This thesis investigates three areas targeted at improving the reliability of machine learning; fairness in machine learning, strategic classification, and algorithmic robustness. Each of these domains has special properties or structure that can complicate learning. A theme throughout this thesis is thinking about ways in which a'plain' empirical risk minimization algorithm will be misleading or ineffective because of a mis-match between classical learning theory assumptions and specific properties of some data distribution in the wild. The overarching research goal for these related topics is to provide a crisp mathematical model for each learning scenario that exposes different failure modes and makes trade-offs between important metrics explicit in order to provide algorithmic advice or recommendations to practitioners and expose gaps for future research. By tuning our learning algorithms to be more distribution specific in these scenarios, the resulting learned system will exhibit higher utility and avoid catastrophic failure modes. This research is grounded in the theory of machine learning and is fundamentally mathematical in nature, with empirical support when appropriate. Theory is particularly important in these sensitive domains as it is unclear which poor behavior in deployed systems is a natural or benign consequence of a learning system with the underlying distribution,contrasting with problematic but correctable behavior caused by an error in algorithm design or implementation, how to mitigate these issues, or what a successful outcome even looks like in each problem. Theoretical understanding in each domain can help guide best practices and allow for the design of effective, reliable, and robust systems.

The detection of cyber-attacks in computer networks is a crucial and ongoing research challenge. Machine learning-based attack classification offers a promising solution, as these models can be continuously updated with new data, enhancing the effectiveness of network intrusion detection systems (NIDS). Unlike binary classification models that simply indicate the presence of an attack, multi-class models can identify specific types of attacks, allowing for more targeted and effective incident responses. However, a significant drawback of these classification models is their sensitivity to imbalanced training data. Recent advances suggest that generative models can assist in data augmentation, claiming to offer superior solutions for imbalanced datasets. Classical balancing methods, although less novel, also provide potential remedies for this issue. Despite these claims, a comprehensive comparison of these methods within the NIDS domain is lacking. Most existing studies focus narrowly on individual methods, making it difficult to compare results due to varying experimental setups. To close this gap, we designed a systematic framework to compare classical and generative resampling methods for class balancing across multiple popular classification models in the NIDS domain, evaluated on several NIDS benchmark datasets. Our experiments indicate that resampling methods for balancing training data do not reliably improve classification performance. Although some instances show performance improvements, the majority of results indicate decreased performance, with no consistent trend in favor of a specific resampling technique enhancing a particular classifier.

Detecting and attributing temperature increases due to climate change is crucial for understanding global warming and guiding adaptation strategies. The complexity of distinguishing human-induced climate signals from natural variability has challenged traditional detection and attribution (D&A) approaches, which seek to identify specific "fingerprints" in climate response variables. Deep learning offers potential for discerning these complex patterns in expansive spatial datasets. However, lack of standard protocols has hindered consistent comparisons across studies. We introduce ClimDetect, a standardized dataset of over 816k daily climate snapshots, designed to enhance model accuracy in identifying climate change signals. ClimDetect integrates various input and target variables used in past research, ensuring comparability and consistency. We also explore the application of vision transformers (ViT) to climate data, a novel and modernizing approach in this context. Our open-access data and code serve as a benchmark for advancing climate science through improved model evaluations.

This paper introduces a group of novel datasets representing real-time time-series and streaming data of energy prices in New Zealand, sourced from the Electricity Market Information (EMI) website maintained by the New Zealand government. The datasets are intended to address the scarcity of proper datasets for streaming regression learning tasks. We conduct extensive analyses and experiments on these datasets, covering preprocessing techniques, regression tasks, prediction intervals, concept drift detection, and anomaly detection. Our experiments demonstrate the datasets' utility and highlight the challenges and opportunities for future research in energy price forecasting.

Reasoning about real-life events is a unifying challenge in AI and NLP that has profound utility in a variety of domains, while fallacy in high-stake applications could be catastrophic. Able to work with diverse text in these domains, large language models (LLMs) have proven capable of answering questions and solving problems. However, I show that end-to-end LLMs still systematically fail to reason about complex events, and they lack interpretability due to their black-box nature. To address these issues, I propose three general approaches to use LLMs in conjunction with a structured representation of events. The first is a language-based representation involving relations of sub-events that can be learned by LLMs via fine-tuning. The second is a semi-symbolic representation involving states of entities that can be predicted and leveraged by LLMs via few-shot prompting. The third is a fully symbolic representation that can be predicted by LLMs trained with structured data and be executed by symbolic solvers. On a suite of event reasoning tasks spanning common-sense inference and planning, I show that each approach greatly outperforms end-to-end LLMs with more interpretability. These results suggest manners of synergy between LLMs and structured representations for event reasoning and beyond.

The increasing sophistication of large language models (LLMs) has allowed for their more prominent presence in political science research. One particular area gathering significant attention in the field is the use of LLMs as annotators. Research has shown promising results, with LLMs often outperforming human coders (Gilardi, Alizadeh and Kubli, 2023) and providing comparable accuracy when labelling political text, across multiple languages (Heseltine and Clemm von Hohenberg, 2024). While researchers have evaluated the performance of LLMs as annotators across different domains, there still little information on how the known biases of LLMs (see Gallegos et al., 2024) can affect their performance. For human annotators, studies show that political cues, such as party, have an effect on their coding decisions (Laver and Garry, 2000; Benoit et al., 2016; Ennser-Jedenastik and Meyer, 2018).

The field of fair AI aims to counter biased algorithms through computational modelling. However, it faces increasing criticism for perpetuating the use of overly technical and reductionist methods. As a result, novel approaches appear in the field to address more socially-oriented and interdisciplinary (SOI) perspectives on fair AI. In this paper, we take this dynamic as the starting point to study the tension between computer science (CS) and SOI research. By drawing on STS and CSCW theory, we position fair AI research as a matter of 'organizational alignment': what makes research 'doable' is the successful alignment of three levels of work organization (the social world, the laboratory, and the experiment). Based on qualitative interviews with CS researchers, we analyze the tasks, resources, and actors required for doable research in the case of fair AI. We find that CS researchers engage with SOI research to some extent, but organizational conditions, articulation work, and ambiguities of the social world constrain the doability of SOI research for them. Based on our findings, we identify and discuss problems for aligning CS and SOI as fair AI continues to evolve.

In order to understand the appropriate field of view (FOV) size of celestial automatic navigation systems for surface ships, we investigate the variations of measurement accuracy of star position and probability of successful star identification with respect to FOV, focusing on the decreasing number of observable star magnitudes and the presence of physically covered stars in marine environments. The results revealed that, although a larger FOV reduces the measurement accuracy of star positions, it increases the number of observable objects and thus improves the probability of star identification using subgraph isomorphism-based methods. It was also found that, although at least four objects need to be observed for accurate identification, four objects may not be sufficient for wider FOVs. On the other hand, from the point of view of celestial navigation systems, a decrease in the measurement accuracy leads to a decrease in positioning accuracy. Therefore, it was found that maximizing the FOV is required for celestial automatic navigation systems as long as the desired positioning accuracy can be ensured. Furthermore, it was found that algorithms incorporating more than four observed celestial objects are required to achieve highly accurate star identification over a wider FOV.

What techniques can be used to verify compliance with international agreements about advanced AI development? In this paper, we examine 10 verification methods that could detect two types of potential violations: unauthorized AI training (e.g., training runs above a certain FLOP threshold) and unauthorized data centers. We divide the verification methods into three categories: (a) national technical means (methods requiring minimal or no access from suspected non-compliant nations), (b) access-dependent methods (methods that require approval from the nation suspected of unauthorized activities), and (c) hardware-dependent methods (methods that require rules around advanced hardware). For each verification method, we provide a description, historical precedents, and possible evasion techniques. We conclude by offering recommendations for future work related to the verification and enforcement of international AI governance agreements.

The widespread adoption of cloud-based proprietary large language models (LLMs) has introduced significant challenges, including operational dependencies, privacy concerns, and the necessity of continuous internet connectivity. In this work, we introduce an LLMOps pipeline, "LlamaDuo", for the seamless migration of knowledge and abilities from service-oriented LLMs to smaller, locally manageable models. This pipeline is crucial for ensuring service continuity in the presence of operational failures, strict privacy policies, or offline requirements. Our LlamaDuo involves fine-tuning a small language model against the service LLM using a synthetic dataset generated by the latter. If the performance of the fine-tuned model falls short of expectations, it is enhanced by further fine-tuning with additional similar data created by the service LLM. This iterative process guarantees that the smaller model can eventually match or even surpass the service LLM's capabilities in specific downstream tasks, offering a practical and scalable solution for managing AI deployments in constrained environments. Extensive experiments with leading-edge LLMs are conducted to demonstrate the effectiveness, adaptability, and affordability of LlamaDuo across various downstream tasks.