Pre-trained transformer models have achieved remarkable success in natural language processing (NLP) and have recently become competitive alternatives to Convolution Neural Networks (CNN) and Recurrent Neural Networks (RNN) in vision and speech tasks, respectively. Due to excellent computational efficiency and scalability, transformer models can be trained on exceedingly large amounts of data; however, model sizes can grow tremendously. As high performance, large-scale, and pre-trained transformer models become available for users to download and fine-tune for customized downstream tasks, the deployment of these models becomes challenging due to the vast amount of operations and large memory footprint. To address this challenge, we introduce methods to deeply compress pre-trained transformer models across three major application domains: NLP, speech, and vision. Specifically, we quantize transformer backbones down to 4-bit and further achieve 50% fine-grained structural sparsity on pre-trained BERT, Wav2vec2.0

Softmax policy gradient is a popular algorithm for policy optimization in single-agent reinforcement learning, particularly since projection is not needed for each gradient update. However, in multi-agent systems, the lack of central coordination introduces significant additional difficulties in the convergence analysis. Even for a stochastic game with identical interest, there can be multiple Nash Equilibria (NEs), which disables proof techniques that rely on the existence of a unique global optimum. Moreover, the softmax parameterization introduces non-NE policies with zero gradient, making it difficult for gradient-based algorithms in seeking NEs. In this paper, we study the finite time convergence of decentralized softmax gradient play in a special form of game, Markov Potential Games (MPGs), which includes the identical interest game as a special case. We investigate both gradient play and natural gradient play, with and without \log -barrier regularization.

We study the asynchronous stochastic gradient descent algorithm, for distributed training over n workers that might be heterogeneous. In this algorithm, workers compute stochastic gradients in parallel at their own pace and return them to the server without any synchronization.Existing convergence rates of this algorithm for non-convex smooth objectives depend on the maximum delay \tau_{\max} and reach an \epsilon -stationary point after O\!\left(\sigma 2\epsilon {-2} \tau_{\max}\epsilon {-1}\right) iterations, where \sigma is the variance of stochastic gradients. We also provide (ii) a simple delay-adaptive learning rate scheme, under which asynchronous SGD achieves a convergence rate of O\!\left(\sigma 2\epsilon {-2} \tau_{avg}\epsilon {-1}\right), and does not require any extra hyperparameter tuning nor extra communications. In addition, (iii) we consider the case of heterogeneous functions motivated by federated learning applications and improve the convergence rate by proving a weaker dependence on the maximum delay compared to prior works.

We present Neural Shape Deformation Priors, a novel method for shape manipulation that predicts mesh deformations of non-rigid objects from user-provided handle movements. State-of-the-art methods cast this problem as an optimization task, where the input source mesh is iteratively deformed to minimize an objective function according to hand-crafted regularizers such as ARAP. In this work, we learn the deformation behavior based on the underlying geometric properties of a shape, while leveraging a large-scale dataset containing a diverse set of non-rigid deformations. Specifically, given a source mesh and desired target locations of handles that describe the partial surface deformation, we predict a continuous deformation field that is defined in 3D space to describe the space deformation. To this end, we introduce transformer-based deformation networks that represent a shape deformation as a composition of local surface deformations. It learns a set of local latent codes anchored in 3D space, from which we can learn a set of continuous deformation functions for local surfaces.

In recent years, federated learning (FL) has emerged as an important distributed machine learning paradigm to collaboratively learn a global model with multiple clients, while keeping data local and private. However, a key assumption in most existing works on FL algorithms' convergence analysis is that the noise in stochastic first-order information has a finite variance. Although this assumption covers all light-tailed (i.e., sub-exponential) and some heavy-tailed noise distributions (e.g., log-normal, Weibull, and some Pareto distributions), it fails for many fat-tailed noise distributions (i.e., heavier-tailed'' with potentially infinite variance) that have been empirically observed in the FL literature. To date, it remains unclear whether one can design convergent algorithms for FL systems that experience fat-tailed noise. Specifically, for the largest \alpha \in (1,2] such that the fat-tailed noise in FL still has a bounded \alpha -moment, we show that both variants achieve \mathcal{O}((mT) {\frac{2-\alpha}{\alpha}}) and \mathcal{O}((mT) {\frac{1-\alpha}{3\alpha-2}}) convergence rates in the strongly-convex and general non-convex settings, respectively, where m and T are the numbers of clients and communication rounds.

Deep learning has achieved many breakthroughs in modern classification tasks. Numerous architectures have been proposed for different data structures but when it comes to the loss function, the cross-entropy loss is the predominant choice. Recently, several alternative losses have seen revived interests for deep classifiers. In particular, empirical evidence seems to promote square loss but a theoretical justification is still lacking. In this work, we contribute to the theoretical understanding of square loss in classification by systematically investigating how it performs for overparametrized neural networks in the neural tangent kernel (NTK) regime.

Multi-agent reinforcement learning (MARL) has witnessed significant progress with the development of value function factorization methods. It allows optimizing a joint action-value function through the maximization of factorized per-agent utilities. In this paper, we show that in partially observable MARL problems, an agent's ordering over its own actions could impose concurrent constraints (across different states) on the representable function class, causing significant estimation errors during training. We tackle this limitation and propose PAC, a new framework leveraging Assistive information generated from Counterfactual Predictions of optimal joint action selection, which enable explicit assistance to value function factorization through a novel counterfactual loss. A variational inference-based information encoding method is developed to collect and encode the counterfactual predictions from an estimated baseline.

We initiate the study of private inference on Transformer-based models in the client-server setting, where clients have private inputs and servers hold proprietary models. Our main contribution is to provide several new secure protocols for matrix multiplication and complex non-linear functions like Softmax, GELU activations, and LayerNorm, which are critical components of Transformers. Specifically, we first propose a customized homomorphic encryption-based protocol for matrix multiplication that crucially relies on a novel compact packing technique. This design achieves \sqrt{m} \times less communication ( m is the number of rows of the output matrix) over the most efficient work. Second, we design efficient protocols for three non-linear functions via integrating advanced underlying protocols and specialized optimizations.

Neural Radiance Field (NeRF) has emerged as a compelling method to represent 3D objects and scenes for photo-realistic rendering. However, its implicit representation causes difficulty in manipulating the models like the explicit mesh representation.Several recent advances in NeRF manipulation are usually restricted by a shared renderer network, or suffer from large model size. To circumvent the hurdle, in this paper, we present a neural field representation that enables efficient and convenient manipulation of models.To achieve this goal, we learn a hybrid tensor rank decomposition of the scene without neural networks. Motivated by the low-rank approximation property of the SVD algorithm, we propose a rank-residual learning strategy to encourage the preservation of primary information in lower ranks. The model size can then be dynamically adjusted by rank truncation to control the levels of detail, achieving near-optimal compression without extra optimization.Furthermore, different models can be arbitrarily transformed and composed into one scene by concatenating along the rank dimension.The growth of storage cost can also be mitigated by compressing the unimportant objects in the composed scene.

A fundamental property of deep learning normalization techniques, such as batch normalization, is making the pre-normalization parameters scale invariant. The intrinsic domain of such parameters is the unit sphere, and therefore their gradient optimization dynamics can be represented via spherical optimization with varying effective learning rate (ELR), which was studied previously. However, the varying ELR may obscure certain characteristics of the intrinsic loss landscape structure. In this work, we investigate the properties of training scale-invariant neural networks directly on the sphere using a fixed ELR. We discover three regimes of such training depending on the ELR value: convergence, chaotic equilibrium, and divergence.