Over fourteen months of field research studying manufacturing firms in South Asia - spread across three industries in fifteen cities in India and Pakistan - uncovered a remarkably consistent pattern of diversity in firm-level industrial strategy. Such a pattern cut across states and provinces, industries and national boundaries, thus presenting a challenge to state-centric explanations of firm behavior. Through interviews with more than two hundred firms, two categories of strategy emerged in every research site and in every sector. Some firms exhibited paternalistic relationships with their workers, conservative modes of financing and tended to see the state as a protector of indigenously developed norms. Other firms deployed professional institutions for the management of workers, liberal modes of financing including high levels of leverage, and tended to see that state as a promoter of international norms.

I argue that instead of strategies being derived as responses to government policy or market conditions, manufacturers in South Asia formulate relationships with other actors in the economy based on their perspective on the character of the economy and the means and ends of industrial production. Those who perceive the economy through technocratic perspectives tend to see the economy as a space ruled by universal norms and practices, and thus establish more systematic relationships with workers, finance and the state. Those who maintain embedded perspectives tend to see the economy as constituted by dense networks of individual relationships, and thus establish more personalistic relationships with workers and the state while building up finance internally. These perspectives often endure within firms over time, as second- and third-generation manufacturers follow the guidelines of the firm's founder. They do not correlate with a particular size of firm, industry or location, but do however correlate with the education and work experience of manufacturers and the strength of institutions within the firm over time.

Such cleavages in industrial strategy and manufacturers' perspectives reflect a broader division within South Asian society of the means and ends of economic development, one that was distilled and legitimated during national resistance against imperial rule. Some among the nationalist movement, who forwarded a vision of development based on socialistic economic planning, saw development as a rationalistic process of industrial catch-up, with the state as the central actor in disciplining and transforming society. Others held that development lay in the rejection of western norms and values, and a return to a mythical pre-colonial society of traditional hierarchies and economic relations submerged in moral frameworks, where the state would serve as a protector of this organic society. These positions in the nationalist movement both reflected and consolidated divisions among indigenous society, divisions that are operative more than six decades later in the technocratic and embedded perspectives among manufacturers in contemporary South Asia. The continued existence of such dual modes of capitalist strategy and practice also represents a portrait of the South Asian state that - far from disciplining society into a unitary institutional framework or being prevented from doing so by capture - actively maintains and supports multiple modes of doing business. Based on the study of Indian regulatory documents, I suggest that this is because the state tries to forward several different developmental interests, from increasing exports to creating employment, simultaneously, rather than maintaining a singular focus on maximizing economic growth.

Why do some political families flourish, and others decline quickly? While the presence of political families has been a long-standing feature of many countries, crucial questions about the nature of political succession and the underlying conditions that make it possible remain under-studied. This dissertation focuses on the aspects of the local political economy, and networks that control it, to propose a theory of dynastic perpetuation in India and elsewhere. Specifically, I argue that the diversification of economic resources and political power are key factors that determine the strength and longevity of political families. To test this theory, I map the familial connections of the winners and the runners-up from all elections held in Uttar Pradesh (India) between 1974 to 2019. The analysis shows that politicians from dominant castes and those with greater economic endowment are more likely to form successful political families. In addition to this, I take a closer look at dynamism within political families by classifying them into three distinct groups – families that have been in power for a long period of time, the old political families that have declined in power and status, and new and rising political families. I show that families that successfully diversify their political and economic portfolios are more likely to perpetuate themselves. This diversification strategy, which in part comes from the political family’s access to local political economy networks, allows them to project influence even during electorally leaner periods. While most successful political families remain largely confined to their local constituencies, I show how some families manage to scale up their network and influence, while retaining their local power base, through a case study. Finally, I suggest that understanding the rise and decline of political families is key to reflecting on the changing nature of representative democracy in India and elsewhere. And as long as the local state remains beholden to the control of elite networks, political families are likely to flourish, under different guises.

Physically based rendering is widely used due to its ability to create compelling, photorealistic images. Generating these images requires evaluating a multidimensional integral and is typically done through Monte Carlo (MC) samples of the scene's light transport. However, these samples are computationally costly and many are needed to obtain a converged result, otherwise there is objectionable noise in the final image.

To address these concerns, there has been substantial work aimed at reducing the variance of the MC estimate. For this dissertation, we observe that this work can be categorized based on where in the rendering pipeline these methods are being applied: pre-rendering, intra-rendering, and post-rendering. The first category refers to methods that perform preprocessing to make subsequent rendering more efficient. Meanwhile, intra-rendering methods will provide feedback during rendering to facilitate convergence of the MC estimate. Finally, post-rendering approaches are used after the sampling budget has been exhausted to output images that better match the ground truth rendering.

Previous methods in these categories are far from optimal as a result of their inaccurate approximations, costly initialization, and fallible heuristics. On the other hand, machine learning approaches in countless applications have seen steady progress in recent years thanks to their ability to model complex relationships with significantly more accuracy and robustness relative to their hand-crafted counterparts. Inspired by these approaches, we propose to apply machine learning to all three portions of the pipeline to avoid the aforementioned pitfalls of prior work in MC variance reduction.

For pre-rendering, we introduce the first deep network for prefiltering that can accurately capture the appearance of large environments with complex geometry and materials for level of detail. Our approach foregoes traditional geometric simplification and volumetric modeling and robustly handles the range of cases.

Within the context of intra-rendering, previous approaches for importance sampling are costly since they require many samples and, in some cases, even additional training per scene before they can be effectively used to guide samples towards paths with higher light contributions. We demonstrate the first offline deep network that can guide sampling for almost the entirety of rendering for an arbitrary scene.

Finally, we bring deep learning to post-rendering MC denoising and replace the heuristics used by previous methods to set the parameters of explicit filters. Our network predicts the filter kernel around each pixel, can even directly filter the image, and works on general scenes and varying levels of noise without retraining.

In this thesis, we demonstrate how our proposed, learning-based solutions offer a significant advantage over prior state-of-the-art methods in each of these cases. We also discuss avenues for future work in the research community to bring machine learning to additional applications within all components of the MC rendering pipeline.

Abstract

Evolutionary constraints on the sequence of Ras

by

Pradeep Bandaru

Doctor of Philosophy in Molecular and Cellular Biology

University of California, Berkeley

Professor John Kuriyan, Chair

Ras proteins are highly conserved signaling molecules that exhibit regulated, nucleotide-dependent switching between an active GTP-bound state that transduces signals by binding to effector proteins, and an inactive GDP-bound state that cannot bind to effectors. The high conservation of Ras requires mechanistic explanation, especially given that proteins are generally robust to mutation, a concept that was first established from early structural and phylogenetic analysis of hemoglobin by Max Perutz and John Kendrew.

During my thesis research, I adapted a two-hybrid selection system to analyze how mutations affect the functional cycle of human H-Ras, with the ultimate goal of understanding the constraints on the sequence of Ras that give rise to its high evolutionary conservation. My strategy was to isolate just the minimal biochemical network that defines this cycle, comprising Ras, its effector Raf, a GTPase accelerating protein (GAP), and a guanine-nucleotide exchange factor (GEF). Using this selection system, I analyzed the sensitivity of every residue in Ras to mutation in the context of this network, while excluding the effects of the membrane and additional regulatory factors. This approach provided an opportunity, for the first time, to use deep mutational scanning approaches to study how local regulatory networks influence the mutational sensitivity and phenotypic plasticity of key signaling molecules.

I found that Ras exhibits global sensitivity to mutation when regulated by a GAP and a GEF, effectively displaying global constraints that result in the majority of mutations leading to a modest decrease in Ras function. In the absence of regulators, Ras shows considerable tolerance to mutation, as seen previously in saturation mutagenesis experiments on other proteins, where the distribution of mutational effects shifted to be largely neutral. Surprisingly, the analysis of Ras in the absence of regulators also revealed allosteric hotspots of activating mutations in residues that restrain Ras dynamics and promote the inactive, GDP-bound state. This showed that structural fold of Ras is intrinsically capable of accommodating sequence changes that, in evolution, could lead to the acquisition of new function, but could also lead to unwanted Ras activation in disease. Indeed, oncogenic mutations that disturb the switching mechanism of Ras result in aberrant signaling and cancer, highlighted by the fact that Ras is one of the most important proto-oncogenes in the human genome.

Altogether, my research shows that the local regulatory network places a stringent constraint on the sequence of Ras, and also creates the potential conditions in which it is susceptible to activating mutations. This extended previous observations that mutational sensitivity in proteins is strongly dependent on the selective conditions in which the protein operates. From a practical perspective, though small molecule inhibitors of Ras have yet to achieve clinical relevance despite a concerted effort to obtain such inhibitors, my research also provides a roadmap of allosteric hotspots of Ras activation that can be exploited to design novel cancer therapeutics.

This dissertation project seeks to understand the political dynamics of ruling parties under strong presidencies at the legislative recruitment stage with a regional focus on South Korea. The first paper introduces my main argument regarding how strong presidencies can determine the form and function of ruling party reforms such as legislative primaries. I argue that under strong presidential systems, wherein the executive abides by formal rules but is still willing to push her de facto power to its constitutional limits, ruling party elites - especially those who do not belong to the president's faction - can counteract presidential discretion by voluntarily democratizing the candidate selection process. I provide empirical evidence of these dynamics from South Korea during 2008-2016. The second paper formalizes the conditions under which ruling party elites endogenously choose to implement legislative primaries as a means to prevent the president from interfering in the party nomination process and provides implications of ruling party reforms on voter welfare. Whereas the first two papers study the contexts under which competing factions within the ruling party are able to agree upon the party's nomination processes and outcomes, the third paper examines a case of a failed coordination. The paper draws upon empirical evidence from the 2008 South Korea Legislative Election, where some party elites chose to defect from the ruling party and successfully formed a transitory electoral alliance.

Cardiovascular diseases including hypertension, arrhythmias, and heart failure are the leading causes of morbidity and mortality in the world. The autonomic nervous system (ANS) regulates the electrical and mechanical function of the heart, and its dysfunction plays an important role in the pathophysiology of cardiovascular diseases, representing an emerging target for therapeutic intervention. However, fundamental gaps in our knowledge regarding autonomic control of the heart in health and the adverse remodeling that occurs in disease have impeded the development of neuromodulation therapies. The cardiac ANS consists of neurons located from the level of the brain to that of the heart itself. At the organ level, the intrinsic cardiac nervous system (ICNS), a distributed network of ganglia and interconnecting nerves, serves as the final common pathway for the integration of neural inputs to the heart. Despite the importance of the ICNS, the impact of acute and chronic stress on the function of this neural network is not well-characterized. Therefore, we investigated the acute effects of premature ventricular contractions (PVCs) and the chronic remodeling induced by myocardial infarction (MI) on the ICNS by obtaining in vivo neuronal recordings from beating hearts. Both PVCs and MI affect a critical population of neurons within the ICNS—local circuit neurons. Given that they receive a convergence of afferent and efferent inputs and have intra- and interganglionic projections, local circuit neurons integrate information and coordinate regional cardiac function. Further, MI results in the formation of a neural sensory border zone, characterized by diminished afferent signals from the infarct and preserved signals from border and remote regions of the heart. These neural effects may underlie, at least in part, the mechanical dyssynchrony, electrical instability, and reflex activation of the ANS noted with PVCs and after MI. In addition, we demonstrated that the evoked cardiac response to vagus nerve stimulation, a neuromodulation therapy being evaluated for cardiac conditions, represents the dynamic interplay between afferent and efferent vagal fibers and reflex responses of the ANS. In conclusion, a better understanding of autonomic regulation of the heart will lead to the development of rationale neuroscience-based therapies for cardiovascular diseases.