Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with L0 regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness.

The exposure sequence is being actively studied for user interest modeling in Click-Through Rate (CTR) prediction. However, the existing methods for exposure sequence modeling bring extensive computational burden and neglect noise problems, resulting in an excessively latency and the limited performance in online recommenders. In this paper, we propose to address the high latency and noise problems via Gating-adapted wavelet multiresolution analysis (Gama), which can effectively denoise the extremely long exposure sequence and adaptively capture the implied multi-dimension user interest with linear computational complexity. This is the first attempt to integrate non-parametric multiresolution analysis technique into deep neural network to model user exposure sequence. Extensive experiments on large scale benchmark dataset and real production dataset confirm the effectiveness of Gama for exposure sequence modeling, especially in cold-start scenarios. Benefited from its low latency and high effecitveness, Gama has been deployed in our real large-scale industrial recommender, successfully serving over hundreds of millions users.

As the source of side information, knowledge graph (KG) plays a critical role in recommender systems. Recently, graph neural networks (GNN) have shown their technical advancements at boosting recommendation performances. Existing GNN-based models mainly focus on aggregation technique and regularization allocation, ignoring the rich entity-aware information hidden in the relation network of KG. In this paper, we explore the relational semantics at the granularity of entities behind a user-item interaction by leveraging knowledge graph, named Entity-aware Collaborative Relation Network (ECRN). Technically, we construct multiple meta-paths from users to entities based on the user-item interaction and item-entity connectivity to obtain user representation, while designing a relation-aware self-attention mechanism to aggregate collaborative signals of items. Empirical results on three benchmarks show that ECRN significantly outperforms state-of-the-art baselines.

Although lots of arrhythmia classification models based on deep neural networks have been proposed, most of them can only be directly applied to inputs of a fixed length, which means the raw ECG records need to be padded or truncated before being put into the model. However, this process brings two main drawbacks: truncation may lead to information loss while padding increases calculation load. Besides, different sampling rates in other datasets may add trouble to transfer learning in this case. To address these problems, we propose a length adaptive arrhythmia classification model that can take advantage of raw ECG records of variable length. To further improve the overall performance, this model is designed with multi-scale convolution networks. We then introduce the dilated convolution that enables small convolution kernels to replace bigger traditional ones so as to enlarge feature’s reception fields and reduce training parameters. Finally, an attention mechanism is implemented so that the model can get efficiently trained and output enhanced results. Extensive experiments on the benchmark MIT-BIH database prove that our method is competitive with other state-of-the-arts.

Personalized tag recommender systems recommend a series of tags for items by leveraging users' historical records, which helps tag-aware recommender systems (TRS) to better depict user profiles and item characteristics. However, existing personalized tag recommendation solutions are insufficient to capture the collaborative signal hidden in the interactions among entities without considering reasonable correlations, since neighborhood messages are treated as the same weights when constructing graph-structured data, resulting in decreased accuracy in making recommendations. In this paper, we propose a Tag-aware Attentional Graph Neural Network (TA-GNN), which integrates the attention mechanism into tag-based graph neural networks to alleviate the above issues. Specifically, we extract the user-tag interaction and the item-tag interaction from the user-tag-item graph structure. For each interaction, we exploit the contextual semantics of multi-hop neighbors by leveraging attentional strategy on graph neural networks to discriminate the importance of different connected nodes. In this way, we effectively extract collaborative signals of neighborhood representations and capture the potential information in an explicit manner. Extensive experiments on three public datasets show that our proposed TA-GNN outperforms the state-of-the-art personalized tag recommendation baselines.

Session-based recommendation aims to suggest items to users mainly by modeling sequential dependencies. However, existing methods mainly focus on resorting either temporal interests with time-series data or inherent interests by leveraging static user behaviors, ignoring benefits of the dynamic aggregation from short- and long-term preferences, resulting in unfavorable performance in various sequence recommendation scenarios. In this paper, we attempt to combine users' short interests with long preferences synthetically, and tackle the sequential recommendation problem from new views of the causal (short) and non-causal (long) perspectives. We propose a dynamic gap-complementary generative framework named DGRec for session-based Recommendation. Specifically, we fuse two practical neural network models: the dilated convolutional neural network (DCNN) and the self-attention block (SAB). DCNN mainly captures users' short-term interests while SAB is responsible for obtaining users' long-term inherent behaviors. Besides, we devise a gap-complementary strategy to dynamically learn item sequence representation, effectively enhancing the robustness and anti-interference ability of model. Extensive experimental results over five public datasets demonstrate that DGRec gains significant improvements against the advanced sequential recommendation methods.

Deep neural networks (DNNs) have shown tremendous success in many areas, such as signal processing, computer vision, and artificial intelligence. However, the DNNs require intensive computation resources, hindering their practical applications on the edge devices with limited storage and computation resources. Filter pruning has been recognized as a useful technique to compress and accelerate the DNNs, but most existing works tend to prune filters in a layerwise manner, facing some significant drawbacks. First, the layerwise pruning methods require prohibitive computation for per-layer sensitivity analysis. Second, layerwise pruning suffers from the accumulation of pruning errors, leading to performance degradation of pruned networks. To address these challenges, we propose a novel global pruning method, namely, EasiEdge, to compress and accelerate the DNNs for efficient edge computing. More specifically, we introduce an alternating direction method of multipliers (ADMMs) to formulate the pruning problem as a performance improving subproblem and a global pruning subproblem. In the global pruning subproblem, we propose to use information gain (IG) to quantify the impact of filters removal on the class probability distributions of network output. Besides, we propose a Taylor-based approximate algorithm (TBAA) to efficiently calculate the IG of filters. Extensive experiments on three data sets and two edge computing platforms verify that our proposed EasiEdge can efficiently accelerate DNNs on edge computing platforms with nearly negligible accuracy loss. For example, when EasiEdge prunes 80% filters in VGG-16, the accuracy drops by 0.22%, but inference latency on CPU of Jetson TX2 decreases from 76.85 to 8.01 ms.

Blood pressure is a key indicator of personal health. In this paper, we propose a novel integrated deep learning method which can accurately determine blood pressure levels under inter-subject scenario without initial calibration. In detail, a convolutional neural network is first introduced to extract features from the raw photoplethysmogram signals. After that, we concatenate the obtained features with personal BMI information and use them as the input of two independent neural networks, which output the estimated blood pressure values and the predicted hypertension class, respectively. These two outputs, in the end, are integrated assessed to generate the final result. Comprehensive experiments demonstrate that our method achieves highly competitive performance compared with others.

In an ECG record, the PQRST waves are of important medical significance which provide ample information reflecting heartbeat activities. In this paper, we propose a novel arrhythmia classification method namely EasiECG, characterized by simplicity and accuracy. Compared with other works, the EasiECG takes the configuration of these five key waves into account and does not require complicated feature engineering. Meanwhile, an additional encoding of the extracted features makes the EasiECG applicable even on samples with missing waves. To automatically capture interactions that contribute to the classification among the processed features, a novel adapted classification model named Attention-based Convolution Factorization Machines (ACFM) is proposed. In detail, the ACFM can learn both linear and high-order interactions from linear regression and convolution on outer-product feature interaction maps, respectively. After that, an attention mechanism implemented in the model can further assign different importance of these interactions when predicting certain types of heartbeats. To validate the effectiveness and practicability of our EasiECG, extensive experiments of inter-patient paradigm on the benchmark MIT-BIH arrhythmia database are conducted. To tackle the imbalanced sample problem in this dataset, an ingenious loss function: focal loss is adopted when training. The experiment results show that our method is competitive compared with other state-of-the-arts, especially in classifying the Supraventricular ectopic beats. Besides, the EasiECG achieves an overall accuracy of 87.6% on samples with a missing wave in the related experiment, demonstrating the robustness of our proposed method.

Convolutional Neural Networks (CNNs) are powerful but computationally demanding and memory intensive, thus impeding their practical applications on resource-constrained hardware. Filter pruning is an efficient approach for deep CNN compression and acceleration, which aims to eliminate some filters with tolerable performance degradation. In the literature, the majority of approaches prune networks by defining the redundant filters or training the networks with a sparsity prior loss function. These approaches mainly use FLOPs as their speed metric. However, the inference latency of pruned networks cannot be directly controlled on the hardware platform, which is an important dimension of practicality. To address this issue, we propose a novel Hardware-aware Filter Pruning method (HFP) which can produce pruned networks that satisfy the actual latency budget on the hardwares of interest. In addition, we propose an iterative pruning framework called Opti-Trim to decrease the accuracy degradation of pruning process and accelerate the pruning procedure whilst meeting the hardware budget. More specifically, HFP first builds up a lookup table for fast estimating the latency of target network about filter configuration layer by layer. Then, HFP leverages information gain (IG) to globally evaluate the filters contribution to network output distribution. HFP utilizes the Opti-Trim framework to globally prune filters with the minimum IG one by one until the latency budget is satisfied. We verify the effectiveness of the proposed method on CIFAR-10 and ImageNet. Compared with the state-of-the-art pruning methods, HFP demonstrates superior performances on VGGNet, ResNet and MobileNet V1/V2.