Industrial search and recommendation systems mostly follow the classic multi-stage information retrieval paradigm: matching, pre-ranking, ranking, and re-ranking stages. To account for system efficiency, simple vector-product based models are commonly deployed in the pre-ranking stage. Recent works consider distilling the high knowledge of large ranking models to small pre-ranking models for better effectiveness. However, two major challenges in pre-ranking system still exist: (i) without explicitly modeling the performance gain versus computation cost, the predefined latency constraint in the pre-ranking stage inevitably leads to suboptimal solutions; (ii) transferring the ranking teacher's knowledge to a pre-ranking student with a predetermined handcrafted architecture still suffers from the loss of model performance. In this work, a novel framework AutoFAS is proposed which jointly optimizes the efficiency and effectiveness of the pre-ranking model: (i) AutoFAS for the first time simultaneously selects the most valuable features and network architectures using Neural Architecture Search (NAS) technique; (ii) equipped with ranking model guided reward during NAS procedure, AutoFAS can select the best pre-ranking architecture for a given ranking teacher without any computation overhead. Experimental results in our real world search system show AutoFAS consistently outperforms the previous state-of-the-art (SOTA) approaches at a lower computing cost. Notably, our model has been adopted in the pre-ranking module in the search system of Meituan, bringing significant improvements.
6G wireless networks are foreseen to speed up the convergence of the physical and cyber worlds and to enable a paradigm-shift in the way we deploy and exploit communication networks. Machine learning, in particular deep learning (DL), is going to be one of the key technological enablers of 6G by offering a new paradigm for the design and optimization of networks with a high level of intelligence. In this article, we introduce an emerging DL architecture, known as the transformer, and discuss its potential impact on 6G network design. We first discuss the differences between the transformer and classical DL architectures, and emphasize the transformer's self-attention mechanism and strong representation capabilities, which make it particularly appealing in tackling various challenges in wireless network design. Specifically, we propose transformer-based solutions for massive multiple-input multiple-output (MIMO) systems and various semantic communication problems in 6G networks. Finally, we discuss key challenges and open issues in transformer-based solutions, and identify future research directions for their deployment in intelligent 6G networks.
Integrated sensing and communication (ISAC) has opened up numerous game-changing opportunities for realizing future wireless systems. In this paper, we propose an ISAC processing framework relying on millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) systems. Specifically, we provide a compressed sampling (CS) perspective to facilitate ISAC processing, which can not only recover the large-scale channel state information or/and radar imaging information, but also significantly reduce pilot overhead. First, an energy-efficient widely spaced array (WSA) architecture is tailored for the radar receiver, which enhances the angular resolution of radar sensing at the cost of angular ambiguity. Then, we propose an ISAC frame structure for time-variant ISAC systems considering different timescales. The pilot waveforms are judiciously designed by taking into account both CS theories and hardware constraints. Next, we design the dedicated dictionary for WSA that serves as a building block for formulating the ISAC processing as sparse signal recovery problems. The orthogonal matching pursuit with support refinement (OMP-SR) algorithm is proposed to effectively solve the problems in the existence of the angular ambiguity. We also provide a framework for estimating and compensating the Doppler frequencies during payload data transmission to guarantee communication performances. Simulation results demonstrate the good performances of both communications and radar sensing under the proposed ISAC framework.
Multi-touch attribution (MTA), aiming to estimate the contribution of each advertisement touchpoint in conversion journeys, is essential for budget allocation and automatically advertising. Existing methods first train a model to predict the conversion probability of the advertisement journeys with historical data and calculate the attribution of each touchpoint using counterfactual predictions. An assumption of these works is the conversion prediction model is unbiased, i.e., it can give accurate predictions on any randomly assigned journey, including both the factual and counterfactual ones. Nevertheless, this assumption does not always hold as the exposed advertisements are recommended according to user preferences. This confounding bias of users would lead to an out-of-distribution (OOD) problem in the counterfactual prediction and cause concept drift in attribution. In this paper, we define the causal MTA task and propose CausalMTA to eliminate the influence of user preferences. It systemically eliminates the confounding bias from both static and dynamic preferences to learn the conversion prediction model using historical data. We also provide a theoretical analysis to prove CausalMTA can learn an unbiased prediction model with sufficient data. Extensive experiments on both public datasets and the impression data in an e-commerce company show that CausalMTA not only achieves better prediction performance than the state-of-the-art method but also generates meaningful attribution credits across different advertising channels.
Modern video-text retrieval frameworks basically consist of three parts: video encoder, text encoder and the similarity head. With the success on both visual and textual representation learning, transformer based encoders and fusion methods have also been adopted in the field of video-text retrieval. In this report, we present CLIP2TV, aiming at exploring where the critical elements lie in transformer based methods. To achieve this, We first revisit some recent works on multi-modal learning, then introduce some techniques into video-text retrieval, finally evaluate them through extensive experiments in different configurations. Notably, CLIP2TV achieves 52.9@R1 on MSR-VTT dataset, outperforming the previous SOTA result by 4.1%.
Sea subsurface temperature, an essential component of aquatic wildlife, underwater dynamics and heat transfer with the sea surface, is affected by global warming in climate change. Existing research is commonly based on either physics-based numerical models or data based models. Physical modeling and machine learning are traditionally considered as two unrelated fields for the sea subsurface temperature prediction task, with very different scientific paradigms (physics-driven and data-driven). However, we believe both methods are complementary to each other. Physical modeling methods can offer the potential for extrapolation beyond observational conditions, while data-driven methods are flexible in adapting to data and are capable of detecting unexpected patterns. The combination of both approaches is very attractive and offers potential performance improvement. In this paper, we propose a novel framework based on generative adversarial network (GAN) combined with numerical model to predict sea subsurface temperature. First, a GAN-based model is used to learn the simplified physics between the surface temperature and the target subsurface temperature in numerical model. Then, observation data are used to calibrate the GAN-based model parameters to obtain better prediction. We evaluate the proposed framework by predicting daily sea subsurface temperature in the South China sea. Extensive experiments demonstrate the effectiveness of the proposed framework compared to existing state-of-the-art methods.
In this paper, sparsification techniques aided online prediction algorithms in a reproducing kernel Hilbert space are studied for nonstationary time series. The online prediction algorithms as usual consist of the selection of kernel structure parameters and the kernel weight vector updating. For structure parameters, the kernel dictionary is selected by some sparsification techniques with online selective modeling criteria, and moreover the kernel covariance matrix is intermittently optimized in the light of the covariance matrix adaptation evolution strategy (CMA-ES). Optimizing the real symmetric covariance matrix can not only improve the kernel structure's flexibility by the cross relatedness of the input variables, but also partly alleviate the prediction uncertainty caused by the kernel dictionary selection for nonstationary time series. In order to sufficiently capture the underlying dynamic characteristics in prediction-error time series, a generalized optimization strategy is designed to construct the kernel dictionary sequentially in multiple kernel connection modes. The generalized optimization strategy provides a more self-contained way to construct the entire kernel connections, which enhances the ability to adaptively track the changing dynamic characteristics. Numerical simulations have demonstrated that the proposed approach has superior prediction performance for nonstationary time series.
Interactive single-image segmentation is ubiquitous in the scientific and commercial imaging software. In this work, we focus on the single-image segmentation problem only with some seeds such as scribbles. Inspired by the dynamic receptive field in the human being's visual system, we propose the Gaussian dynamic convolution (GDC) to fast and efficiently aggregate the contextual information for neural networks. The core idea is randomly selecting the spatial sampling area according to the Gaussian distribution offsets. Our GDC can be easily used as a module to build lightweight or complex segmentation networks. We adopt the proposed GDC to address the typical single-image segmentation tasks. Furthermore, we also build a Gaussian dynamic pyramid Pooling to show its potential and generality in common semantic segmentation. Experiments demonstrate that the GDC outperforms other existing convolutions on three benchmark segmentation datasets including Pascal-Context, Pascal-VOC 2012, and Cityscapes. Additional experiments are also conducted to illustrate that the GDC can produce richer and more vivid features compared with other convolutions. In general, our GDC is conducive to the convolutional neural networks to form an overall impression of the image.
Camera scene detection is among the most popular computer vision problem on smartphones. While many custom solutions were developed for this task by phone vendors, none of the designed models were available publicly up until now. To address this problem, we introduce the first Mobile AI challenge, where the target is to develop quantized deep learning-based camera scene classification solutions that can demonstrate a real-time performance on smartphones and IoT platforms. For this, the participants were provided with a large-scale CamSDD dataset consisting of more than 11K images belonging to the 30 most important scene categories. The runtime of all models was evaluated on the popular Apple Bionic A11 platform that can be found in many iOS devices. The proposed solutions are fully compatible with all major mobile AI accelerators and can demonstrate more than 100-200 FPS on the majority of recent smartphone platforms while achieving a top-3 accuracy of more than 98%. A detailed description of all models developed in the challenge is provided in this paper.
In Goal-oriented Reinforcement learning, relabeling the raw goals in past experience to provide agents with hindsight ability is a major solution to the reward sparsity problem. In this paper, to enhance the diversity of relabeled goals, we develop FGI (Foresight Goal Inference), a new relabeling strategy that relabels the goals by looking into the future with a learned dynamics model. Besides, to improve sample efficiency, we propose to use the dynamics model to generate simulated trajectories for policy training. By integrating these two improvements, we introduce the MapGo framework (Model-Assisted Policy Optimization for Goal-oriented tasks). In our experiments, we first show the effectiveness of the FGI strategy compared with the hindsight one, and then show that the MapGo framework achieves higher sample efficiency when compared to model-free baselines on a set of complicated tasks.