Human activity recognition (HAR) technology that analyzes data acquired from various types of sensing devices, including wearable sensors and vision sensors, is getting considerable attention in the field of Artificial Intelligence (AI) driven healthcare systems. Human activities can be used to provide remote healthcare solutions by identifying particular movements such as falls, gait, and breathing disorders. HAR healthcare system can allow people to live more independent lifestyles and still have the safety of being monitored if more direct care is needed. Thanks to the development of machine learning technology, many machine learning methods have been employed in human activity recognition systems in healthcare. However, this field still faces many technical challenges. Some challenges are shared with other pattern recognition fields, such as a limited number of labeled data, while other challenges are unique to sensor-based activity recognition in healthcare and require dedicated methods for real-life healthcare applications, such as data noise of sensor factors in the data collection process.

In this dissertation, we start with the challenges of healthcare-oriented HAR systems and summarize the challenge-related machine learning approaches. To overview HAR healthcare applications with wearable sensors, we cover essential components of designing HAR healthcare systems, including sensor factors (e.g., type, number, and placement location), AI model selection (e.g., classical machine learning models versus deep learning models), and feature engineering.

Next, we present a new healthcare application of HAR, that is, Early Mobility Activity (EMA) recognition for Intensive Care Unit (ICU) patients, to illustrate the system design of HAR applications for healthcare. We identify insensitive wearable sensor orientation features and propose a segment voting process to improve the model accuracy and stability.

We further apply the state-of-the-art vision sensor-based HAR approaches in healthcare. We present a healthcare system (BWCNN) to use eye blinks to communicate with the outside world for Amyotrophic Lateral Sclerosis (ALS) patients. The system uses a Convolutional Neural Network (CNN) to predict the eyes' state, which is used to find the blinking pattern.

Then, we propose a MASTAF that can quickly learn from a few examples efficiently to solve the limited number of video samples in real-life HAR applications, a common challenge shared with computer vision. MASTAF takes input from a general video spatial and temporal representation,e.g., using 2D CNN, 3D CNN, and video Transformer. Then, to make the most of such representations, we use self- and cross-attention models to highlight the critical spatio-temporal region to increase the inter-class distance and decrease the intra-class distance. Last, MASTAF applies a lightweight fusion network and the nearest neighbor classifier to classify each query video. We demonstrate that MASTAF improves the state-of-the-art performance on three few-shot HAR video benchmarks.

Last, we present Multimodal Masked Autoencoders-Based One-Shot Learning (Mu-MAE), which represents a significant advancement in the field of HAR using multimodal sensors. Addressing the challenges posed by labor-intensive data collection and reliance on external pretrained models, MU-MAE introduces a synchronized masking strategy tailored for wearable sensors, coupled with a multimodal masked autoencoder architecture. This innovative approach compels the networks to capture more meaningful spatiotemporal features, facilitating effective self-supervised pretraining without the need for additional data. Furthermore, MU-MAE leverages the representations extracted from multimodal masked autoencoders to enhance cross-attention fusion, which highlights critical spatiotemporal features across different modalities while emphasizing differences between activity classes. Through comprehensive evaluations on MMAct one-shot classification datasets, MU-MAE demonstrates superior performance, achieving up to an 80.17% accuracy for five-way one-shot multimodal classification, thus establishing itself as a state-of-the-art solution in HAR for healthcare applications.

In this paper, I attempt to examine the process of reception of the thoughts of the great Italian thinker Antonio Gramsci in China.

On the one hand, as a founding member and onetime leader of the Communist Party of Italy, Antonio Gramsci is widely regarded as a highly original and influential thinker within the Marxist tradition, and well-known for his theoretical contributions to the Western Communist movements. Obviously influenced by Marx and Lenin, Gramsci also built his intellectual formation upon a number of Italian philosophers: Niccolò Machiavelli, Antonio Labriola, Giovanni Gentile and Benedetto Croce. When Italian Fascism finalized its dictatorship over the country in the 1920s, Gramsci was arrested at Mussolini’s orders and spent nine terrible years in prison, which eventually killed him in 1937. Prison Notebooks, the most significant work of Antonio Gramsci, contains numerous enigmas because of the censorship and the insufficiency of the resource in the prison. Some of these enigmas still remain unsolved by contemporary scholarship.

On the other hand, the People’s Republic of China is the most populous country in the world with over 1.3 billion people. Located in East Asia, China is a single-party state governed by the Communist Party of China after its victory in the Chinese Civil War since World War II. China is unique in the contemporary world for the reason that it is considered to be the largest existing socialist country.

Gramsci’s presence in China is a combination of two particular issues in the development of the modern Communist movement. In order to explore this topic, I have analyzed the existing articles and books published by Chinese intellectuals who intended to introduce the thoughts of Gramsci to China.

Time and the Other in the Anthropocene is a reflection on the contemporary transformation of the figure of the human (anthropos) into a geological agent and the way this is linked to a pluralization of time, on the one hand, and a pluralization of ontology, on the other hand. Specifically, it examines the conceptual link between the disintegration of the notion of Nature in the contemporary and its connection to the new pluralism in anthropology which has arisen as a response to the climate crisis. I consider the way this new pluralism, associated with what has been dubbed “the ontological turn” in anthropology, is strikingly being formulated, no longer through the question of the relativity of cultures, but rather that of natures. This move towards an analysis of various ways of being or what could be called a pluralist realism, I suggest, is one of the most original and crucial responses to climate crisis in that it foregrounds the question of the multiplicity of the modes of composing worlds which the Anthropocene has made a matter of dire necessity to think today.

Using the novel framework Philippe Descola and Bruno Latour’s comparative anthropology, I foreground the anthropologies of Amazonia and Melanesia as two examples of other modes of composing the world. I reflect on the manner in which the ontological predicates of the Amazonian (or Amerindian) and Melanesian world allow us to view fundamental modern concepts – nature and culture, subject and object – otherwise. As essentially anthropomorphic worlds, my implicit suggestion throughout is that Amazonia and Melanesia worlds stand in a stark counterpoint to anthropocentric ones with regard to the most basic principles of what counts as “being” and “human” being. By extending or attributing what we call “agency” and “subjectivity” to every other being in the world, Amazonians and Melanesians also turn these very categories from our point of view unrecognizable. Showing us that being and the world have been thought on entirely different premises and with entirely different concepts, the properly transcendental equivocation of these categories enables us to develop a new comparative vision and invites us to begin the collective work of imagining another mode of being human per se on the new Earth in which we live.

Recovering from a decades-long civil war, the Angolan government agreed in the early 2000s to contract the majority of projects under its National Reconstruction Program to Chinese state-owned enterprises, in exchange for multiple billions of dollars in oil-backed loans. This partnership has facilitated migration and commerce between China to Angola on an unprecedented scale. While some have denounced Chinese investments in postwar Angola as a form of neo-imperialism, others have celebrated Chinese financing of Angolan reconstruction as a practical mode of cooperation, or even a non-hegemonic expression of South-South solidarity. Without engaging in a moralistic debate between two “sides,” this dissertation examines the contradictions of Chinese-Angolan relations through an ethnographic study of everyday life and work at a Chinese state-owned construction firm in Luanda. By attending to ordinary interactions among employees of this firm and their co-workers, business partners, competitors, and critics, it explores how ethical relations are forged and broken in a context of political instability and moral controversy.

This dissertation analyzes the various scientific, political and cultural narratives about and the official public health responses to the recent 2009 H1N1 influenza pandemic. The historical planning emphasis on avian influenza, or `bird flu,' unintentionally created a large amount of uncertainty about how to respond to the threat from a milder, less severe, strain of influenza. I suggest that the specter of a future deadly global pandemic of avian influenza spurred a fascination or myopia in global health policy reminiscent of the mythical danger of listening to the Greek Sirens' song. Throughout this dissertation, I attempt to make sense out of the various and still-emerging accounts of the "swine flu" or 2009 H1N1 outbreak. Thus, I define `pathography' here as the combined historical, biological, social, political, economic and cultural narratives of the 2009 pandemic. The first section explores the historical and biological origins of influenza. Chapter one examines influenza research and early attempts to sequence the H1N1 virus. I use the genetic structure of the virus to suggest that public health agencies are related through a `viral kinship.' Chapter two details the material processes involved in the sequencing and discovery of influenza viruses. In it, I argue that what virology laboratories construct through these material processes of DNA sequencing is not - reductio ad absurdum - the virus itself, or even simply `knowledge' about a virus, but rather a complex network of scientists, laboratories, farms, public health institutions and other `actors' involved in the circulation of influenza samples and genetic information about influenza viruses. Chapter Three explores Hong Kong's history as an international `lab' for disease research, its local identity crisis as a former colony and current position as a Chinese city, and its unique role within global scientific and public health networks. The H1N1 virus is not simply a symbol of the complex global forces that shaped its emergence and its spread, the microscopic 2009 H1N1 virus embodies those macroscopic forces. Using the terms of molecular biology itself, I would like to suggest here that influenza viruses are not born sui generis out of larger economic, political or social processes, but are both created from and used to create the worlds in which they inhabit. The second section of the dissertation deals with this macro-level of analysis, or the political and cultural ramifications of influenza pandemics. Chapter four examines the seemingly new paradigm shift within global public health from the use of a scientific "certainty" to a biological and situational "uncertainty" as one of the foundations of response to infectious disease outbreaks. Chapter five analyzes the production, collection and sharing of epidemiological information during a pandemic. Scientific facts about the virus and the pandemic were freely circulated and agreed upon, but their cultural and political interpretations needed to be continuously negotiated. I argue that not only are cultural politics alive and well, but that they played a vital role in the global response to the influenza outbreak. Finally, I argue that to understand the pandemic as more than just a biological, social, political or historical event, one must look at all these narratives at once. The influenza virus is not thus merely a symbol of our times or for an increasingly globalized world; instead, I argue that it has partially constructed - and continues to shape - the contours of our world.

Reinforcement learning (RL), a critical subfield of machine learning, effectively models sequential decision-making scenarios for agents operating in various environments. Despite its extensive applications, RL encounters significant challenges in real world settings, particularly regarding limited data availability, the exploration-exploitation trade-off, and the lack of explainability. In this dissertation, I explore these issues through three distinct lenses. Firstly, I improve data efficiency in situations involving multiple agents or tasks. Secondly, I propose opportunistic learning algorithms in environments with varying exploration costs. Thirdly, I interpret the agent's learned policy through causal explanations.

The following sections outline the contributions. Initially, I study online global optimization in multi-agent situations. Cellular network configuration is a suitable application area experiencing these challenges, including the scarcity of diverse historical data, constrained experimental budgets imposed by network operators, and highly complex and unknown network performance functions. To overcome these challenges, I introduce an online-learning-based joint-optimization algorithm combining neural network regression with Gibbs sampling, which considerably outperforms distributed Q-learning in overall performance and ramp-up time. By leveraging similarities among tasks/base stations, I propose a kernel-based multi-task contextual bandit algorithm with the similarity estimated via conditional kernel embedding. These algorithms notably outperform the default cellular network configuration and the respective baseline algorithms.

Next, I focus on opportunistic learning, where the exploration cost in RL varies based on different environmental conditions. Given that exploration cost directly impacts the regret of selecting a sub-optimal action, I design the learning strategy to explore more when the cost is low and exploit when the cost is high. I propose an AdaLinUCB algorithm for opportunistic contextual bandits to balance the exploration-exploitation trade-off adaptively. My algorithm significantly outperforms existing contextual bandit algorithms in scenarios with large exploration cost fluctuations. I further develop two algorithms OppUCRL2 and OppPSRL for the finite-horizon episodic Markov decision process, demonstrating the benefits of opportunistic RL. My algorithms balance the exploration-exploitation trade-off dynamically through a variation factor dependent optimism, leading to superior performance. These results are supported by theoretical regret bound analyses ensuring their performance.

Lastly, I aim to enhance RL's interpretability by providing causal explanations. My approach quantifies the causal influence of states on actions and their temporal impact, thereby surpassing associative methods in RL policy explanation. I propose a mechanism to quantify the individual-level causal counterfactual path-specific importance score for a structural causal model. This mechanism effectively evaluates causal influence in decision chains, allowing us to comprehend better how a specific decision variable influences an outcome variable.

Recently, deep learning models have been extensively adopted in numerous applications, from health care to finance and entertainment industry. This wide-spread deployment of deep models raised concern over the privacy of data used to train deep models. This is a huge concern particularly for data-sensitive applications, such as health records, personal data, bio-metric data, etc. As a result, a new direction of research focusing on possible attacks aiming to identify training data of deep models emerged, called membership inference.Membership inference (MI) attacks identify which samples have been used during training and which samples have not. The first generation of membership inference attacks mainly used deep models' prediction confidence as a feature to identify training samples. The intuition is that deep models are more confident on samples they have seen during training than non-training samples. Despite their sound intuition and apparent successful reports, we, along a few other parallel studies, showed that the first generation of membership inference attacks are ineffective in practice for multiple reasons. First, they could not significantly outperform a naive baseline that labels a sample as a member (training sample) if it is correctly classified by the deep model and as a non-member (non-training sample) otherwise. Second, the confidence distribution of correctly classified samples, which cover the majority of a dataset, are not distinguishable between train and non-train samples. Only a small portion of mis-classified samples exhibit discrepant distribution. Third, all these membership inference attacks report average-case success metric (e.g., accuracy or ROC-AUC). However, privacy is not an average case-metric, and it should be treated similar to other security and privacy related problems. Similar to other security problems, the attack is reliable if it can identify a few training samples while almost on non-training samples are falsely labeled as a training sample. In other words, a reliable membership inference attack should have a decent true-positive rate (TPR) at low false-positive rates (FPR). In this dissertation, we aim to move the membership inference research in a more practical direction, either by showing the limitations of the current attacks or by proposing more reliable attacks. As stated earlier, we first show that the current generation of membership inference attacks are not reliable in practice. Then, we propose several new membership inference attacks that achieve more reliable performance in more realistic scenarios. The first attack focuses on the model's behavior in the entire sub-population, instead of a single sample in vacuum. More specifically, we compare the model's confidence on a target sample and other samples from the same sub-population. If the confidence of a sample is significantly higher than the average confidence on that sub-population, that is an indication of a training sample. We show that this attack can achieve moderate true positive with very low false positive. Additionally, we propose a BiGAN architecture to generate samples from the same sub-population, in case it is not available. The second attack aims to focus on user-level MI attack instead of the record-level MI attack. In this scenario, we identify if a user's data has been used during training instead if identifying which samples from the user have been used. Not only this attack is more realistic in privacy domain, but we show that we can achieve the state-of-the-art accuracy if multiple samples from a user are used to draw the membership inference. In another study, we show that MI attacks are generally more successful when deep ensemble is used. We show that deep ensemble shifts the distribution of train and non-train samples in a different way where they become significantly more distinguishable. Finally, we show that are a few simple aggregation mechanisms instead of ensemble averaging that can improve the accuracy and privacy of deep models in deep ensemble context. Finally, we illustrate a fundamental issue with current MI attacks, including the state-of-the-art attacks, that limits their applications in certain scenarios. We elaborate the issues with a practical scenario where membership inference attacks are used by an auditor (investigator) to prove to a judge/jury that the auditee unlawfully used sensitive data during training. Although the current SOTA attacks can identify some training samples with low false positive ratio in a common experimental setting extensively used for MI attacks, an auditee can generate unlimited number of samples on which MI attacks catastrophically fail. This can be used in court to easily discredit the allegation of the auditor and make the case dismissed. Interestingly, we show that auditee does not need to know anything about the auditor's membership inference attack to generate those challenging samples. We called this problem, discredibility. Currently, there is no attack immune to discredibility. We hope that our research sheds light on this newly-discovered issue and encourage researchers to investigate it.

The brain is the most complex part of human body. It is the source of intelligence and human behavior. Understanding how the brain works is essential for treatment of neurological disorders and development of more powerful artificial intelligence algorithms. To investigate its function, novel neurotechnologies are needed to record large-scale neural activity at high spatiotemporal resolution. Besides that, computational models could also be built to investigate possible mechanisms for neural activities at different levels. This dissertation presents research work on novel neurotechnologies and computational frameworks based on machine learning and biophysical modeling for investigation of brain functions. Chapter 1 presents a flexible insertable transparent microelectrode array (Neuro-FITM) for simultaneous electrical recordings from hippocampus and wide-field calcium imaging of dorsal cortex. We successfully recorded simultaneous activity of large-scale cortex and hippocampus. Using two-stage tensor component analysis and decoding analysis with support vector machines, we demonstrated distinct associations between hippocampus and cortex during hippocampal sharp-wave ripples. Chapter 2 presents a compact closed-loop optogenetics system based on transparent graphene microelectrode array. We extensively investigated light-induced artifacts for graphene electrodes and gold electrodes. We also validated the system for different frequencies of interest for neural recordings. The developed system can be used for various applications involving optogenetic stimulation and electrophysiological recordings. Chapter 3 presents the decoding of cortex-wide brain activity using surface electrical recordings over the cortex. We performed simultaneous local electrical recording and wide-field calcium imaging in awake mice. Using recurrent neural network, we demonstrated successful decoding of pixel-level cortical activity using locally recorded surface potentials. These results show that locally recorded surface potentials indeed contain rich information of large-scale neural activities. Chapter 4 presents a biophysical model for coupled network of hippocampal CA1 and prefrontal cortex to study mechanisms of memory trace transfer and reactivation. We found that stored memory traces in cortex can be retrieved through two different mechanisms, namely cell-specific inputs and non-specific spontaneous activities. Our study presents mechanistic explanations for memory transfer and retrieval. Chapter 5 is the conclusion of this dissertation. The future directions of neurotechnologies and computational models for neuroscience research are discussed

More than one decade after his infamous interventions of Pour Marx and Lire “Le Capital,” writing in a Preface to an uncelebrated book written by Gérard Duménil entitled Le Concept de loi économique dans “Le Capital,” Louis Althusser delivers a disarmingly devastating and simple question: “how does Marx think in Capital?” This question, Althusser writes, “refers us…to the rudiments…in Marx himself.” Motivated by his longstanding dissatisfaction with the doxa that Marx’s materialism inverts Hegel’s idealism, Althusser goes on to argue not only that the discursive orders and concepts of Capital do not bear such an inversion of Hegel’s method and system, but moreover that Capital discloses a mode of thinking that interdicts recourse to discursive constructions which bind an appropriation (Aneignung) and exposition (Darstellung) of material (Stoff) in a “purely theoretical” manner. In doing so, Louis Althusser, this dissertation argues, fundamentally rethinks the field and function of inversion in the critique of political economy.

In detecting the consequences of this reconsideration, the dissertation identifies how the original and sustained limit of Das Kapital: Kritik der politischen Oekonomie (1867), that is the chapter “Die moderne Kolonisationstheorie” (“The Modern Theory of Colonization”), brings forth such a critical inversion in Capital. Part One of the dissertation demonstrates both why “The Modern Theory of Colonization” has historically presented epistemological obstacles for its readers as well as the conceptual effects which are generated by grasping its notion of inversion. Foremost amongst these effects is the identification of the limit of the critique of political economy as a process, not a space, of modern colonization. Part One concludes that the critique identified by “The Modern Theory of Colonization” avails a place of thinking and speaking in modern colonization just beyond the limit of “pure theoretical” discourse in the activity that this dissertation calls, following as well as critically reconsidering Althusser, interpellation.

Oriented by these considerations, Part Two of the dissertation exposits an ethnography of a discourse of interpellation in a set of Indonesian interventions in this critique of modern colonization. The analysis focuses on three discursive and practical localities. The first is the novel written by the pseudonymous Multatuli, Max Havelaar, Or the Coffee Auctions of a Dutch Trading Company (1860) and its subsequent formal and aesthetic transformations through the 20th and 21st centuries, including its contemporary settings in the sculptures of Multatuli (2017, 2019) made by Dolorosa Sinaga (b. 1952) with her Somalaing studio in Jakarta. The second is the investigations and disputes regarding ontology and the social formation in the contemporary Indonesian Renaissance in philosophy, centering on the writings of Martin Suryajaya (b. 1989) and Muhammad Al-Fayyadl (b. 1986). The third is a practice and discourse of parrhesia in the artwork of Dolorosa Sinaga and the world of the verandah beside her Somalaing studio, where ethnographic research for this dissertation was based from June 2016 to March 2020. Part Two of the dissertation identifies how these interventions as interpellations are marked by a critical, rather than a “purely theoretical,” mode of inversion and effectivity.

The dissertation concludes that the exposition of “The Modern Theory of Colonization” orients the limit of Capital as a critique of the process whereby a colonizing state acts indirectly on the forceful conjuncture of capital and labour on the behalf of capital. The dissertation shows how this process is both identified and critiqued by Indonesian interventions within the conjuncture of modern colonization. These works both enrich and develop practices for interrupting the actuality, rather than the pure theoretical possibility, of the mistake (Verwechslung) of political economy. In lieu of reconstituting a priori theoretical impasses generated by the production processes of capital, the dissertation shows that these interventions initiate inversions in political, physical, and metaphysical modes of existence by obstructing the processes of modern colonization that mute domination and compulsion.