9 October 2023 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Federated learning (FL) allows multiple clients to collaboratively learn a globally shared model through cycles of model aggregation and local model training, without the need to share data. Most existing FL methods train local models separately on different clients, and then simply average their parameters to obtain a centralized model on the server side. However, these approaches generally suffer from large aggregation errors and severe local forgetting, which are particularly bad in heterogeneous data settings. To tackle these issues, in this paper, we propose a novel FL framework that uses online Laplace approximation to approximate posteriors on both the client and server side. On the server side, a multivariate Gaussian product mechanism is employed to construct and maximize a global posterior, largely reducing the aggregation errors induced by large discrepancies between local models. On the client side, a prior loss that uses the global posterior probabilistic parameters delivered from the server is designed to guide the local training. Binding such learning constraints from other clients enables our method to mitigate local forgetting. Finally, we achieve state-of-the-art results on several benchmarks, clearly demonstrating the advantages of the proposed method.
23 July 2020 in IEEE Transactions on Neural Networks and Learning Systems
Sparse additive models have been successfully applied to high-dimensional data analysis due to the flexibility and interpretability of their representation. However, the existing methods are often formulated using the least-squares loss with learning the conditional mean, which is sensitive to data with the non-Gaussian noises, e.g., skewed noise, heavy-tailed noise, and outliers. To tackle this problem, we propose a new robust regression method, called as sparse modal additive model (SpMAM), by integrating the modal regression metric, the data-dependent hypothesis space, and the weighted ℓq,1-norm regularizer (q ≥ 1) into the additive models. Specifically, the modal regression metric assures the model robustness to complex noises via learning the conditional mode, the data-dependent hypothesis space offers the model adaptivity via sample-based presentation, and the ℓq,1-norm regularizer addresses the algorithmic interpretability via sparse variable selection. In theory, the proposed SpMAM enjoys statistical guarantees on asymptotic consistency for regression estimation and variable selection simultaneously. Experimental results on both synthetic and real-world benchmark data sets validate the effectiveness and robustness of the proposed model.
1 October 2020 in IEEE Transactions on Cybernetics
This article presents a generalized collaborative representation-based classification (GCRC) framework, which includes many existing representation-based classification (RC) methods, such as collaborative RC (CRC) and sparse RC (SRC) as special cases. This article also advances the GCRC theory by exploring theoretical conditions on the general regularization matrix. A key drawback of CRC and SRC is that they fail to use the label information of training data and are essentially unsupervised in computing the representation vector. This largely compromises the discriminative ability of the learned representation vector and impedes the classification performance. Guided by the GCRC theory, we propose a novel RC method referred to as discriminative RC (DRC). The proposed DRC method has the following three desirable properties: 1) discriminability: DRC can leverage the label information of training data and is supervised in both representation and classification, thus improving the discriminative ability of the representation vector; 2) efficiency: it has a closed-form solution and is efficient in computing the representation vector and performing classification; and 3) theory: it also has theoretical guarantees for classification. Experimental results on benchmark databases demonstrate both the efficacy and efficiency of DRC for multiclass classification.
1 January 2018 in Applied and Computational Harmonic Analysis
21 March 2019 in IEEE Transactions on Cybernetics
Representation-based classification (RC) methods, such as sparse RC, have shown great potential in face recognition (FR) in recent years. Most previous RC methods are based on the conventional regression models, such as lasso regression, ridge regression, or group lasso regression. These regression models essentially impose a predefined assumption on the distribution of the noise variable in the query sample, such as the Gaussian or Laplacian distribution. However, the complicated noises in practice may violate the assumptions and impede the performance of these RC methods. In this paper, we propose a modal regression (MR)-based atomic representation and classification (MRARC) framework to alleviate such limitations. MR is a robust regression framework which aims to reveal the relationship between the input and response variables by regressing toward the conditional mode function. Atomic representation is a general atomic norm regularized linear representation framework which includes many popular representation methods, such as sparse representation, collaborative representation, and low-rank representation as special cases. Unlike previous RC methods, the MRARC framework does not require the noise variable to follow any specific predefined distributions. This gives rise to the capability of MRARC in handling various complex noises in reality. Using MRARC as a general platform, we also develop four novel RC methods for unimodal and multimodal FR, respectively. In addition, we devise a general optimization algorithm for the unified MRARC framework based on the alternating direction method of multipliers and half-quadratic theory. The experiments on real-world data validate the efficacy of MRARC for robust FR and reconstruction.
1 April 2013 in Neural Computation
In this letter, we consider a density-level detection (DLD) problem by a coefficient-based classification framework with [Formula: see text]-regularizer and data-dependent hypothesis spaces. Although the data-dependent characteristic of the algorithm provides flexibility and adaptivity for DLD, it leads to difficulty in generalization error analysis. To overcome this difficulty, an error decomposition is introduced from an established classification framework. On the basis of this decomposition, the estimate of the learning rate is obtained by using Rademacher average and stepping-stone techniques. In particular, the estimate is independent of the capacity assumption used in the previous literature.