Hu Ding

Zixiu Wang, Yiwen Guo, Hu Ding

In many machine learning tasks, a common approach for dealing with large-scale data is to build a small summary, {\em e.g.,} coreset, that can efficiently represent the original input. However, real-world datasets usually contain outliers and most existing coreset construction methods are not resilient against outliers (in particular, an outlier can be located arbitrarily in the space by an adversarial attacker). In this paper, we propose a novel robust coreset method for the {\em continuous-and-bounded learning} problems (with outliers) which includes a broad range of popular optimization objectives in machine learning, {\em e.g.,} logistic regression and $ k $-means clustering. Moreover, our robust coreset can be efficiently maintained in fully-dynamic environment. To the best of our knowledge, this is the first robust and fully-dynamic coreset construction method for these optimization problems. Another highlight is that our coreset size can depend on the doubling dimension of the parameter space, rather than the VC dimension of the objective function which could be very large or even challenging to compute. Finally, we conduct the experiments on real-world datasets to evaluate the effectiveness of our proposed robust coreset method.

Guannan Hu, Wu Zhang, Hu Ding et al.

Catastrophic forgetting in continual learning is a common destructive phenomenon in gradient-based neural networks that learn sequential tasks, and it is much different from forgetting in humans, who can learn and accumulate knowledge throughout their whole lives. Catastrophic forgetting is the fatal shortcoming of a large decrease in performance on previous tasks when the model is learning a novel task. To alleviate this problem, the model should have the capacity to learn new knowledge and preserve learned knowledge. We propose an average gradient episodic memory (A-GEM) with a soft constraint $ε\in [0, 1]$, which is a balance factor between learning new knowledge and preserving learned knowledge; our method is called gradient episodic memory with a soft constraint $ε$ ($ε$-SOFT-GEM). $ε$-SOFT-GEM outperforms A-GEM and several continual learning benchmarks in a single training epoch; additionally, it has state-of-the-art average accuracy and efficiency for computation and memory, like A-GEM, and provides a better trade-off between the stability of preserving learned knowledge and the plasticity of learning new knowledge.

Hu Ding, Fan Yang, Jiawei Huang

Adversarial machine learning has attracted a great amount of attention in recent years. In a poisoning attack, the adversary can inject a small number of specially crafted samples into the training data which make the decision boundary severely deviate and cause unexpected misclassification. Due to the great importance and popular use of support vector machines (SVM), we consider defending SVM against poisoning attacks in this paper. We study two commonly used strategies for defending: designing robust SVM algorithms and data sanitization. Though several robust SVM algorithms have been proposed before, most of them either are in lack of adversarial-resilience, or rely on strong assumptions about the data distribution or the attacker's behavior. Moreover, the research on their complexities is still quite limited. We are the first, to the best of our knowledge, to prove that even the simplest hard-margin one-class SVM with outliers problem is NP-complete, and has no fully PTAS unless P$=$NP (that means it is hard to achieve an even approximate algorithm). For the data sanitization defense, we link it to the intrinsic dimensionality of data; in particular, we provide a sampling theorem in doubling metrics for explaining the effectiveness of DBSCAN (as a density-based outlier removal method) for defending against poisoning attacks. In our empirical experiments, we compare several defenses including the DBSCAN and robust SVM methods, and investigate the influences from the intrinsic dimensionality and data density to their performances.

Hu Ding, Ruizhe Qin, Jiawei Huang

In this paper, we consider robust optimization problems in high dimensions. Because a real-world dataset may contain significant noise or even specially crafted samples from some attacker, we are particularly interested in the optimization problems with arbitrary (and potentially adversarial) outliers. We focus on two fundamental optimization problems: {\em SVM with outliers} and {\em $k$-center clustering with outliers}. They are in fact extremely challenging combinatorial optimization problems, since we cannot impose any restriction on the adversarial outliers. Therefore, their computational complexities are quite high especially when we consider the instances in high dimensional spaces. The {\em Johnson-Lindenstrauss (JL) Transform} is one of the most popular methods for dimension reduction. Though the JL transform has been widely studied in the past decades, its effectiveness for dealing with adversarial outliers has never been investigated before (to the best of our knowledge). Based on some novel insights from the geometry, we prove that the complexities of these two problems can be significantly reduced through the JL transform. Moreover, we prove that the solution in the dimensionality-reduced space can be efficiently recovered in the original $\mathbb{R}^d$ while the quality is still preserved. In the experiments, we compare JL transform with several other well known dimension reduction methods, and study their performances on synthetic and real datasets.

Hu Ding, Jiawei Huang, Haikuo Yu

Clustering has many important applications in computer science, but real-world datasets often contain outliers. Moreover, the presence of outliers can make the clustering problems to be much more challenging. To reduce the complexities, various sampling methods have been proposed in past years. Namely, we take a small sample (uniformly or non-uniformly) from input and run an existing approximation algorithm on the sample. Comparing with existing non-uniform sampling methods, the uniform sampling approach has several significant benefits. For example, it only needs to read the data in one-pass and is very easy to implement in practice. Thus, the effectiveness of uniform sampling for clustering with outliers is a natural and fundamental problem deserving to study in both theory and practice. In this paper, we propose a new and unified framework for analyzing the effectiveness of uniform sampling for three representative center-based clustering with outliers problems, $k$-center/median/means clustering with outliers. We introduce a "significance" criterion and prove that the performance of uniform sampling depends on the significance degree of the given instance. In particular, we show that the sample size can be independent of the ratio $n/z$ and the dimensionality. More importantly, to the best of our knowledge, our method is the first uniform sampling approach that allows to discard exactly $z$ outliers for these three center-based clustering with outliers problems. The results proposed in this paper also can be viewed as an extension of the previous sub-linear time algorithms for the ordinary clustering problems (without outliers). The experiments suggest that the uniform sampling method can achieve comparable clustering results with other existing methods, but greatly reduce the running times.

Hu Ding

In this paper, we revisit the Minimum Enclosing Ball (MEB) problem and its robust version, MEB with outliers, in Euclidean space $\mathbb{R}^d$. Though the problem has been extensively studied before, most of the existing algorithms need at least linear time (in the number of input points $n$ and the dimensionality $d$) to achieve a $(1+ε)$-approximation. Motivated by some recent developments on beyond worst-case analysis, we introduce the notion of stability for MEB (with outliers), which is natural and easy to understand. Roughly speaking, an instance of MEB is stable, if the radius of the resulting ball cannot be significantly reduced by removing a small fraction of the input points. Under the stability assumption, we present two sampling algorithms for computing approximate MEB with sample complexities independent of the number of input points $n$. In particular, the second algorithm has the sample complexity even independent of the dimensionality $d$. Further, we extend the idea to achieve a sub-linear time approximation algorithm for the MEB with outliers problem. Note that most existing sub-linear time algorithms for the problems of MEB and MEB with outliers usually result in bi-criteria approximations, where the "bi-criteria" means that the solution has to allow the approximations on the radius and the number of covered points. Differently, all the algorithms proposed in this paper yield single-criterion approximations (with respect to radius). We expect that our proposed notion of stability and techniques will be applicable to design sub-linear time algorithms for other optimization problems.

Hu Ding, Haikuo Yu, Zixiu Wang

We study the problem of $k$-center clustering with outliers in arbitrary metrics and Euclidean space. Though a number of methods have been developed in the past decades, it is still quite challenging to design quality guaranteed algorithm with low complexity for this problem. Our idea is inspired by the greedy method, Gonzalez's algorithm, for solving the problem of ordinary $k$-center clustering. Based on some novel observations, we show that this greedy strategy actually can handle $k$-center clustering with outliers efficiently, in terms of clustering quality and time complexity. We further show that the greedy approach yields small coreset for the problem in doubling metrics, so as to reduce the time complexity significantly. Our algorithms are easy to implement in practice. We test our method on both synthetic and real datasets. The experimental results suggest that our algorithms can achieve near optimal solutions and yield lower running times comparing with existing methods.

Hu Ding, Mingquan Ye

In real-world, many problems can be formulated as the alignment between two geometric patterns. Previously, a great amount of research focus on the alignment of 2D or 3D patterns, especially in the field of computer vision. Recently, the alignment of geometric patterns in high dimension finds several novel applications, and has attracted more and more attentions. However, the research is still rather limited in terms of algorithms. To the best of our knowledge, most existing approaches for high dimensional alignment are just simple extensions of their counterparts for 2D and 3D cases, and often suffer from the issues such as high complexities. In this paper, we propose an effective framework to compress the high dimensional geometric patterns and approximately preserve the alignment quality. As a consequence, existing alignment approach can be applied to the compressed geometric patterns and thus the time complexity is significantly reduced. Our idea is inspired by the observation that high dimensional data often has a low intrinsic dimension. We adopt the widely used notion "doubling dimension" to measure the extents of our compression and the resulting approximation. Finally, we test our method on both random and real datasets, the experimental results reveal that running the alignment algorithm on compressed patterns can achieve similar qualities, comparing with the results on the original patterns, but the running times (including the times cost for compression) are substantially lower.

Hu Ding, Jinhui Xu

In this paper, we consider a class of constrained clustering problems of points in $\mathbb{R}^{d}$, where $d$ could be rather high. A common feature of these problems is that their optimal clusterings no longer have the locality property (due to the additional constraints), which is a key property required by many algorithms for their unconstrained counterparts. To overcome the difficulty caused by the loss of locality, we present in this paper a unified framework, called {\em Peeling-and-Enclosing (PnE)}, to iteratively solve two variants of the constrained clustering problems, {\em constrained $k$-means clustering} ($k$-CMeans) and {\em constrained $k$-median clustering} ($k$-CMedian). Our framework is based on two standalone geometric techniques, called {\em Simplex Lemma} and {\em Weaker Simplex Lemma}, for $k$-CMeans and $k$-CMedian, respectively. The simplex lemma (or weaker simplex lemma) enables us to efficiently approximate the mean (or median) point of an unknown set of points by searching a small-size grid, independent of the dimensionality of the space, in a simplex (or the surrounding region of a simplex), and thus can be used to handle high dimensional data. If $k$ and $\frac{1}ε$ are fixed numbers, our framework generates, in nearly linear time ({\em i.e.,} $O(n(\log n)^{k+1}d)$), $O((\log n)^{k})$ $k$-tuple candidates for the $k$ mean or median points, and one of them induces a $(1+ε)$-approximation for $k$-CMeans or $k$-CMedian, where $n$ is the number of points. Combining this unified framework with a problem-specific selection algorithm (which determines the best $k$-tuple candidate), we obtain a $(1+ε)$-approximation for each of the constrained clustering problems. We expect that our technique will be applicable to other constrained clustering problems without locality.

Hu Ding

The problem of constrained clustering has attracted significant attention in the past decades. In this paper, we study the balanced $k$-center, $k$-median, and $k$-means clustering problems where the size of each cluster is constrained by the given lower and upper bounds. The problems are motivated by the applications in processing large-scale data in high dimension. Existing methods often need to compute complicated matchings (or min cost flows) to satisfy the balance constraint, and thus suffer from high complexities especially in high dimension. We develop an effective framework for the three balanced clustering problems to address this issue, and our method is based on a novel spatial partition idea in geometry. For the balanced $k$-center clustering, we provide a $4$-approximation algorithm that improves the existing approximation factors; for the balanced $k$-median and $k$-means clusterings, our algorithms yield constant and $(1+ε)$-approximation factors with any $ε>0$. More importantly, our algorithms achieve linear or nearly linear running times when $k$ is a constant, and significantly improve the existing ones. Our results can be easily extended to metric balanced clusterings and the running times are sub-linear in terms of the complexity of $n$-point metric.