The application of data-intensive automatic speech recognition (ASR) technologies to dysarthric and elderly adult speech is confronted by their mismatch against healthy and nonaged voices, data scarcity and large speaker-level variability. To this end, this paper proposes two novel data-efficient methods to learn homogeneous dysarthric and elderly speaker-level features for rapid, on-the-fly test-time adaptation of DNN/TDNN and Conformer ASR models. These include: 1) speaker-level variance-regularized spectral basis embedding (VR-SBE) features that exploit a special regularization term to enforce homogeneity of speaker features in adaptation; and 2) feature-based learning hidden unit contributions (f-LHUC) transforms that are conditioned on VR-SBE features. Experiments are conducted on four tasks across two languages: the English UASpeech and TORGO dysarthric speech datasets, the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech corpora. The proposed on-the-fly speaker adaptation techniques consistently outperform baseline iVector and xVector adaptation by statistically significant word or character error rate reductions up to 5.32% absolute (18.57% relative) and batch-mode LHUC speaker adaptation by 2.24% absolute (9.20% relative), while operating with real-time factors speeding up to 33.6 times against xVectors during adaptation. The efficacy of the proposed adaptation techniques is demonstrated in a comparison against current ASR technologies including SSL pre-trained systems on UASpeech, where our best system produces a state-of-the-art WER of 23.33%. Analyses show VR-SBE features and f-LHUC transforms are insensitive to speaker-level data quantity in testtime adaptation. T-SNE visualization reveals they have stronger speaker-level homogeneity than baseline iVectors, xVectors and batch-mode LHUC transforms.

Medication recommendation combines patient medical history with biomedical knowledge to assist doctors in determining medication combinations more accurately and safely. Existing works based on molecular knowledge neglect the 3D geometric structure of molecules and fail to learn the high-dimensional information of medications, leading to structural confusion. Additionally, it does not extract key substructures from a single patient visit, resulting in the failure to identify medication molecules suitable for the current patient visit. To address the above limitations, we propose a bimodal molecular recommendation framework named BiMoRec, which introduces 3D molecular structures to obtain atomic 3D coordinates and edge indices, overcoming the inherent lack of high-dimensional molecular information in 2D molecular structures. To retain the fast training and prediction efficiency of the recommendation system, we use bimodal graph contrastive pretraining to maximize the mutual information between the two molecular modalities, achieving the fusion of 2D and 3D molecular graphs and re-evaluating substructures at the visit level. Specifically, we use deep learning networks to construct a pretraining method that acquires 2D and 3D molecular structure representations and substructure representations, and obtain mutual information through contrastive learning. We then generate fused molecular representations using the trained GNN module and re-determine the relevance of substructure representations in combination with the patient's clinical history. Finally, we generate the final medication combination based on the extracted substructure sequences. Our implementation on the MIMIC-III and MIMIC-IV datasets demonstrates that our method achieves state-of-the-art performance. Compared to the second-best baseline, our model improves accuracy by 2.07%, with DDI at the same level as the baseline.

Federated Edge Learning (FEEL) emerges as a pioneering distributed machine learning paradigm for the 6G Hyper-Connectivity, harnessing data from the Internet of Things (IoT) devices while upholding data privacy. However, current FEEL algorithms struggle with non-independent and non-identically distributed (non-IID) data, leading to elevated communication costs and compromised model accuracy. To address these statistical imbalances within FEEL, we introduce a clustered data sharing framework, mitigating data heterogeneity by selectively sharing partial data from cluster heads to trusted associates through sidelink-aided multicasting. The collective communication pattern is integral to FEEL training, where both cluster formation and the efficiency of communication and computation impact training latency and accuracy simultaneously. To tackle the strictly coupled data sharing and resource optimization, we decompose the overall optimization problem into the clients clustering and effective data sharing subproblems. Specifically, a distribution-based adaptive clustering algorithm (DACA) is devised basing on three deductive cluster forming conditions, which ensures the maximum sharing yield. Meanwhile, we design a stochastic optimization based joint computed frequency and shared data volume optimization (JFVO) algorithm, determining the optimal resource allocation with an uncertain objective function. The experiments show that the proposed framework facilitates FEEL on non-IID datasets with faster convergence rate and higher model accuracy in a limited communication environment.

In the context of ground robot navigation in unstructured hazardous environments, the coupling of efficient path planning with an adequate environment representation is a crucial topic in order to guarantee the robot safety while ensuring the accomplishment of its mission. This paper discusses the exploitation of an environment representation obtained via Gaussian process regression (GPR) for smooth path planning using gradient descent B\'ezier curve optimisation (BCO). A continuous differentiable GPR of the terrain traversability and obstacle distance is used to plan paths with a weighted A* discrete planner, a T-RRT sampling-based planner and BCO using A* or T-RRT computed paths as prior. Numerical experiments in procedurally generated 2D environments allowed to compare the paths planned by the described methods and highlight the benefits of the joint use of the GPR continuous representation and the BCO smooth path planning with these different priors.

Differentially Private Stochastic Gradient Descent (DP-SGD) is a popular iterative algorithm used to train machine learning models while formally guaranteeing the privacy of users. However the privacy analysis of DP-SGD makes the unrealistic assumption that all intermediate iterates (aka internal state) of the algorithm are released since in practice, only the final trained model, i.e., the final iterate of the algorithm is released. In this hidden state setting, prior work has provided tighter analyses, albeit only when the loss function is constrained, e.g., strongly convex and smooth or linear. On the other hand, the privacy leakage observed empirically from hidden state DP-SGD, even when using non-convex loss functions suggest that there is in fact a gap between the theoretical privacy analysis and the privacy guarantees achieved in practice. Therefore, it remains an open question whether privacy amplification for DP-SGD is possible in the hidden state setting for general loss functions. Unfortunately, this work answers the aforementioned research question negatively. By carefully constructing a loss function for DP-SGD, we show that for specific loss functions, the final iterate of DP-SGD alone leaks as much information as the sequence of all iterates combined. Furthermore, we empirically verify this result by evaluating the privacy leakage from the final iterate of DP-SGD with our loss function and show that this matches the theoretical upper bound guaranteed by DP exactly. Therefore, we show that the current privacy analysis fo DP-SGD is tight for general loss functions and conclude that no privacy amplification is possible for DP-SGD in general for all (possibly non-convex) loss functions.

Recent studies suggest that deep learning models inductive bias towards favoring simpler features may be one of the sources of shortcut learning. Yet, there has been limited focus on understanding the complexity of the myriad features that models learn. In this work, we introduce a new metric for quantifying feature complexity, based on $\mathscr{V}$-information and capturing whether a feature requires complex computational transformations to be extracted. Using this $\mathscr{V}$-information metric, we analyze the complexities of 10,000 features, represented as directions in the penultimate layer, that were extracted from a standard ImageNet-trained vision model. Our study addresses four key questions: First, we ask what features look like as a function of complexity and find a spectrum of simple to complex features present within the model. Second, we ask when features are learned during training. We find that simpler features dominate early in training, and more complex features emerge gradually. Third, we investigate where within the network simple and complex features flow, and find that simpler features tend to bypass the visual hierarchy via residual connections. Fourth, we explore the connection between features complexity and their importance in driving the networks decision. We find that complex features tend to be less important. Surprisingly, important features become accessible at earlier layers during training, like a sedimentation process, allowing the model to build upon these foundational elements.

Based on audio recordings made once a month during the first 12 months of a child's life, we propose a new method for clustering this set of vocalizations. We use a topologically augmented representation of the vocalizations, employing two persistence diagrams for each vocalization: one computed on the surface of its spectrogram and one on the Takens' embeddings of the vocalization. A synthetic persistent variable is derived for each diagram and added to the MFCCs (Mel-frequency cepstral coefficients). Using this representation, we fit a non-parametric Bayesian mixture model with a Dirichlet process prior to model the number of components. This procedure leads to a novel data-driven categorization of vocal productions. Our findings reveal the presence of 8 clusters of vocalizations, allowing us to compare their temporal distribution and acoustic profiles in the first 12 months of life.

As the size of datasets used in statistical learning continues to grow, distributed training of models has attracted increasing attention. These methods partition the data and exploit parallelism to reduce memory and runtime, but suffer increasingly from communication costs as the data size or the number of iterations grows. Recent work on linear models has shown that a surrogate likelihood can be optimized locally to iteratively improve on an initial solution in a communication-efficient manner. However, existing versions of these methods experience multiple shortcomings as the data size becomes massive, including diverging updates and efficiently handling sparsity. In this work we develop solutions to these problems which enable us to learn a communication-efficient distributed logistic regression model even beyond millions of features. In our experiments we demonstrate a large improvement in accuracy over distributed algorithms with only a few distributed update steps needed, and similar or faster runtimes. Our code is available at \url{https://github.com/FutureComputing4AI/ProxCSL}.

We consider a covariate-assisted ranking model grounded in the Plackett--Luce framework. Unlike existing works focusing on pure covariates or individual effects with fixed covariates, our approach integrates individual effects with dynamic covariates. This added flexibility enhances realistic ranking yet poses significant challenges for analyzing the associated estimation procedures. This paper makes an initial attempt to address these challenges. We begin by discussing the sufficient and necessary condition for the model's identifiability. We then introduce an efficient alternating maximization algorithm to compute the maximum likelihood estimator (MLE). Under suitable assumptions on the topology of comparison graphs and dynamic covariates, we establish a quantitative uniform consistency result for the MLE with convergence rates characterized by the asymptotic graph connectivity. The proposed graph topology assumption holds for several popular random graph models under optimal leading-order sparsity conditions. A comprehensive numerical study is conducted to corroborate our theoretical findings and demonstrate the application of the proposed model to real-world datasets, including horse racing and tennis competitions.

We revisit data selection in a modern context of finetuning from a fundamental perspective. Extending the classical wisdom of variance minimization in low dimensions to high-dimensional finetuning, our generalization analysis unveils the importance of additionally reducing bias induced by low-rank approximation. Inspired by the variance-bias tradeoff in high dimensions from the theory, we introduce Sketchy Moment Matching (SkMM), a scalable data selection scheme with two stages. (i) First, the bias is controlled using gradient sketching that explores the finetuning parameter space for an informative low-dimensional subspace $\mathcal{S}$; (ii) then the variance is reduced over $\mathcal{S}$ via moment matching between the original and selected datasets. Theoretically, we show that gradient sketching is fast and provably accurate: selecting $n$ samples by reducing variance over $\mathcal{S}$ preserves the fast-rate generalization $O(\dim(\mathcal{S})/n)$, independent of the parameter dimension. Empirically, we concretize the variance-bias balance via synthetic experiments and demonstrate the effectiveness of SkMM for finetuning in real vision tasks.