Recent studies indicate that deep neural networks (DNNs) are prone to show discrimination towards certain demographic groups. We observe that algorithmic discrimination can be explained by the high reliance of the models on fairness sensitive features. Motivated by this observation, we propose to achieve fairness by suppressing the DNN models from capturing the spurious correlation between those fairness sensitive features with the underlying task. Specifically, we firstly train a bias-only teacher model which is explicitly encouraged to maximally employ fairness sensitive features for prediction. The teacher model then counter-teaches a debiased student model so that the interpretation of the student model is orthogonal to the interpretation of the teacher model. The key idea is that since the teacher model relies explicitly on fairness sensitive features for prediction, the orthogonal interpretation loss enforces the student network to reduce its reliance on sensitive features and instead capture more task relevant features for prediction. Experimental analysis indicates that our framework substantially reduces the model's attention on fairness sensitive features. Experimental results on four datasets further validate that our framework has consistently improved the fairness with respect to three group fairness metrics, with a comparable or even better accuracy.

Over the last decade, deep neural networks (DNNs) are regarded as black-box methods, and their decisions are criticized for the lack of explainability. Existing attempts based on local explanations offer each input a visual saliency map, where the supporting features that contribute to the decision are emphasized with high relevance scores. In this paper, we improve the saliency map based on differentiated explanations, of which the saliency map not only distinguishes the supporting features from backgrounds but also shows the different degrees of importance of the various parts within the supporting features. To do this, we propose to learn a differentiated relevance estimator called DRE, where a carefully-designed distribution controller is introduced to guide the relevance scores towards right-skewed distributions. DRE can be directly optimized under pure classification losses, enabling higher faithfulness of explanations and avoiding non-trivial hyper-parameter tuning. The experimental results on three real-world datasets demonstrate that our differentiated explanations significantly improve the faithfulness with high explainability. Our code and trained models are available at https://github.com/fuweijie/DRE.

Knowledge graphs have received intensive research interests. When the labels of most nodes or datapoints are missing, anchor graph and hierarchical anchor graph models can be employed. With an anchor graph or hierarchical anchor graph, we only need to optimize the labels of the coarsest anchors, and the labels of datapoints can be inferred from these anchors in a coarse-to-fine manner. The complexity of optimization is therefore reduced to a cubic cost with respect to the number of the coarsest anchors. However, to obtain a high accuracy when a data distribution is complex, the scale of this anchor set still needs to be large, which thus inevitably incurs an expensive computational burden. As such, a challenge in scaling up these models is how to efficiently estimate the labels of these anchors while keeping classification performance. To address this problem, we propose a novel approach that adds an anchor label predictor in the conventional anchor graph and hierarchical anchor graph models. In the proposed approach, the labels of the coarsest anchors are not directly optimized, and instead, we learn a label predictor which estimates the labels of these anchors with their spectral representations. The predictor is optimized with a regularization on all datapoints based on a hierarchical anchor graph, and we show that its solution only involves the inversion of a small-size matrix. Built upon the anchor hierarchy, we design a sparse intra-layer adjacency matrix over these anchors, which can simultaneously accelerate spectral embedding and enhance effectiveness. Our approach is named Faster Learning on Anchor Graph (FLAG) as it improves conventional anchor-graph-based methods in terms of efficiency. Experiments on a variety of publicly available datasets with sizes varying from thousands to millions of samples demonstrate the effectiveness of our approach.

Machine learning can provide deep insights into data, allowing machines to make high-quality predictions and having been widely used in real-world applications, such as text mining, visual classification, and recommender systems. However, most sophisticated machine learning approaches suffer from huge time costs when operating on large-scale data. This issue calls for the need of Large-scale Machine Learning (LML), which aims to learn patterns from big data with comparable performance efficiently. In this paper, we offer a systematic survey on existing LML methods to provide a blueprint for the future developments of this area. We first divide these LML methods according to the ways of improving the scalability: 1) model simplification on computational complexities, 2) optimization approximation on computational efficiency, and 3) computation parallelism on computational capabilities. Then we categorize the methods in each perspective according to their targeted scenarios and introduce representative methods in line with intrinsic strategies. Lastly, we analyze their limitations and discuss potential directions as well as open issues that are promising to address in the future.

We study the problem of active learning for multi-class classification on large-scale datasets. In this setting, the existing active learning approaches built upon uncertainty measures are ineffective for discovering unknown regions, and those based on expected error reduction are inefficient owing to their huge time costs. To overcome the above issues, this paper proposes a novel query selection criterion called approximated error reduction (AER). In AER, the error reduction of each candidate is estimated based on an expected impact over all datapoints and an approximated ratio between the error reduction and the impact over its nearby datapoints. In particular, we utilize hierarchical anchor graphs to construct the candidate set as well as the nearby datapoint sets of these candidates. The benefit of this strategy is that it enables a hierarchical expansion of candidates with the increase of labels, and allows us to further accelerate the AER estimation. We finally introduce AER into an efficient semi-supervised classifier for scalable active learning. Experiments on publicly available datasets with the sizes varying from thousands to millions demonstrate the effectiveness of our approach.

Several models have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). The AGR approach significantly accelerates graph-based learning by exploring a set of anchors. However, when a dataset becomes much larger, AGR still faces a big graph which brings dramatically increasing computational costs. To overcome this issue, we propose a novel Hierarchical Anchor Graph Regularization (HAGR) approach by exploring multiple-layer anchors with a pyramid-style structure. In HAGR, the labels of datapoints are inferred from the coarsest anchors layer by layer in a coarse-to-fine manner. The label smoothness regularization is performed on all datapoints, and we demonstrate that the optimization process only involves a small-size reduced Laplacian matrix. We also introduce a fast approach to construct our hierarchical anchor graph based on an approximate nearest neighbor search technique. Experiments on million-scale datasets demonstrate the effectiveness and efficiency of the proposed HAGR approach over existing methods. Results show that the HAGR approach is even able to achieve a good performance within 3 minutes in an 8-million-example classification task.

Many graph-based semi-supervised learning methods for large datasets have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). This model builds a regularization framework by exploring the underlying structure of the whole dataset with both datapoints and anchors. Nevertheless, AGR still has limitations in its two components: (1) in anchor graph construction, the estimation of the local weights between each datapoint and its neighboring anchors could be biased and relatively slow; and (2) in anchor graph regularization, the adjacency matrix that estimates the relationship between datapoints, is not sufficiently effective. In this paper, we develop an Efficient Anchor Graph Regularization (EAGR) by tackling these issues. First, we propose a fast local anchor embedding method, which reformulates the optimization of local weights and obtains an analytical solution. We show that this method better reconstructs datapoints with anchors and speeds up the optimizing process. Second, we propose a new adjacency matrix among anchors by considering the commonly linked datapoints, which leads to a more effective normalized graph Laplacian over anchors. We show that, with the novel local weight estimation and normalized graph Laplacian, EAGR is able to achieve better classification accuracy with much less computational costs. Experimental results on several publicly available datasets demonstrate the effectiveness of our approach.