Generative Pretraining at Scale: Transformer-Based Encoding of Transactional Behavior for Fraud Detection

In this work, we introduce an innovative autoregressive model leveraging Generative Pretrained Transformer (GPT) architectures, tailored for fraud detection in payment systems. Our approach innovatively confronts token explosion and reconstructs behavioral sequences, providing a nuanced understanding of transactional behavior through temporal and contextual analysis. Utilizing unsupervised pretraining, our model excels in feature representation without the need for labeled data. Additionally, we integrate a differential convolutional approach to enhance anomaly detection, bolstering the security and efficacy of one of the largest online payment merchants in China. The scalability and adaptability of our model promise broad applicability in various transactional contexts.

Realization Scheme for Visual Cryptography with Computer-generated Holograms

We propose to realize visual cryptography in an indirect way with the help of computer-generated hologram. At present, the recovery method of visual cryptography is mainly superimposed on transparent film or superimposed by computer equipment, which greatly limits the application range of visual cryptography. In this paper, the shares of the visual cryptography were encoded with computer-generated hologram, and the shares is reproduced by optical means, and then superimposed and decrypted. This method can expand the application range of visual cryptography and further increase the security of visual cryptography.

CROLoss: Towards a Customizable Loss for Retrieval Models in Recommender Systems

In large-scale recommender systems, retrieving top N relevant candidates accurately with resource constrain is crucial. To evaluate the performance of such retrieval models, Recall@N, the frequency of positive samples being retrieved in the top N ranking, is widely used. However, most of the conventional loss functions for retrieval models such as softmax cross-entropy and pairwise comparison methods do not directly optimize Recall@N. Moreover, those conventional loss functions cannot be customized for the specific retrieval size N required by each application and thus may lead to sub-optimal performance. In this paper, we proposed the Customizable Recall@N Optimization Loss (CROLoss), a loss function that can directly optimize the Recall@N metrics and is customizable for different choices of N. This proposed CROLoss formulation defines a more generalized loss function space, covering most of the conventional loss functions as special cases. Furthermore, we develop the Lambda method, a gradient-based method that invites more flexibility and can further boost the system performance. We evaluate the proposed CROLoss on two public benchmark datasets. The results show that CROLoss achieves SOTA results over conventional loss functions for both datasets with various choices of retrieval size N. CROLoss has been deployed onto our online E-commerce advertising platform, where a fourteen-day online A/B test demonstrated that CROLoss contributes to a significant business revenue growth of 4.75%.

DropNAS: Grouped Operation Dropout for Differentiable Architecture Search

Neural architecture search (NAS) has shown encouraging results in automating the architecture design. Recently, DARTS relaxes the search process with a differentiable formulation that leverages weight-sharing and SGD where all candidate operations are trained simultaneously. Our empirical results show that such procedure results in the co-adaption problem and Matthew Effect: operations with fewer parameters would be trained maturely earlier. This causes two problems: firstly, the operations with more parameters may never have the chance to express the desired function since those with less have already done the job; secondly, the system will punish those underperforming operations by lowering their architecture parameter, and they will get smaller loss gradients, which causes the Matthew Effect. In this paper, we systematically study these problems and propose a novel grouped operation dropout algorithm named DropNAS to fix the problems with DARTS. Extensive experiments demonstrate that DropNAS solves the above issues and achieves promising performance. Specifically, DropNAS achieves 2.26% test error on CIFAR-10, 16.39% on CIFAR-100 and 23.4% on ImageNet (with the same training hyperparameters as DARTS for a fair comparison). It is also observed that DropNAS is robust across variants of the DARTS search space. Code is available at https://github.com/wiljohnhong/DropNAS.

Relaxed Conditional Image Transfer for Semi-supervised Domain Adaptation

Semi-supervised domain adaptation (SSDA), which aims to learn models in a partially labeled target domain with the assistance of the fully labeled source domain, attracts increasing attention in recent years. To explicitly leverage the labeled data in both domains, we naturally introduce a conditional GAN framework to transfer images without changing the semantics in SSDA. However, we identify a label-domination problem in such an approach. In fact, the generator tends to overlook the input source image and only memorizes prototypes of each class, which results in unsatisfactory adaptation performance. To this end, we propose a simple yet effective Relaxed conditional GAN (Relaxed cGAN) framework. Specifically, we feed the image without its label to our generator. In this way, the generator has to infer the semantic information of input data. We formally prove that its equilibrium is desirable and empirically validate its practical convergence and effectiveness in image transfer. Additionally, we propose several techniques to make use of unlabeled data in the target domain, enhancing the model in SSDA settings. We validate our method on the well-adopted datasets: Digits, DomainNet, and Office-Home. We achieve state-of-the-art performance on DomainNet, Office-Home and most digit benchmarks in low-resource and high-resource settings.

AutoFIS: Automatic Feature Interaction Selection in Factorization Models for Click-Through Rate Prediction

Learning effective feature interactions is crucial for click-through rate (CTR) prediction tasks in recommender systems. In most of the existing deep learning models, feature interactions are either manually designed or simply enumerated. However, enumerating all feature interactions brings large memory and computation cost. Even worse, useless interactions may introduce unnecessary noise and complicate the training process. In this work, we propose a two-stage algorithm called Automatic Feature Interaction Selection (AutoFIS). AutoFIS can automatically identify all the important feature interactions for factorization models with just the computational cost equivalent to training the target model to convergence. In the \emph{search stage}, instead of searching over a discrete set of candidate feature interactions, we relax the choices to be continuous by introducing the architecture parameters. By implementing a regularized optimizer over the architecture parameters, the model can automatically identify and remove the redundant feature interactions during the training process of the model. In the \emph{re-train stage}, we keep the architecture parameters serving as an attention unit to further boost the performance. Offline experiments on three large-scale datasets (two public benchmarks, one private) demonstrate that the proposed AutoFIS can significantly improve various FM based models. AutoFIS has been deployed onto the training platform of Huawei App Store recommendation service, where a 10-day online A/B test demonstrated that AutoFIS improved the DeepFM model by 20.3\% and 20.1\% in terms of CTR and CVR respectively.

StacNAS: Towards stable and consistent optimization for differentiable Neural Architecture Search

Earlier methods for Neural Architecture Search were computationally expensive. Recently proposed Differentiable Neural Architecture Search algorithms such as DARTS can effectively speed up the computation. However, the current formulation relies on a relaxation of the original problem that leads to unstable and suboptimal solutions. We argue that these problems are caused by three fundamental reasons: (1) The difficulty of bi-level optimization; (2) Multicollinearity of correlated operations such as max pooling and average pooling; (3) The discrepancy between the optimization complexity of the search stage and the final training. In this paper, we propose a grouped variable pruning algorithm based on one-level optimization, which leads to a more stable and consistent optimization solution for differentiable NAS. Extensive experiments verify the superiority of the proposed method regarding both accuracy and stability. Our new approach obtains state-of-the-art accuracy on CIFAR-10, CIFAR-100 and ImageNet.

Revisit Knowledge Distillation: a Teacher-free Framework

Knowledge Distillation (KD) aims to distill the knowledge of a cumbersome teacher model into a lightweight student model. Its success is generally attributed to the privileged information on similarities among categories provided by the teacher model, and in this sense, only strong teacher models are deployed to teach weaker students in practice. In this work, we challenge this common belief by following experimental observations: 1) beyond the acknowledgment that the teacher can improve the student, the student can also enhance the teacher significantly by reversing the KD procedure; 2) a poorly-trained teacher with much lower accuracy than the student can still improve the latter significantly. To explain these observations, we provide a theoretical analysis of the relationships between KD and label smoothing regularization. We prove that 1) KD is a type of learned label smoothing regularization and 2) label smoothing regularization provides a virtual teacher model for KD. From these results, we argue that the success of KD is not fully due to the similarity information between categories, but also to the regularization of soft targets, which is equally or even more important. Based on these analyses, we further propose a novel Teacher-free Knowledge Distillation (Tf-KD) framework, where a student model learns from itself or manually-designed regularization distribution. The Tf-KD achieves comparable performance with normal KD from a superior teacher, which is well applied when teacher model is unavailable. Meanwhile, Tf-KD is generic and can be directly deployed for training deep neural networks. Without any extra computation cost, Tf-KD achieves up to 0.65\% improvement on ImageNet over well-established baseline models, which is superior to label smoothing regularization. The codes are in: \url{https://github.com/yuanli2333/Teacher-free-Knowledge-Distillation}