, the current approaches offer unsatisfactory explanations of migrants' remitting practices. This dissertation aims to proposes a corrective by developing an alternative model of understanding the determinants of migrants' remitting practices. It synthesizes three strands of social science literature and derive the conception of remitting as a collective social act. From a realist approach following Durkheim, it begins at the empirical manifestations of migrants' remitting practices, but looks at deeper levels of social reality (the actual and the real) in identifying the determinants. It focuses on the internal social relations between the individuals involved in the acts in question as well as the external relations between the individuals and their social world to recognize the determinants of acts. Thus, it adopts a mixed-methods approach involving ethnographic fieldworks supplemented by small-scale survey with convenient samples. Following the theoretical case selection strategy of global ethnography, it investigates remitting practices among Bangladeshi migrants in Tokyo and Los Angeles. Holding the origin of the migrants constant allows for isolating the role of their different destinations in determining their remitting practices. This dissertation recognizes social norms in origin culture, advances in family life-cycle, legal status and social incorporation and/or marginalization in the destination as the determinants of migrants' remitting practices. While confirming the centrality of migrant family in remitting practices in the NELM perspective, it goes beyond by including how internal family relations and external social relations collectively determine migrants' remitting. It advances our knowledge of migrants' remitting by incorporating both migrants' agency and social structures embedding remitting practices. Besides generating satisfactory explanations of migrants' remitting, this study demonstrates how to combine agency and structure in migration studies in particular and sociology in general. It also offers valuable insights for development policy-making utilizing migrants' remittances in their origin communities. It concludes with a call for extending the scope of the study by extensive and systematic use of quantitative data on the general patterns of remitting and to include familiar destination countries of temporary migration in future studies.

Near-future prediction in videos has crucial impact on a wide range of practical applications which require anticipatory response. In videos, prediction can be performed in different spaces such as labels, captions and frames. Labels can be predicted for a longer horizon in future but are less informative than frames. Video frames are much richer in content than labels but only a few frames can be predicted ahead. Captions lie in between these two extremes: they can describe changes in activities for a longer prediction horizon and provide a much richer description than labels. In this thesis, we provide three distinct prediction frameworks leveraged upon different computer vision and machine learning techniques. However, these solution methods require lots of labeled data which is challenging due to high annotation cost. Thus, we also propose a novel early prediction framework so that video annotation becomes scalable.

Most of the existing works on labeling human activities focus on the recognition or early recognition problem where complete or partial observations of the activity are available. However, in the prediction problem we are addressing, no observation of the future activity is available beforehand. We propose a system that can infer about the labels and the starting time of a sequence of future unobserved activities combining different context attributes from the observed portion of the video . Next, we propose a sequence-to-sequence learning-based approach using an encoder-decoder LSTM pair for captioning the near-future unobserved activity sequences.

Building upon the prediction framework, we also work on the frame reconstruction problem in a multi-camera scenario. When a camera has multiple missing frames and available frames within the camera are far apart, the corresponding frames from other overlapping cameras become crucial for reconstruction . We propose an adversarial approach using conditional Generative Adversarial Network (cGAN) where the conditional input is the preceding or following frames within the camera or the corresponding frames from other cameras, all of which are merged together using a weighted average. We also propose an adversarial learning solution to the multi-modal frame reconstruction problem where we learn a mapping between 3D LIDAR point clouds and RGB images. This facilitates faster processing since fusion-based approaches which try to combine the advantages from both sources of data consume huge computing resources.

We also consider the video annotation problem, as it crucial for machine learning approaches described above. State-of-the-art video annotation approaches assume that there is no latency for looking up the correct category of label and the annotator is required to watch the whole video segment. However, choosing the correct label from thousands of categories is not instantaneous and the long viewing time adds to the annotation cost. We propose an LSTM-based early prediction framework which can be combined with any existing active learning approach to provide a list of early suggestions to the annotator. This reduces annotation time and cost by a significant margin.