arXiv subjects

Fei Sun

Publications and source records attributed to Fei Sun.

At least 163 records · Page 9Linked to original sources

arXiv2020

A Unified DNN Weight Compression Framework Using Reweighted Optimization Methods

To address the large model size and intensive computation requirement of deep neural networks (DNNs), weight pruning techniques have been proposed and generally fall into two categories, i.e., static regularization-based pruning and dynamic regularization-based pruning. However, the former method currently suffers either complex workloads or accuracy degradation, while the latter one takes a long time to tune the parameters to achieve the desired pruning rate without accuracy loss. In this paper, we propose a unified DNN weight pruning framework with dynamically updated regularization terms bounded by the designated constraint, which can generate both non-structured sparsity and different kinds of structured sparsity. We also extend our method to an integrated framework for the combination of different DNN compression tasks.

cs.LG↗

arXiv2020

Learning in the Frequency Domain

Deep neural networks have achieved remarkable success in computer vision tasks. Existing neural networks mainly operate in the spatial domain with fixed input sizes. For practical applications, images are usually large and have to be downsampled to the predetermined input size of neural networks. Even though the downsampling operations reduce computation and the required communication bandwidth, it removes both redundant and salient information obliviously, which results in accuracy degradation. Inspired by digital signal processing theories, we analyze the spectral bias from the frequency perspective and propose a learning-based frequency selection method to identify the trivial frequency components which can be removed without accuracy loss. The proposed method of learning in the frequency domain leverages identical structures of the well-known neural networks, such as ResNet-50, MobileNetV2, and Mask R-CNN, while accepting the frequency-domain information as the input. Experiment results show that learning in the frequency domain with static channel selection can achieve higher accuracy than the conventional spatial downsampling approach and meanwhile further reduce the input data size. Specifically for ImageNet classification with the same input size, the proposed method achieves 1.41% and 0.66% top-1 accuracy improvements on ResNet-50 and MobileNetV2, respectively. Even with half input size, the proposed method still improves the top-1 accuracy on ResNet-50 by 1%. In addition, we observe a 0.8% average precision improvement on Mask R-CNN for instance segmentation on the COCO dataset.

cs.CV↗

arXiv2020

Skeleton Based Action Recognition using a Stacked Denoising Autoencoder with Constraints of Privileged Information

Recently, with the availability of cost-effective depth cameras coupled with real-time skeleton estimation, the interest in skeleton-based human action recognition is renewed. Most of the existing skeletal representation approaches use either the joint location or the dynamics model. Differing from the previous studies, we propose a new method called Denoising Autoencoder with Temporal and Categorical Constraints (DAE_CTC)} to study the skeletal representation in a view of skeleton reconstruction. Based on the concept of learning under privileged information, we integrate action categories and temporal coordinates into a stacked denoising autoencoder in the training phase, to preserve category and temporal feature, while learning the hidden representation from a skeleton. Thus, we are able to improve the discriminative validity of the hidden representation. In order to mitigate the variation resulting from temporary misalignment, a new method of temporal registration, called Locally-Warped Sequence Registration (LWSR), is proposed for registering the sequences of inter- and intra-class actions. We finally represent the sequences using a Fourier Temporal Pyramid (FTP) representation and perform classification using a combination of LWSR registration, FTP representation, and a linear Support Vector Machine (SVM). The experimental results on three action data sets, namely MSR-Action3D, UTKinect-Action, and Florence3D-Action, show that our proposal performs better than many existing methods and comparably to the state of the art.

cs.CV↗

arXiv2020

Privileged Features Distillation at Taobao Recommendations

Features play an important role in the prediction tasks of e-commerce recommendations. To guarantee the consistency of off-line training and on-line serving, we usually utilize the same features that are both available. However, the consistency in turn neglects some discriminative features. For example, when estimating the conversion rate (CVR), i.e., the probability that a user would purchase the item if she clicked it, features like dwell time on the item detailed page are informative. However, CVR prediction should be conducted for on-line ranking before the click happens. Thus we cannot get such post-event features during serving. We define the features that are discriminative but only available during training as the privileged features. Inspired by the distillation techniques which bridge the gap between training and inference, in this work, we propose privileged features distillation (PFD). We train two models, i.e., a student model that is the same as the original one and a teacher model that additionally utilizes the privileged features. Knowledge distilled from the more accurate teacher is transferred to the student to improve its accuracy. During serving, only the student part is extracted and it relies on no privileged features. We conduct experiments on two fundamental prediction tasks at Taobao recommendations, i.e., click-through rate (CTR) at coarse-grained ranking and CVR at fine-grained ranking. By distilling the interacted features that are prohibited during serving for CTR and the post-event features for CVR, we achieve significant improvements over their strong baselines. During the on-line A/B tests, the click metric is improved by +5.0% in the CTR task. And the conversion metric is improved by +2.3% in the CVR task. Besides, by addressing several issues of training PFD, we obtain comparable training speed as the baselines without any distillation.

cs.IR↗

arXiv2020

Flat-top interleavers based on single MMIs

We demonstrate for the first time flat-top interleavers based on cascaded Mach-Zehnder interferometers (MZIs) which use only single multimode interferometers (MMIs) as power splitters. Our previous designs were based on 4-stage cascades of MZIs, where we used single MMIs and double MMIs to achieve 85:15 splitting ratio and 31:69 splitting ratio respectively. This time, we propose instead a greatly simplified 2-stage configuration using only single MMIs, including a standard 50:50 MMI, and two tapered MMIs to achieve 71:29 and 92:08 splitting ratios. We have designed the interleaver based on its geometrical representation on the Bloch sphere, then confirmed by efficient 2D simulations of the building blocks and of the whole structure, based on the eigenmode expansion method. We show how important is to take into account the phase relations between the outputs of all MMIs in order to make a working design. We have successfully fabricated devices with different channel spacing on our micron-scale silicon photonics platform, and measurement results confirmed their expected flat-top operation on a broad band. Using only single MMI splitters we can not only greatly outperform the bandwidth achieved by standard directional couplers, but we can also ensure much higher robustness to fabrication errors, also compared to previous demonstrations based on double MMIs. Indeed, when compared to those previous attempts, the new results prove tapered MMIs to be the most robust approach to achieve arbitrary splitting ratios.

physics.optics↗

arXiv2019

SDM: Sequential Deep Matching Model for Online Large-scale Recommender System

Capturing users' precise preferences is a fundamental problem in large-scale recommender system. Currently, item-based Collaborative Filtering (CF) methods are common matching approaches in industry. However, they are not effective to model dynamic and evolving preferences of users. In this paper, we propose a new sequential deep matching (SDM) model to capture users' dynamic preferences by combining short-term sessions and long-term behaviors. Compared with existing sequence-aware recommendation methods, we tackle the following two inherent problems in real-world applications: (1) there could exist multiple interest tendencies in one session. (2) long-term preferences may not be effectively fused with current session interests. Long-term behaviors are various and complex, hence those highly related to the short-term session should be kept for fusion. We propose to encode behavior sequences with two corresponding components: multi-head self-attention module to capture multiple types of interests and long-short term gated fusion module to incorporate long-term preferences. Successive items are recommended after matching between sequential user behavior vector and item embedding vectors. Offline experiments on real-world datasets show the superior performance of the proposed SDM. Moreover, SDM has been successfully deployed on online large-scale recommender system at Taobao and achieves improvements in terms of a range of commercial metrics.

cs.IR↗

arXiv2019

Thermal-null medium (TNM): a novel material to achieve feasible thermodynamics devices beyond conventional challenges

Recently, heat manipulation has gained the attention of scientific community due to its several applications. In this letter, based on transformation thermodynamic (TT) methodology, a novel material, which is called thermal-null medium (TNM), is proposed that enables us to design various thermal functionalities such as thermal bending devices, arbitrary shape heat concentrators and omnidirectional thermal cloaks. In contrary to the conventional TT-based conductivities, which are inhomogeneous and anisotropic, TNMs are homogeneous and easy to realize. In addition, the attained TNMs are independent of the device shape. That is if the geometry of the desired device is changed, there is no need to recalculate the necessitating conductivities. This feature of TNM will make it suitable for scenarios where re-configurability is of utmost importance. Several numerical simulations are carried out to demonstrate the TNM capability and its applications in directional bending devices, heat concentrators and thermal cloaks. The proposed TNM could open a new avenue for potential applications in solar thermal panels and thermal-electric devices.

physics.app-ph↗

arXiv2019

Nonlinear optical response spatial self-hase modulation in MoTe2: correlations between \c{hi}(3) and mobility or effective mass

We report an unambiguous observation of third-order nonlinear optical effect, spatial self-phase modulation (SSPM), in a MoTe2 dispersion. The values of the third order nonlinear optical coefficients effectively for one layer MoTe2, \c{hi}(3)one-layer, are obtained through the SSPM method at excitation wavelengths 473, 532, 750 and 801 nm, respectively. The wind-chime model is used to explain the ring formation time. The wavelength dependence of \c{hi}(3)one-layer compares well with the photo-absorption spectra. Significantly, we find a correlation between \c{hi}(3) and the carrier mobility μ or effective mass m*, which again further supports the laser-induced ac electron coherence in 2D materials.

physics.optics↗

arXiv2019

BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer

Modeling users' dynamic and evolving preferences from their historical behaviors is challenging and crucial for recommendation systems. Previous methods employ sequential neural networks (e.g., Recurrent Neural Network) to encode users' historical interactions from left to right into hidden representations for making recommendations. Although these methods achieve satisfactory results, they often assume a rigidly ordered sequence which is not always practical. We argue that such left-to-right unidirectional architectures restrict the power of the historical sequence representations. For this purpose, we introduce a Bidirectional Encoder Representations from Transformers for sequential Recommendation (BERT4Rec). However, jointly conditioning on both left and right context in deep bidirectional model would make the training become trivial since each item can indirectly "see the target item". To address this problem, we train the bidirectional model using the Cloze task, predicting the masked items in the sequence by jointly conditioning on their left and right context. Comparing with predicting the next item at each position in a sequence, the Cloze task can produce more samples to train a more powerful bidirectional model. Extensive experiments on four benchmark datasets show that our model outperforms various state-of-the-art sequential models consistently.

cs.IR↗

arXiv2019

Compositional Network Embedding

Network embedding has proved extremely useful in a variety of network analysis tasks such as node classification, link prediction, and network visualization. Almost all the existing network embedding methods learn to map the node IDs to their corresponding node embeddings. This design principle, however, hinders the existing methods from being applied in real cases. Node ID is not generalizable and, thus, the existing methods have to pay great effort in cold-start problem. The heterogeneous network usually requires extra work to encode node types, as node type is not able to be identified by node ID. Node ID carries rare information, resulting in the criticism that the existing methods are not robust to noise. To address this issue, we introduce Compositional Network Embedding, a general inductive network representation learning framework that generates node embeddings by combining node features based on the principle of compositionally. Instead of directly optimizing an embedding lookup based on arbitrary node IDs, we learn a composition function that infers node embeddings by combining the corresponding node attribute embeddings through a graph-based loss. For evaluation, we conduct the experiments on link prediction under four different settings. The results verified the effectiveness and generalization ability of compositional network embeddings, especially on unseen nodes.

cs.LG↗

arXiv2019

Personalized Re-ranking for Recommendation

Ranking is a core task in recommender systems, which aims at providing an ordered list of items to users. Typically, a ranking function is learned from the labeled dataset to optimize the global performance, which produces a ranking score for each individual item. However, it may be sub-optimal because the scoring function applies to each item individually and does not explicitly consider the mutual influence between items, as well as the differences of users' preferences or intents. Therefore, we propose a personalized re-ranking model for recommender systems. The proposed re-ranking model can be easily deployed as a follow-up modular after any ranking algorithm, by directly using the existing ranking feature vectors. It directly optimizes the whole recommendation list by employing a transformer structure to efficiently encode the information of all items in the list. Specifically, the Transformer applies a self-attention mechanism that directly models the global relationships between any pair of items in the whole list. We confirm that the performance can be further improved by introducing pre-trained embedding to learn personalized encoding functions for different users. Experimental results on both offline benchmarks and real-world online e-commerce systems demonstrate the significant improvements of the proposed re-ranking model.

cs.IR↗

arXiv2019

Quasiconvex risk measures with markets volatility

Since the quasiconvex risk measures is a bigger class than the well known convex risk measures, the study of quasiconvex risk measures makes sense especially in the financial markets with volatility. In this paper, we will study the quasiconvex risk measures defined on a special space $L^{p(\cdot)}$ where the variable exponent $p(\cdot)$ is no longer a given real number like the space $L^{p}$, but a random variable, which reflects the possible volatility of the financial markets. The dual representation for this quasiconvex risk measures will also provided.

q-fin.RM↗

arXiv2019

Improving Multi-turn Dialogue Modelling with Utterance ReWriter

Recent research has made impressive progress in single-turn dialogue modelling. In the multi-turn setting, however, current models are still far from satisfactory. One major challenge is the frequently occurred coreference and information omission in our daily conversation, making it hard for machines to understand the real intention. In this paper, we propose rewriting the human utterance as a pre-process to help multi-turn dialgoue modelling. Each utterance is first rewritten to recover all coreferred and omitted information. The next processing steps are then performed based on the rewritten utterance. To properly train the utterance rewriter, we collect a new dataset with human annotations and introduce a Transformer-based utterance rewriting architecture using the pointer network. We show the proposed architecture achieves remarkably good performance on the utterance rewriting task. The trained utterance rewriter can be easily integrated into online chatbots and brings general improvement over different domains.

cs.CL↗

arXiv2019

FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search

Designing accurate and efficient ConvNets for mobile devices is challenging because the design space is combinatorially large. Due to this, previous neural architecture search (NAS) methods are computationally expensive. ConvNet architecture optimality depends on factors such as input resolution and target devices. However, existing approaches are too expensive for case-by-case redesigns. Also, previous work focuses primarily on reducing FLOPs, but FLOP count does not always reflect actual latency. To address these, we propose a differentiable neural architecture search (DNAS) framework that uses gradient-based methods to optimize ConvNet architectures, avoiding enumerating and training individual architectures separately as in previous methods. FBNets, a family of models discovered by DNAS surpass state-of-the-art models both designed manually and generated automatically. FBNet-B achieves 74.1% top-1 accuracy on ImageNet with 295M FLOPs and 23.1 ms latency on a Samsung S8 phone, 2.4x smaller and 1.5x faster than MobileNetV2-1.3 with similar accuracy. Despite higher accuracy and lower latency than MnasNet, we estimate FBNet-B's search cost is 420x smaller than MnasNet's, at only 216 GPU-hours. Searched for different resolutions and channel sizes, FBNets achieve 1.5% to 6.4% higher accuracy than MobileNetV2. The smallest FBNet achieves 50.2% accuracy and 2.9 ms latency (345 frames per second) on a Samsung S8. Over a Samsung-optimized FBNet, the iPhone-X-optimized model achieves a 1.4x speedup on an iPhone X.

cs.CV↗

arXiv2019

POG: Personalized Outfit Generation for Fashion Recommendation at Alibaba iFashion

Increasing demand for fashion recommendation raises a lot of challenges for online shopping platforms and fashion communities. In particular, there exist two requirements for fashion outfit recommendation: the Compatibility of the generated fashion outfits, and the Personalization in the recommendation process. In this paper, we demonstrate these two requirements can be satisfied via building a bridge between outfit generation and recommendation. Through large data analysis, we observe that people have similar tastes in individual items and outfits. Therefore, we propose a Personalized Outfit Generation (POG) model, which connects user preferences regarding individual items and outfits with Transformer architecture. Extensive offline and online experiments provide strong quantitative evidence that our method outperforms alternative methods regarding both compatibility and personalization metrics. Furthermore, we deploy POG on a platform named Dida in Alibaba to generate personalized outfits for the users of the online application iFashion. This work represents a first step towards an industrial-scale fashion outfit generation and recommendation solution, which goes beyond generating outfits based on explicit queries, or merely recommending from existing outfit pools. As part of this work, we release a large-scale dataset consisting of 1.01 million outfits with rich context information, and 0.28 billion user click actions from 3.57 million users. To the best of our knowledge, this dataset is the largest, publicly available, fashion related dataset, and the first to provide user behaviors relating to both outfits and fashion items.

cs.IR↗

arXiv2019

Exact-K Recommendation via Maximal Clique Optimization

This paper targets to a novel but practical recommendation problem named exact-K recommendation. It is different from traditional top-K recommendation, as it focuses more on (constrained) combinatorial optimization which will optimize to recommend a whole set of K items called card, rather than ranking optimization which assumes that "better" items should be put into top positions. Thus we take the first step to give a formal problem definition, and innovatively reduce it to Maximum Clique Optimization based on graph. To tackle this specific combinatorial optimization problem which is NP-hard, we propose Graph Attention Networks (GAttN) with a Multi-head Self-attention encoder and a decoder with attention mechanism. It can end-to-end learn the joint distribution of the K items and generate an optimal card rather than rank individual items by prediction scores. Then we propose Reinforcement Learning from Demonstrations (RLfD) which combines the advantages in behavior cloning and reinforcement learning, making it sufficient- and-efficient to train the model. Extensive experiments on three datasets demonstrate the effectiveness of our proposed GAttN with RLfD method, it outperforms several strong baselines with a relative improvement of 7.7% and 4.7% on average in Precision and Hit Ratio respectively, and achieves state-of-the-art (SOTA) performance for the exact-K recommendation problem.

cs.IR↗

arXiv2019

Deep Session Interest Network for Click-Through Rate Prediction

Click-Through Rate (CTR) prediction plays an important role in many industrial applications, such as online advertising and recommender systems. How to capture users' dynamic and evolving interests from their behavior sequences remains a continuous research topic in the CTR prediction. However, most existing studies overlook the intrinsic structure of the sequences: the sequences are composed of sessions, where sessions are user behaviors separated by their occurring time. We observe that user behaviors are highly homogeneous in each session, and heterogeneous cross sessions. Based on this observation, we propose a novel CTR model named Deep Session Interest Network (DSIN) that leverages users' multiple historical sessions in their behavior sequences. We first use self-attention mechanism with bias encoding to extract users' interests in each session. Then we apply Bi-LSTM to model how users' interests evolve and interact among sessions. Finally, we employ the local activation unit to adaptively learn the influences of various session interests on the target item. Experiments are conducted on both advertising and production recommender datasets and DSIN outperforms other state-of-the-art models on both datasets.

cs.IR↗

arXiv2019

Pressure enhanced interplay among lattice, spin and charge in La2FeMnO6 mixed perovskite

Spin crossover plays a central role in the structural instability, net magnetic moment modification, metallization, and even in superconductivity in corresponding materials. Most reports on the pressure-induced spin crossover with a large volume collapse so far focused on compounds with single transition metal. Here we report a comprehensive high-pressure investigation of a mixed Fe-Mn perovskite La2FeMnO6. Under pressure, the strong coupling between Fe and Mn leads to a combined valence/spin transition: Fe3+(S = 5/2) to Fe2+(S = 0) and Mn3+(S = 2) to Mn4+(S = 3/2), with an isostructural phase transition. The spin transitions of both Fe and Mn are offset by ~ 20 GPa of the onset pressure, and the lattice collapse occurs in between. Interestingly, Fe3+ ion shows an abnormal behavior when it reaches a lower valence state (Fe2+) accompanied by a + 0.5 eV energy shift in Fe K-absorption edge at 15 GPa. This process is associated with the charge-spin-orbital state transition from high spin Fe3+ to low spin Fe2+, caused by the significantly enhanced t2g-eg crystal field splitting in the compressed lattice under high pressure. Density Functional Theory calculations confirm the energy preference of the high-pressure state with charge redistribution accompanied by spin state transition of Fe ions. Moreover, La2FeMnO6 maintains semiconductor behaviors even when the pressure reached 144.5 GPa as evidenced by the electrical transport measurements, despite the huge resistivity decreasing 7 orders of magnitude compared with that at ambient pressure. The investigation carried out here demonstrates high flexibility of double perovskites and their good potentials for optimizing the functionality of these materials.

cond-mat.str-el↗