Unsupervised Semantic Segmentation (USS) involves segmenting images without relying on predefined labels, aiming to alleviate the burden of extensive human labeling. Existing methods utilize features generated by self-supervised models and specific priors for clustering. However, their clustering objectives are not involved in the optimization of the features during training. Additionally, due to the lack of clear class definitions in USS, the resulting segments may not align well with the clustering objective. In this paper, we introduce a novel approach called Optimally Matched Hierarchy (OMH) to simultaneously address the above issues. The core of our method lies in imposing structured sparsity on the feature space, which allows the features to encode information with different levels of granularity. The structure of this sparsity stems from our hierarchy (OMH). To achieve this, we learn a soft but sparse hierarchy among parallel clusters through Optimal Transport. Our OMH yields better unsupervised segmentation performance compared to existing USS methods. Our extensive experiments demonstrate the benefits of OMH when utilizing our differentiable paradigm. We will make our code publicly available.

In recent years, advancements in deep learning techniques have considerably enhanced the efficiency and accuracy of medical diagnostics. In this work, a novel approach using multi-task learning (MTL) for the simultaneous classification of lung sounds and lung diseases is proposed. Our proposed model leverages MTL with four different deep learning models such as 2D CNN, ResNet50, MobileNet and Densenet to extract relevant features from the lung sound recordings. The ICBHI 2017 Respiratory Sound Database was employed in the current study. The MTL for MobileNet model performed better than the other models considered, with an accuracy of74\% for lung sound analysis and 91\% for lung diseases classification. Results of the experimentation demonstrate the efficacy of our approach in classifying both lung sounds and lung diseases concurrently. In this study,using the demographic data of the patients from the database, risk level computation for Chronic Obstructive Pulmonary Disease is also carried out. For this computation, three machine learning algorithms namely Logistic Regression, SVM and Random Forest classifierswere employed. Among these ML algorithms, the Random Forest classifier had the highest accuracy of 92\%.This work helps in considerably reducing the physician's burden of not just diagnosing the pathology but also effectively communicating to the patient about the possible causes or outcomes.

This thesis studies advanced probabilistic models, including both their theoretical foundations and practical applications, for different semi-supervised learning (SSL) tasks. The proposed probabilistic methods are able to improve the safety of AI systems in real applications by providing reliable uncertainty estimates quickly, and at the same time, achieve competitive performance compared to their deterministic counterparts. The experimental results indicate that the methods proposed in the thesis have great value in safety-critical areas, such as the autonomous driving or medical imaging analysis domain, and pave the way for the future discovery of highly effective and efficient probabilistic approaches in the SSL sector.

LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question: "What would the preference be if the model's and baseline's output had the same length?". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ .

We apply Bayesian Additive Regression Tree (BART) principles to training an ensemble of small neural networks for regression tasks. Using Markov Chain Monte Carlo, we sample from the posterior distribution of neural networks that have a single hidden layer. To create an ensemble of these, we apply Gibbs sampling to update each network against the residual target value (i.e. subtracting the effect of the other networks). We demonstrate the effectiveness of this technique on several benchmark regression problems, comparing it to equivalent shallow neural networks, BART, and ordinary least squares. Our Bayesian Additive Regression Networks (BARN) provide more consistent and often more accurate results. On test data benchmarks, BARN averaged between 5 to 20 percent lower root mean square error. This error performance does come at the cost, however, of greater computation time. BARN sometimes takes on the order of a minute where competing methods take a second or less. But, BARN without cross-validated hyperparameter tuning takes about the same amount of computation time as tuned other methods. Yet BARN is still typically more accurate.

We propose Deep Longitudinal Targeted Minimum Loss-based Estimation (Deep LTMLE), a novel approach to estimate the counterfactual mean of outcome under dynamic treatment policies in longitudinal problem settings. Our approach utilizes a transformer architecture with heterogeneous type embedding trained using temporal-difference learning. After obtaining an initial estimate using the transformer, following the targeted minimum loss-based likelihood estimation (TMLE) framework, we statistically corrected for the bias commonly associated with machine learning algorithms. Furthermore, our method also facilitates statistical inference by enabling the provision of 95% confidence intervals grounded in asymptotic statistical theory. Simulation results demonstrate our method's superior performance over existing approaches, particularly in complex, long time-horizon scenarios. It remains effective in small-sample, short-duration contexts, matching the performance of asymptotically efficient estimators. To demonstrate our method in practice, we applied our method to estimate counterfactual mean outcomes for standard versus intensive blood pressure management strategies in a real-world cardiovascular epidemiology cohort study.

Community detection in multi-layer networks has emerged as a crucial area of modern network analysis. However, conventional approaches often assume that nodes belong exclusively to a single community, which fails to capture the complex structure of real-world networks where nodes may belong to multiple communities simultaneously. To address this limitation, we propose novel spectral methods to estimate the common mixed memberships in the multi-layer mixed membership stochastic block model. The proposed methods leverage the eigen-decomposition of three aggregate matrices: the sum of adjacency matrices, the debiased sum of squared adjacency matrices, and the sum of squared adjacency matrices. We establish rigorous theoretical guarantees for the consistency of our methods. Specifically, we derive per-node error rates under mild conditions on network sparsity, demonstrating their consistency as the number of nodes and/or layers increases under the multi-layer mixed membership stochastic block model. Our theoretical results reveal that the method leveraging the sum of adjacency matrices generally performs poorer than the other two methods for mixed membership estimation in multi-layer networks. We conduct extensive numerical experiments to empirically validate our theoretical findings. For real-world multi-layer networks with unknown community information, we introduce two novel modularity metrics to quantify the quality of mixed membership community detection. Finally, we demonstrate the practical applications of our algorithms and modularity metrics by applying them to real-world multi-layer networks, demonstrating their effectiveness in extracting meaningful community structures.

We present a nonparametric construction for deep learning compatible modern Hopfield models and utilize this framework to debut an efficient variant. Our key contribution stems from interpreting the memory storage and retrieval processes in modern Hopfield models as a nonparametric regression problem subject to a set of query-memory pairs. Crucially, our framework not only recovers the known results from the original dense modern Hopfield model but also fills the void in the literature regarding efficient modern Hopfield models, by introducing \textit{sparse-structured} modern Hopfield models with sub-quadratic complexity. We establish that this sparse model inherits the appealing theoretical properties of its dense analogue -- connection with transformer attention, fixed point convergence and exponential memory capacity -- even without knowing details of the Hopfield energy function. Additionally, we showcase the versatility of our framework by constructing a family of modern Hopfield models as extensions, including linear, random masked, top-$K$ and positive random feature modern Hopfield models. Empirically, we validate the efficacy of our framework in both synthetic and realistic settings.

This paper considers the problem of regression analysis with random covariance matrix as outcome and Euclidean covariates in the framework of Fr\'echet regression on the Bures-Wasserstein manifold. Such regression problems have many applications in single cell genomics and neuroscience, where we have covariance matrix measured over a large set of samples. Fr\'echet regression on the Bures-Wasserstein manifold is formulated as estimating the conditional Fr\'echet mean given covariates $x$. A non-asymptotic $\sqrt{n}$-rate of convergence (up to $\log n$ factors) is obtained for our estimator $\hat{Q}_n(x)$ uniformly for $\left\|x\right\| \lesssim \sqrt{\log n}$, which is crucial for deriving the asymptotic null distribution and power of our proposed statistical test for the null hypothesis of no association. In addition, a central limit theorem for the point estimate $\hat{Q}_n(x)$ is obtained, giving insights to a test for covariate effects. The null distribution of the test statistic is shown to converge to a weighted sum of independent chi-squares, which implies that the proposed test has the desired significance level asymptotically. Also, the power performance of the test is demonstrated against a sequence of contiguous alternatives. Simulation results show the accuracy of the asymptotic distributions. The proposed methods are applied to a single cell gene expression data set that shows the change of gene co-expression network as people age.

This paper explores the intersection of Natural Language Processing (NLP) and financial analysis, focusing on the impact of sentiment analysis in stock price prediction. We employ BERTopic, an advanced NLP technique, to analyze the sentiment of topics derived from stock market comments. Our methodology integrates this sentiment analysis with various deep learning models, renowned for their effectiveness in time series and stock prediction tasks. Through comprehensive experiments, we demonstrate that incorporating topic sentiment notably enhances the performance of these models. The results indicate that topics in stock market comments provide implicit, valuable insights into stock market volatility and price trends. This study contributes to the field by showcasing the potential of NLP in enriching financial analysis and opens up avenues for further research into real-time sentiment analysis and the exploration of emotional and contextual aspects of market sentiment. The integration of advanced NLP techniques like BERTopic with traditional financial analysis methods marks a step forward in developing more sophisticated tools for understanding and predicting market behaviors.