Goto

Collaborating Authors

 Chen, Wenlin


External Large Foundation Model: How to Efficiently Serve Trillions of Parameters for Online Ads Recommendation

arXiv.org Artificial Intelligence

Ads recommendation is a prominent service of online advertising systems and has been actively studied. Recent studies indicate that scaling-up and advanced design of the recommendation model can bring significant performance improvement. However, with a larger model scale, such prior studies have a significantly increasing gap from industry as they often neglect two fundamental challenges in industrial-scale applications. First, training and inference budgets are restricted for the model to be served, exceeding which may incur latency and impair user experience. Second, large-volume data arrive in a streaming mode with data distributions dynamically shifting, as new users/ads join and existing users/ads leave the system. We propose the External Large Foundation Model (ExFM) framework to address the overlooked challenges. Specifically, we develop external distillation and a data augmentation system (DAS) to control the computational cost of training/inference while maintaining high performance. We design the teacher in a way like a foundation model (FM) that can serve multiple students as vertical models (VMs) to amortize its building cost. We propose Auxiliary Head and Student Adapter to mitigate the data distribution gap between FM and VMs caused by the streaming data issue. Comprehensive experiments on internal industrial-scale applications and public datasets demonstrate significant performance gain by ExFM.


Towards Training One-Step Diffusion Models Without Distillation

arXiv.org Artificial Intelligence

Recent advances in one-step generative models typically follow a two-stage process: first training a teacher diffusion model and then distilling it into a one-step student model. This distillation process traditionally relies on both the teacher model's score function to compute the distillation loss and its weights for student initialization. In this paper, we explore whether one-step generative models can be trained directly without this distillation process. First, we show that the teacher's score function is not essential and propose a family of distillation methods that achieve competitive results without relying on score estimation. Next, we demonstrate that initialization from teacher weights is indispensable in successful training. Surprisingly, we find that this benefit is not due to improved ``input-output" mapping but rather the learned feature representations, which dominate distillation quality. Our findings provide a better understanding of the role of initialization in one-step model training and its impact on distillation quality.


Your Image is Secretly the Last Frame of a Pseudo Video

arXiv.org Artificial Intelligence

Diffusion models, which can be viewed as a special case of hierarchical variational autoencoders (HVAEs), have shown profound success in generating photo-realistic images. In contrast, standard HVAEs often produce images of inferior quality compared to diffusion models. In this paper, we hypothesize that the success of diffusion models can be partly attributed to the additional self-supervision information for their intermediate latent states provided by corrupted images, which along with the original image form a pseudo video. Based on this hypothesis, we explore the possibility of improving other types of generative models with such pseudo videos. Specifically, we first extend a given image generative model to their video generative model counterpart, and then train the video generative model on pseudo videos constructed by applying data augmentation to the original images. Furthermore, we analyze the potential issues of first-order Markov data augmentation methods, which are typically used in diffusion models, and propose to use more expressive data augmentation to construct more useful information in pseudo videos. Our empirical results on the CIFAR10 and CelebA datasets demonstrate that improved image generation quality can be achieved with additional self-supervised information from pseudo videos.


Training Neural Samplers with Reverse Diffusive KL Divergence

arXiv.org Machine Learning

Training generative models to sample from unnormalized density functions is an important and challenging task in machine learning. Traditional training methods often rely on the reverse Kullback-Leibler (KL) divergence due to its tractability. However, the mode-seeking behavior of reverse KL hinders effective approximation of multi-modal target distributions. To address this, we propose to minimize the reverse KL along diffusion trajectories of both model and target densities. We refer to this objective as the reverse diffusive KL divergence, which allows the model to capture multiple modes. Leveraging this objective, we train neural samplers that can efficiently generate samples from the target distribution in one step. We demonstrate that our method enhances sampling performance across various Boltzmann distributions, including both synthetic multi-modal densities and n-body particle systems.


Wukong: Towards a Scaling Law for Large-Scale Recommendation

arXiv.org Artificial Intelligence

Scaling laws play an instrumental role in the sustainable improvement in model quality. Unfortunately, recommendation models to date do not exhibit such laws similar to those observed in the domain of large language models, due to the inefficiencies of their upscaling mechanisms. This limitation poses significant challenges in adapting these models to increasingly more complex real-world datasets. In this paper, we propose an effective network architecture based purely on stacked factorization machines, and a synergistic upscaling strategy, collectively dubbed Wukong, to establish a scaling law in the domain of recommendation. Wukong's unique design makes it possible to capture diverse, any-order of interactions simply through taller and wider layers. We conducted extensive evaluations on six public datasets, and our results demonstrate that Wukong consistently outperforms state-of-the-art models quality-wise. Further, we assessed Wukong's scalability on an internal, large-scale dataset. The results show that Wukong retains its superiority in quality over state-of-the-art models, while holding the scaling law across two orders of magnitude in model complexity, extending beyond 100 GFLOP/example, where prior arts fall short.


Diffusive Gibbs Sampling

arXiv.org Artificial Intelligence

The inadequate mixing of conventional Markov Chain Monte Carlo (MCMC) methods for multi-modal distributions presents a significant challenge in practical applications such as Bayesian inference and molecular dynamics. Addressing this, we propose Diffusive Gibbs Sampling (DiGS), an innovative family of sampling methods designed for effective sampling from distributions characterized by distant and disconnected modes. DiGS integrates recent developments in diffusion models, leveraging Gaussian convolution to create an auxiliary noisy distribution that bridges isolated modes in the original space and applying Gibbs sampling to alternately draw samples from both spaces. Our approach exhibits a better mixing property for sampling multi-modal distributions than state-of-the-art methods such as parallel tempering. We demonstrate that our sampler attains substantially improved results across various tasks, including mixtures of Gaussians, Bayesian neural networks and molecular dynamics.


It HAS to be Subjective: Human Annotator Simulation via Zero-shot Density Estimation

arXiv.org Artificial Intelligence

Human annotator simulation (HAS) serves as a cost-effective substitute for human evaluation such as data annotation and system assessment. Human perception and behaviour during human evaluation exhibit inherent variability due to diverse cognitive processes and subjective interpretations, which should be taken into account in modelling to better mimic the way people perceive and interact with the world. This paper introduces a novel meta-learning framework that treats HAS as a zeroshot density estimation problem, which incorporates human variability and allows for the efficient generation of human-like annotations for unlabelled test inputs. Under this framework, we propose two new model classes, conditional integer flows and conditional softmax flows, to account for ordinal and categorical annotations, respectively. The proposed method is evaluated on three real-world human evaluation tasks and shows superior capability and efficiency to predict the aggregated behaviours of human annotators, match the distribution of human annotations, and simulate the inter-annotator disagreements. Collecting human annotations or evaluations often requires substantial resources and may expose human annotators to distressing and harmful content in sensitive tasks (e.g., toxic speech detection, suicidal risk prediction, and depression detection). This inspires the exploration of human annotator simulation (HAS) as a scalable and cost-effective alternative, which facilitates large-scale dataset evaluation, benchmarking, and system comparisons. Variability is a unique aspect of real-world human evaluation, since individual variations in cognitive biases, cultural backgrounds, and personal experiences (Hirschberg et al., 2003; Wiebe et al., 2004; Haselton et al., 2015) can lead to variability in human interpretation (Lotfian & Busso, 2019; Mathew et al., 2021; Maniati et al., 2022). HAS aims to incorporate the variability present in human evaluation rather than solely relying on majority opinions, which mitigates potential biases and over-representation in scenarios where dominant opinions could potentially overshadow minority viewpoints (Dixon et al., 2018; Hutchinson et al., 2020), thus promoting fairness and inclusivity. In this work, we investigate HAS for the automatic generation of human-like annotations that take into account the variability in human evaluation.


Leveraging Task Structures for Improved Identifiability in Neural Network Representations

arXiv.org Machine Learning

This work extends the theory of identifiability in supervised learning by considering the consequences of having access to a distribution of tasks. In such cases, we show that identifiability is achievable even in the case of regression, extending prior work restricted to linear identifiability in the single-task classification case. Furthermore, we show that the existence of a task distribution which defines a conditional prior over latent factors reduces the equivalence class for identifiability to permutations and scaling, a much stronger and more useful result than linear identifiability. When we further assume a causal structure over these tasks, our approach enables simple maximum marginal likelihood optimization together with downstream applicability to causal representation learning. Empirically, we validate that our model outperforms more general unsupervised models in recovering canonical representations for both synthetic and real-world molecular data. Multi-task regression is a common problem in machine learning, which naturally arises in many scientific applications such as molecular property prediction (Stanley et al., 2021; Chen et al., 2023). Despite this, most deep learning approaches to this problem attempt to model the relationships between tasks through heuristic approaches, such as fitting a shared neural network in an attempt to capture the joint structures between tasks.


Neural Characteristic Activation Value Analysis for Improved ReLU Network Feature Learning

arXiv.org Machine Learning

This work examines the characteristic activation values of individual ReLU units in neural networks. We refer to the set of input locations corresponding to such characteristic activation values as the characteristic activation set of a ReLU unit. We draw an explicit connection between the characteristic activation set and learned features in ReLU networks. This connection leads to new insights into how various neural network normalization techniques used in modern deep learning architectures regularize and stabilize stochastic gradient optimization. Utilizing these insights, we propose geometric parameterization for ReLU networks to improve feature learning, which decouples the radial and angular parameters in the hyperspherical coordinate system. We empirically verify its usefulness with less carefully chosen initialization schemes and larger learning rates. We report significant improvements in optimization stability, convergence speed, and generalization performance for various models on a variety of datasets, including the ResNet-50 network on ImageNet.


Towards the Better Ranking Consistency: A Multi-task Learning Framework for Early Stage Ads Ranking

arXiv.org Artificial Intelligence

Dividing ads ranking system into retrieval, early, and final stages is a common practice in large scale ads recommendation to balance the efficiency and accuracy. The early stage ranking often uses efficient models to generate candidates out of a set of retrieved ads. The candidates are then fed into a more computationally intensive but accurate final stage ranking system to produce the final ads recommendation. As the early and final stage ranking use different features and model architectures because of system constraints, a serious ranking consistency issue arises where the early stage has a low ads recall, i.e., top ads in the final stage are ranked low in the early stage. In order to pass better ads from the early to the final stage ranking, we propose a multi-task learning framework for early stage ranking to capture multiple final stage ranking components (i.e. ads clicks and ads quality events) and their task relations. With our multi-task learning framework, we can not only achieve serving cost saving from the model consolidation, but also improve the ads recall and ranking consistency. In the online A/B testing, our framework achieves significantly higher click-through rate (CTR), conversion rate (CVR), total value and better ads-quality (e.g. reduced ads cross-out rate) in a large scale industrial ads ranking system.