Adaptive Dense-to-Sparse Paradigm for Pruning Online Recommendation System with Non-Stationary Data

Large scale deep learning provides a tremendous opportunity to improve the quality of content recommendation systems by employing both wider and deeper models, but this comes at great infrastructural cost and carbon footprint in modern data centers. Pruning is an effective technique that reduces both memory and compute demand for model inference. However, pruning for online recommendation systems is challenging due to the continuous data distribution shift (a.k.a non-stationary data). Although incremental training on the full model is able to adapt to the non-stationary data, directly applying it on the pruned model leads to accuracy loss. This is because the sparsity pattern after pruning requires adjustment to learn new patterns. To the best of our knowledge, this is the first work to provide in-depth analysis and discussion of applying pruning to online recommendation systems with non-stationary data distribution. Overall, this work makes the following contributions: 1) We present an adaptive dense to sparse paradigm equipped with a novel pruning algorithm for pruning a large scale recommendation system with non-stationary data distribution; 2) We design the pruning algorithm to automatically learn the sparsity across layers to avoid repeating hand-tuning, which is critical for pruning the heterogeneous architectures of recommendation systems trained with non-stationary data.

Go Wide, Then Narrow: Efficient Training of Deep Thin Networks

For deploying a deep learning model into production, it needs to be both accurate and compact to meet the latency and memory constraints. This usually results in a network that is deep (to ensure performance) and yet thin (to improve computational efficiency). In this paper, we propose an efficient method to train a deep thin network with a theoretic guarantee. Our method is motivated by model compression. It consists of three stages. In the first stage, we sufficiently widen the deep thin network and train it until convergence. In the second stage, we use this well-trained deep wide network to warm up (or initialize) the original deep thin network. This is achieved by letting the thin network imitate the immediate outputs of the wide network from layer to layer. In the last stage, we further fine tune this well initialized deep thin network. The theoretical guarantee is established by using mean field analysis, which shows the advantage of layerwise imitation over traditional training deep thin networks from scratch by backpropagation. We also conduct large-scale empirical experiments to validate our approach. By training with our method, ResNet50 can outperform ResNet101, and BERT_BASE can be comparable with BERT_LARGE, where both the latter models are trained via the standard training procedures as in the literature.

SAFER: A Structure-free Approach for Certified Robustness to Adversarial Word Substitutions

State-of-the-art NLP models can often be fooled by human-unaware transformations such as synonymous word substitution. For security reasons, it is of critical importance to develop models with certified robustness that can provably guarantee that the prediction is can not be altered by any possible synonymous word substitution. In this work, we propose a certified robust method based on a new randomized smoothing technique, which constructs a stochastic ensemble by applying random word substitutions on the input sentences, and leverage the statistical properties of the ensemble to provably certify the robustness. Our method is simple and structure-free in that it only requires the black-box queries of the model outputs, and hence can be applied to any pre-trained models (such as BERT) and any types of models (world-level or subword-level). Our method significantly outperforms recent state-of-the-art methods for certified robustness on both IMDB and Amazon text classification tasks. To the best of our knowledge, we are the first work to achieve certified robustness on large systems such as BERT with practically meaningful certified accuracy.

Unsupervised Feature Selection via Multi-step Markov Transition Probability

Feature selection is a widely used dimension reduction technique to select feature subsets because of its interpretability. Many methods have been proposed and achieved good results, in which the relationships between adjacent data points are mainly concerned. But the possible associations between data pairs that are may not adjacent are always neglected. Different from previous methods, we propose a novel and very simple approach for unsupervised feature selection, named MMFS (Multi-step Markov transition probability for Feature Selection). The idea is using multi-step Markov transition probability to describe the relation between any data pair. Two ways from the positive and negative viewpoints are employed respectively to keep the data structure after feature selection. From the positive viewpoint, the maximum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. Then, the features which can keep the compact data structure are selected. From the viewpoint of negative, the minimum transition probability that can be reached in a certain number of steps is used to describe the relation between two points. On the contrary, the features that least maintain the loose data structure are selected. And the two ways can also be combined. Thus three algorithms are proposed. Our main contributions are a novel feature section approach which uses multi-step transition probability to characterize the data structure, and three algorithms proposed from the positive and negative aspects for keeping data structure. The performance of our approach is compared with the state-of-the-art methods on eight real-world data sets, and the experimental results show that the proposed MMFS is effective in unsupervised feature selection.

Improving Generalized Zero-Shot Learning by Semantic Discriminator

It is a recognized fact that the classification accuracy of unseen classes in the setting of Generalized Zero-Shot Learning (GZSL) is much lower than that of traditional Zero-Shot Leaning (ZSL). One of the reasons is that an instance is always misclassified to the wrong domain. Here we refer to the seen and unseen classes as two domains respectively. We propose a new approach to distinguish whether the instances come from the seen or unseen classes. First the visual feature of instance is projected into the semantic space. Then the absolute norm difference between the projected semantic vector and the class semantic embedding vector, and the minimum distance between the projected semantic vectors and the semantic embedding vectors of the seen classes are used as discrimination basis. This approach is termed as SD (Semantic Discriminator) because domain judgement of instance is performed in the semantic space. Our approach can be combined with any existing ZSL method and fully supervision classification model to form a new GZSL method. Furthermore, our approach is very simple and does not need any fixed parameters. A large number of experiments show that the accuracy of our approach is 8.5% to 21.9% higher than the current best method.

Disentanglement Then Reconstruction: Learning Compact Features for Unsupervised Domain Adaptation

Recent works in domain adaptation always learn domain invariant features to mitigate the gap between the source and target domains by adversarial methods. The category information are not sufficiently used which causes the learned domain invariant features are not enough discriminative. We propose a new domain adaptation method based on prototype construction which likes capturing data cluster centers. Specifically, it consists of two parts: disentanglement and reconstruction. First, the domain specific features and domain invariant features are disentangled from the original features. At the same time, the domain prototypes and class prototypes of both domains are estimated. Then, a reconstructor is trained by reconstructing the original features from the disentangled domain invariant features and domain specific features. By this reconstructor, we can construct prototypes for the original features using class prototypes and domain prototypes correspondingly. In the end, the feature extraction network is forced to extract features close to these prototypes. Our contribution lies in the technical use of the reconstructor to obtain the original feature prototypes which helps to learn compact and discriminant features. As far as we know, this idea is proposed for the first time. Experiment results on several public datasets confirm the state-of-the-art performance of our method.

Learning Various Length Dependence by Dual Recurrent Neural Networks

Recurrent neural networks (RNNs) are widely used as a memory model for sequence-related problems. Many variants of RNN have been proposed to solve the gradient problems of training RNNs and process long sequences. Although some classical models have been proposed, capturing long-term dependence while responding to short-term changes remains a challenge. To this problem, we propose a new model named Dual Recurrent Neural Networks (DuRNN). The DuRNN consists of two parts to learn the short-term dependence and progressively learn the long-term dependence. The first part is a recurrent neural network with constrained full recurrent connections to deal with short-term dependence in sequence and generate short-term memory. Another part is a recurrent neural network with independent recurrent connections which helps to learn long-term dependence and generate long-term memory. A selection mechanism is added between two parts to help the needed long-term information transfer to the independent neurons. Multiple modules can be stacked to form a multi-layer model for better performance. Our contributions are: 1) a new recurrent model developed based on the divide-and-conquer strategy to learn long and short-term dependence separately, and 2) a selection mechanism to enhance the separating and learning of different temporal scales of dependence. Both theoretical analysis and extensive experiments are conducted to validate the performance of our model, and we also conduct simple visualization experiments and ablation analyses for the model interpretability. Experimental results indicate that the proposed DuRNN model can handle not only very long sequences (over 5000 time steps), but also short sequences very well. Compared with many state-of-the-art RNN models, our model has demonstrated efficient and better performance.

Steepest Descent Neural Architecture Optimization: Escaping Local Optimum with Signed Neural Splitting

We propose signed splitting steepest descent (S3D), which progressively grows neural architectures by splitting critical neurons into multiple copies, following a theoretically-derived optimal scheme. Our algorithm is a generalization of the splitting steepest descent (S2D) of Liu et al. (2019b), but significantly improves over it by incorporating a rich set of new splitting schemes that allow negative output weights. By doing so, we can escape local optima that the original S2D can not escape. Theoretically, we show that our method provably learns neural networks with much smaller sizes than these needed for standard gradient descent in overparameterized regimes. Empirically, our method outperforms S2D and prior arts on various challenging benchmarks, including CIFAR-100, ImageNet and ModelNet40.

Joint COCO and Mapillary Workshop at ICCV 2019 Keypoint Detection Challenge Track Technical Report: Distribution-Aware Coordinate Representation for Human Pose Estimation

In this paper, we focus on the coordinate representation in human pose estimation. While being the standard choice, heatmap based representation has not been systematically investigated. We found that the process of coordinate decoding (i.e. transforming the predicted heatmaps to the coordinates) is surprisingly significant for human pose estimation performance, which nevertheless was not recognised before. In light of the discovered importance, we further probe the design limitations of the standard coordinate decoding method and propose a principled distribution-aware decoding method. Meanwhile, we improve the standard coordinate encoding process (i.e. transforming ground-truth coordinates to heatmaps) by generating accurate heatmap distributions for unbiased model training. Taking them together, we formulate a novel Distribution-Aware coordinate Representation for Keypoint (DARK) method. Serving as a model-agnostic plug-in, DARK significantly improves the performance of a variety of state-of-the-art human pose estimation models. Extensive experiments show that DARK yields the best results on COCO keypoint detection challenge, validating the usefulness and effectiveness of our novel coordinate representation idea. The project page containing more details is at https://ilovepose.github.io/coco