Fine-tuning large language models (LLM) can be costly. Parameter-efficient fine-tuning (PEFT) addresses the problems by training a fraction of the parameters, whose success reveals the expressiveness and flexibility of pretrained models. This paper studies the limit of PEFT, by further simplifying its design and reducing the number of trainable parameters beyond standard setups. To this end, we use Random Masking to fine-tune the pretrained model. Despite its simplicity, we show that Random Masking is surprisingly effective: with a larger-than-expected learning rate, Random Masking can match the performance of standard PEFT algorithms such as LoRA on various tasks, using fewer trainable parameters. We provide both empirical and theoretical explorations into the success of Random Masking. We show that masking induces a flatter loss landscape and more distant solutions, which allows for and necessitates large learning rates.

In this paper, we propose a differentially private decentralized learning method (termed PrivSGP-VR) which employs stochastic gradient push with variance reduction and guarantees $(\epsilon, \delta)$-differential privacy (DP) for each node. Our theoretical analysis shows that, under DP Gaussian noise with constant variance, PrivSGP-VR achieves a sub-linear convergence rate of $\mathcal{O}(1/\sqrt{nK})$, where $n$ and $K$ are the number of nodes and iterations, respectively, which is independent of stochastic gradient variance, and achieves a linear speedup with respect to $n$. Leveraging the moments accountant method, we further derive an optimal $K$ to maximize the model utility under certain privacy budget in decentralized settings. With this optimized $K$, PrivSGP-VR achieves a tight utility bound of $\mathcal{O}\left( \sqrt{d\log \left( \frac{1}{\delta} \right)}/(\sqrt{n}J\epsilon) \right)$, where $J$ and $d$ are the number of local samples and the dimension of decision variable, respectively, which matches that of the server-client distributed counterparts, and exhibits an extra factor of $1/\sqrt{n}$ improvement compared to that of the existing decentralized counterparts, such as A(DP)$^2$SGD. Extensive experiments corroborate our theoretical findings, especially in terms of the maximized utility with optimized $K$, in fully decentralized settings.

Predicting the minimum operating voltage ($V_{min}$) of chips is one of the important techniques for improving the manufacturing testing flow, as well as ensuring the long-term reliability and safety of in-field systems. Current $V_{min}$ prediction methods often provide only point estimates, necessitating additional techniques for constructing prediction confidence intervals to cover uncertainties caused by different sources of variations. While some existing techniques offer region predictions, but they rely on certain distributional assumptions and/or provide no coverage guarantees. In response to these limitations, we propose a novel distribution-free $V_{min}$ interval estimation methodology possessing a theoretical guarantee of coverage. Our approach leverages conformalized quantile regression and on-chip monitors to generate reliable prediction intervals. We demonstrate the effectiveness of the proposed method on an industrial 5nm automotive chip dataset. Moreover, we show that the use of on-chip monitors can reduce the interval length significantly for $V_{min}$ prediction.

We study the problem of privacy-preserving $k$-means clustering in the horizontally federated setting. Existing federated approaches using secure computation, suffer from substantial overheads and do not offer output privacy. At the same time, differentially private (DP) $k$-means algorithms assume a trusted central curator and do not extend to federated settings. Naively combining the secure and DP solutions results in a protocol with impractical overhead. Instead, our work provides enhancements to both the DP and secure computation components, resulting in a design that is faster, more private, and more accurate than previous work. By utilizing the computational DP model, we design a lightweight, secure aggregation-based approach that achieves four orders of magnitude speed-up over state-of-the-art related work. Furthermore, we not only maintain the utility of the state-of-the-art in the central model of DP, but we improve the utility further by taking advantage of constrained clustering techniques.

This thesis aims to study some of the mathematical challenges that arise in the analysis of statistical sequential decision-making algorithms for postoperative patients follow-up. Stochastic bandits (multiarmed, contextual) model the learning of a sequence of actions (policy) by an agent in an uncertain environment in order to maximise observed rewards. To learn optimal policies, bandit algorithms have to balance the exploitation of current knowledge and the exploration of uncertain actions. Such algorithms have largely been studied and deployed in industrial applications with large datasets, low-risk decisions and clear modelling assumptions, such as clickthrough rate maximisation in online advertising. By contrast, digital health recommendations call for a whole new paradigm of small samples, risk-averse agents and complex, nonparametric modelling. To this end, we developed new safe, anytime-valid concentration bounds, (Bregman, empirical Chernoff), introduced a new framework for risk-aware contextual bandits (with elicitable risk measures) and analysed a novel class of nonparametric bandit algorithms under weak assumptions (Dirichlet sampling). In addition to the theoretical guarantees, these results are supported by in-depth empirical evidence. Finally, as a first step towards personalised postoperative follow-up recommendations, we developed with medical doctors and surgeons an interpretable machine learning model to predict the long-term weight trajectories of patients after bariatric surgery.

Traditional mathematical programming solvers require long computational times to solve constrained minimization problems of complex and large-scale physical systems. Therefore, these problems are often transformed into unconstrained ones, and solved with computationally efficient optimization approaches based on first-order information, such as the gradient descent method. However, for unconstrained problems, balancing the minimization of the objective function with the reduction of constraint violations is challenging. We consider the class of time-dependent minimization problems with increasing (possibly) nonlinear and non-convex objective function and non-decreasing (possibly) nonlinear and non-convex inequality constraints. To efficiently solve them, we propose a penalty-based guardrail algorithm (PGA). This algorithm adapts a standard penalty-based method by dynamically updating the right-hand side of the constraints with a guardrail variable which adds a margin to prevent violations. We evaluate PGA on two novel application domains: a simplified model of a district heating system and an optimization model derived from learned deep neural networks. Our method significantly outperforms mathematical programming solvers and the standard penalty-based method, and achieves better performance and faster convergence than a state-of-the-art algorithm (IPDD) within a specified time limit.

Emphasizing a data-centric AI approach, this paper introduces a novel two-stage active learning (AL) pipeline for automatic speech recognition (ASR), combining unsupervised and supervised AL methods. The first stage utilizes unsupervised AL by using x-vectors clustering for diverse sample selection from unlabeled speech data, thus establishing a robust initial dataset for the subsequent supervised AL. The second stage incorporates a supervised AL strategy, with a batch AL method specifically developed for ASR, aimed at selecting diverse and informative batches of samples. Here, sample diversity is also achieved using x-vectors clustering, while the most informative samples are identified using a Bayesian AL method tailored for ASR with an adaptation of Monte Carlo dropout to approximate Bayesian inference. This approach enables precise uncertainty estimation, thereby enhancing ASR model training with significantly reduced data requirements. Our method has shown superior performance compared to competing methods on homogeneous, heterogeneous, and OOD test sets, demonstrating that strategic sample selection and innovative Bayesian modeling can substantially optimize both labeling effort and data utilization in deep learning-based ASR applications.

The negative consequence of this openness of social media platforms is the spread of false information disguised as truth, i.e., fake news. Fake news can be defined as deceptive posts with an intention to mislead consumers in their purchase or approaching the context of misinformation and disinformation (Xiao et al, 2020). Specifically, while misinformation is an inadvertent action, disinformation is a deliberate creation/sharing of false information. The authenticity and intention can be distinguished as: (i) non-factual and mislead, i.e., deceptive news and disinformation; (ii) factual and mislead (cherry-picking); (iii) undefined and mislead (click-bait); and (iv) non-factual and undefined, i.e., misinformation. Misinformation and fake news are characterized by their big volume, uncertainty, and short-lived nature. Furthermore, they disseminate faster and further on social media sites causing serious impact on politics and economics (Tandoc, 2019). Accordingly, the report on digital transformation of media and the rise of disinformation/fake news of the European Union (EU) (Martens et al, 2018) reinforces the need to strengthen trust in digital media.

The Internet has become an essential tool for people in the modern world. Humans, like all living organisms, have essential requirements for survival. These include access to atmospheric oxygen, potable water, protective shelter, and sustenance. The constant flux of the world is making our existence less complicated. A significant portion of the population utilizes online food ordering services to have meals delivered to their residences. Although there are numerous methods for ordering food, customers sometimes experience disappointment with the food they receive. Our endeavor was to establish a model that could determine if food is of good or poor quality. We compiled an extensive dataset of over 1484 online reviews from prominent food ordering platforms, including Food Panda and HungryNaki. Leveraging the collected data, a rigorous assessment of various deep learning and machine learning techniques was performed to determine the most accurate approach for predicting food quality. Out of all the algorithms evaluated, logistic regression emerged as the most accurate, achieving an impressive 90.91% accuracy. The review offers valuable insights that will guide the user in deciding whether or not to order the food.

Explainable Recommender Systems is an important field of study which provides reasons behind the suggested recommendations. Explanations with recommender systems are useful for developers while debugging anomalies within the system and for consumers while interpreting the model's effectiveness in capturing their true preferences towards items. However, most of the existing state-of-the-art (SOTA) explainable recommenders could not retain their explanation capability under noisy circumstances and moreover are not generalizable across different datasets. The robustness of the explanations must be ensured so that certain malicious attackers do not manipulate any high-stake decision scenarios to their advantage, which could cause severe consequences affecting large groups of interest. In this work, we present a general framework for feature-aware explainable recommenders that can withstand external attacks and provide robust and generalized explanations. This paper presents a novel framework which could be utilized as an additional defense tool, preserving the global explainability when subject to model-based white box attacks. Our framework is simple to implement and supports different methods regardless of the internal model structure and intrinsic utility within any model. We experimented our framework on two architecturally different feature-based SOTA explainable algorithms by training them on three popular e-commerce datasets of increasing scales. We noticed that both the algorithms displayed an overall improvement in the quality and robustness of the global explainability under normal as well as noisy environments across all the datasets, indicating the flexibility and mutability of our framework.