However, the pretrained model may not provide an ideal optimization starting point for the test-time optimization.

Abstract--Domain Incremental Learning (DIL) is a continual learning sub-branch that aims to address never-ending arrivals of new domains without catastrophic forgetting problems. Despite the advent of parameter-efficient fine-tuning (PEFT) approaches, existing works create task-specific LoRAs overlooking shared knowledge across tasks. Inaccurate selection of task-specific LORAs during inference results in significant drops in accuracy, while existing works rely on linear or prototype-based classifiers, which have suboptimal generalization powers. Our paper proposes continual knowledge consolidation low rank adaptation (CONEC-LoRA) addressing the DIL problems. CONEC-LoRA is developed from consolidations between task-shared LORA to extract common knowledge and task-specific LORA to embrace domain-specific knowledge. Unlike existing approaches, CONEC-LoRA integrates the concept of a stochastic classifier whose parameters are sampled from a distribution, thus enhancing the likelihood of correct classifications. Last but not least, an auxiliary network is deployed to optimally predict the task-specific LoRAs for inferences and implements the concept of a different-depth network structure in which every layer is connected with a local classifier to take advantage of intermediate representations. This module integrates the ball-generator loss and transformation module to address the synthetic sample bias problem. Our rigorous experiments demonstrate the advantage of CONEC-LoRA over prior arts in 4 popular benchmark problems with over 5% margins. ONTINUAL learning (CL) constitutes a research area of growing interests where the main goal is to develop a learning agent that can accumulate knowledge overtime [1], [2], [3], [4].

High-frequency physiological waveform modality offers deep, real-time insights into patient status. Recently, physiological foundation models based on Photoplethysmography (PPG), such as PPG-GPT, have been shown to predict critical events, including Cardiac Arrest (CA). However, their powerful representation still needs to be leveraged suitably, especially when the downstream data/label is scarce. We offer three orthogonal improvements to improve PPG-only CA systems by using minimal auxiliary information. First, we propose to use time-to-event modeling, either through simple regression to the event onset time or by pursuing fine-grained discrete survival modeling. Second, we encourage the model to learn CA-focused features by making them patient-identity invariant. This is achieved by first training the largest-scale de-identified biometric identification model, referred to as the p-vector, and subsequently using it adversarially to deconfound cues, such as person identity, that may cause overfitting through memorization. Third, we propose regression on the pseudo-lab values generated by pre-trained auxiliary estimator networks. This is crucial since true blood lab measurements, such as lactate, sodium, troponin, and potassium, are collected sparingly. Via zero-shot prediction, the auxiliary networks can enrich cardiac arrest waveform labels and generate pseudo-continuous estimates as targets. Our proposals can independently improve the 24-hour time-averaged AUC from the 0.74 to the 0.78-0.80 range. We primarily improve over longer time horizons with minimal degradation near the event, thus pushing the Early Warning System research. Finally, we pursue multi-task formulation and diagnose it with a high gradient conflict rate among competing losses, which we alleviate via the PCGrad optimization technique.

--Federated learning (FL) is one of the popular distributed machine learning (ML) solutions but incurs significant communication and computation costs at edge devices. Federated split learning (FSL) can train sub-models in parallel and reduce the computational burden of edge devices by splitting the model architecture. However, it still requires a high communication overhead due to transmitting the smashed data and gradients between clients and the server in every global round. Furthermore, the server must maintain separate partial models for every client, leading to a significant storage requirement. T o address these challenges, this paper proposes a novel communication and storage efficient federated split learning method, termed CSE-FSL, which utilizes an auxiliary network to locally update the weights of the clients while keeping a single model at the server, hence avoiding frequent transmissions of gradients from the server and greatly reducing the storage requirement of the server . Additionally, a new model update method of transmitting the smashed data in selected epochs can reduce the amount of smashed data sent from the clients. We provide a theoretical analysis of CSE-FSL, rigorously guaranteeing its convergence under non-convex loss functions. The extensive experimental results further indicate that CSE-FSL achieves a significant communication reduction over existing FSL solutions using real-world FL tasks. S an emerging distributed machine learning (ML) paradigm, federated learning (FL) [2] enables distributed clients to collaboratively train ML models without uploading their sensitive data to a central server. While FL addresses privacy concerns by keeping data localized, most existing FL approaches assume that clients possess sufficient computational and storage resources to perform local updates on large (potentially deep) models. However, this assumption breaks down in scenarios where clients, such as mobile and Internet-of-Things (IoT) devices, are resource-constrained. Consequently, these clients struggle to handle the heavy computational and storage demands of training deep ML models, making FL impractical in such settings. To address this issue, split learning (SL) [3]-[5] is proposed.

Knowledge distillation (KD) has become a powerful tool for training compact student models using larger, pretrained teacher models, often requiring less data and computational resources. Teacher models typically possess more layers and thus exhibit richer feature representations compared to their student counterparts. Furthermore, student models tend to learn simpler, surface-level features in their early layers. This discrepancy can increase errors in groups where labels spuriously correlate with specific input attributes, leading to a decline in group fairness even when overall accuracy remains comparable to the teacher. To mitigate these challenges, Early-Exit Neural Networks (EENNs), which enable predictions at multiple intermediate layers, have been employed. Confidence margins derived from these early exits have been utilized to reweight both cross-entropy and distillation losses on a per-instance basis. In this paper, we propose that leveraging Laplace approximation-based methods to obtain well-calibrated uncertainty estimates can also effectively reweight challenging instances and improve group fairness. We hypothesize that Laplace approximation offers a more robust identification of difficult or ambiguous instances compared to margin-based approaches. To validate our claims, we benchmark our approach using a Bert-based model on the MultiNLI dataset.

Recently, the rapid development of LEO satellite networks spurs another widespread concern-data processing at satellites. However, achieving efficient computation at LEO satellites in highly dynamic satellite networks is challenging and remains an open problem when considering the constrained computation capability of LEO satellites. For the first time, we propose a novel distributed learning framework named SFL-LEO by combining Federated Learning (FL) with Split Learning (SL) to accommodate the high dynamics of LEO satellite networks and the constrained computation capability of LEO satellites by leveraging the periodical orbit traveling feature. The proposed scheme allows training locally by introducing an asynchronous training strategy, i.e., achieving local update when LEO satellites disconnect with the ground station, to provide much more training space and thus increase the training performance. Meanwhile, it aggregates client-side sub-models at the ground station and then distributes them to LEO satellites by borrowing the idea from the federated learning scheme. Experiment results driven by satellite-ground bandwidth measured in Starlink demonstrate that SFL-LEO provides a similar accuracy performance with the conventional SL scheme because it can perform local training even within the disconnection duration.

Target-speaker speech processing (TS) tasks, such as target-speaker automatic speech recognition (TS-ASR), target speech extraction (TSE), and personal voice activity detection (p-VAD), are important for extracting information about a desired speaker's speech even when it is corrupted by interfering speakers. While most studies have focused on training schemes or system architectures for each specific task, the auxiliary network for embedding target-speaker cues has not been investigated comprehensively in a unified cross-task evaluation. Therefore, this paper aims to address a fundamental question: what is the preferred speaker embedding for TS tasks? To this end, for the TS-ASR, TSE, and p-VAD tasks, we compare pre-trained speaker encoders (i.e., self-supervised or speaker recognition models) that compute speaker embeddings from pre-recorded enrollment speech of the target speaker with ideal speaker embeddings derived directly from the target speaker's identity in the form of a one-hot vector. To further understand the properties of ideal speaker embedding, we optimize it using a gradient-based approach to improve performance on the TS task. Our analysis reveals that speaker verification performance is somewhat unrelated to TS task performances, the one-hot vector outperforms enrollment-based ones, and the optimal embedding depends on the input mixture.

Partial-Label Learning (PLL) is a typical problem of weakly supervised learning, where each training instance is annotated with a set of candidate labels. Self-training PLL models achieve state-of-the-art performance but suffer from error accumulation problem caused by mistakenly disambiguated instances. Although co-training can alleviate this issue by training two networks simultaneously and allowing them to interact with each other, most existing co-training methods train two structurally identical networks with the same task, i.e., are symmetric, rendering it insufficient for them to correct each other due to their similar limitations. Therefore, in this paper, we propose an asymmetric dual-task co-training PLL model called AsyCo, which forces its two networks, i.e., a disambiguation network and an auxiliary network, to learn from different views explicitly by optimizing distinct tasks. Specifically, the disambiguation network is trained with self-training PLL task to learn label confidence, while the auxiliary network is trained in a supervised learning paradigm to learn from the noisy pairwise similarity labels that are constructed according to the learned label confidence. Finally, the error accumulation problem is mitigated via information distillation and confidence refinement. Extensive experiments on both uniform and instance-dependent partially labeled datasets demonstrate the effectiveness of AsyCo. The code is available at https://github.com/libeibeics/AsyCo.