With the tremendous success of large transformer models in natural language understanding, down-sizing them for cost-effective deployments has become critical. Recent studies have explored the low-rank weight factorization techniques which are efficient to train, and apply out-of-the-box to any transformer architecture. Unfortunately, the low-rank assumption tends to be over-restrictive and hinders the expressiveness of the compressed model. This paper proposes, DSFormer, a simple alternative factorization scheme which expresses a target weight matrix as the product of a small dense and a semi-structured sparse matrix. The resulting approximation is more faithful to the weight distribution in transformers and therefore achieves a stronger efficiency-accuracy trade-off. Another concern with existing factorizers is their dependence on a task-unaware initialization step which degrades the accuracy of the resulting model. DSFormer addresses this issue through a novel Straight-Through Factorizer (STF) algorithm that jointly learns all the weight factorizations to directly maximize the final task accuracy. Extensive experiments on multiple natural language understanding benchmarks demonstrate that DSFormer obtains up to 40% better compression than the state-of-the-art low-rank factorizers, leading semi-structured sparsity baselines and popular knowledge distillation approaches. Our approach is also orthogonal to mainstream compressors and offers up to 50% additional compression when added to popular distilled, layer-shared and quantized transformers. We empirically evaluate the benefits of STF over conventional optimization practices.
相關內容
We present GenEFT: an effective theory framework for shedding light on the statics and dynamics of neural network generalization, and illustrate it with graph learning examples. We first investigate the generalization phase transition as data size increases, comparing experimental results with information-theory-based approximations. We find generalization in a Goldilocks zone where the decoder is neither too weak nor too powerful. We then introduce an effective theory for the dynamics of representation learning, where latent-space representations are modeled as interacting particles (repons), and find that it explains our experimentally observed phase transition between generalization and overfitting as encoder and decoder learning rates are scanned. This highlights the power of physics-inspired effective theories for bridging the gap between theoretical predictions and practice in machine learning.
Spatio-temporal forecasting, pivotal in numerous fields, hinges on the delicate equilibrium between isolating nuanced patterns and sifting out noise. To tackle this, we introduce Sparse Regression-based Vector Quantization (SVQ), a novel technique that leverages sparse regression for succinct representation, an approach theoretically and practically favored over classical clustering-based vector quantization methods. This approach preserves critical details from the original vectors using a regression model while filtering out noise via sparse design. Moreover, we approximate the sparse regression process using a blend of a two-layer MLP and an extensive codebook. This approach not only substantially cuts down on computational costs but also grants SVQ differentiability and training simplicity, resulting in a notable enhancement of performance. Our empirical studies on five spatial-temporal benchmark datasets demonstrate that SVQ achieves state-of-the-art results. Specifically, on the WeatherBench-S temperature dataset, SVQ improves the top baseline by 7.9%. In video prediction benchmarks-Human, KTH, and KittiCaltech-it reduces MAE by an average of 9.4% and improves image quality by 17.3% (LPIPS).
In response to the evolving landscape of quantum computing and the escalating vulnerabilities in classical cryptographic systems, our paper introduces a unified cryptographic framework. Rooted in the innovative work of Kuang et al., we leverage two novel primitives: the Quantum Permutation Pad (QPP) for symmetric key encryption and the Homomorphic Polynomial Public Key (HPPK) for Key Encapsulation Mechanism (KEM) and Digital Signatures (DS). Our approach adeptly confronts the challenges posed by quantum advancements. Utilizing the Galois Permutation Group's matrix representations and inheriting its bijective and non-commutative properties, QPP achieves quantum-secure symmetric key encryption, seamlessly extending Shannon's perfect secrecy to both classical and quantum-native systems. Meanwhile, HPPK, free from NP-hard problems, fortifies symmetric encryption for the plain public key. It accomplishes this by concealing the mathematical structure through modular multiplications or arithmetic representations of Galois Permutation Group over hidden rings, harnessing their partial homomorphic properties. This allows for secure computation on encrypted data during secret encapsulations, bolstering the security of the plain public key. The seamless integration of KEM and DS within HPPK cryptography yields compact key, cipher, and signature sizes, demonstrating exceptional performance. This paper organically unifies QPP and HPPK under the Galois Permutation Group, marking a significant advancement in laying the groundwork for quantum-resistant cryptographic protocols. Our contribution propels the development of secure communication systems amid the era of quantum computing.
This study investigates the asymptotic dynamics of alternating minimization applied to optimize a bilinear non-convex function with normally distributed covariates. We employ the replica method from statistical physics in a multi-step approach to precisely trace the algorithm's evolution. Our findings indicate that the dynamics can be described effectively by a two--dimensional discrete stochastic process, where each step depends on all previous time steps, revealing a memory dependency in the procedure. The theoretical framework developed in this work is broadly applicable for the analysis of various iterative algorithms, extending beyond the scope of alternating minimization.
We prove an inverse approximation theorem for the approximation of nonlinear sequence-to-sequence relationships using recurrent neural networks (RNNs). This is a so-called Bernstein-type result in approximation theory, which deduces properties of a target function under the assumption that it can be effectively approximated by a hypothesis space. In particular, we show that nonlinear sequence relationships that can be stably approximated by nonlinear RNNs must have an exponential decaying memory structure - a notion that can be made precise. This extends the previously identified curse of memory in linear RNNs into the general nonlinear setting, and quantifies the essential limitations of the RNN architecture for learning sequential relationships with long-term memory. Based on the analysis, we propose a principled reparameterization method to overcome the limitations. Our theoretical results are confirmed by numerical experiments. The code has been released in //github.com/radarFudan/Curse-of-memory
We conduct a systematic study of the approximation properties of Transformer for sequence modeling with long, sparse and complicated memory. We investigate the mechanisms through which different components of Transformer, such as the dot-product self-attention, positional encoding and feed-forward layer, affect its expressive power, and we study their combined effects through establishing explicit approximation rates. Our study reveals the roles of critical parameters in the Transformer, such as the number of layers and the number of attention heads, and these insights also provide natural suggestions for alternative architectures.
We study inference on the long-term causal effect of a continual exposure to a novel intervention, which we term a long-term treatment, based on an experiment involving only short-term observations. Key examples include the long-term health effects of regularly-taken medicine or of environmental hazards and the long-term effects on users of changes to an online platform. This stands in contrast to short-term treatments or ``shocks," whose long-term effect can reasonably be mediated by short-term observations, enabling the use of surrogate methods. Long-term treatments by definition have direct effects on long-term outcomes via continual exposure, so surrogacy conditions cannot reasonably hold. We connect the problem with offline reinforcement learning, leveraging doubly-robust estimators to estimate long-term causal effects for long-term treatments and construct confidence intervals.
With the exponential growth in data volume and the emergence of data-intensive applications, particularly in the field of machine learning, concerns related to resource utilization, privacy, and fairness have become paramount. This paper focuses on the textual domain of data and addresses challenges regarding encoding sentences to their optimized representations through the lens of information-theory. In particular, we use empirical estimates of mutual information, using the Donsker-Varadhan definition of Kullback-Leibler divergence. Our approach leverages this estimation to train an information-theoretic sentence embedding, called TexShape, for (task-based) data compression or for filtering out sensitive information, enhancing privacy and fairness. In this study, we employ a benchmark language model for initial text representation, complemented by neural networks for information-theoretic compression and mutual information estimations. Our experiments demonstrate significant advancements in preserving maximal targeted information and minimal sensitive information over adverse compression ratios, in terms of predictive accuracy of downstream models that are trained using the compressed data.
With the continuous growth in the number of parameters of transformer-based pretrained language models (PLMs), particularly the emergence of large language models (LLMs) with billions of parameters, many natural language processing (NLP) tasks have demonstrated remarkable success. However, the enormous size and computational demands of these models pose significant challenges for adapting them to specific downstream tasks, especially in environments with limited computational resources. Parameter Efficient Fine-Tuning (PEFT) offers an effective solution by reducing the number of fine-tuning parameters and memory usage while achieving comparable performance to full fine-tuning. The demands for fine-tuning PLMs, especially LLMs, have led to a surge in the development of PEFT methods, as depicted in Fig. 1. In this paper, we present a comprehensive and systematic review of PEFT methods for PLMs. We summarize these PEFT methods, discuss their applications, and outline future directions. Furthermore, we conduct experiments using several representative PEFT methods to better understand their effectiveness in parameter efficiency and memory efficiency. By offering insights into the latest advancements and practical applications, this survey serves as an invaluable resource for researchers and practitioners seeking to navigate the challenges and opportunities presented by PEFT in the context of PLMs.
The advent of large language models marks a revolutionary breakthrough in artificial intelligence. With the unprecedented scale of training and model parameters, the capability of large language models has been dramatically improved, leading to human-like performances in understanding, language synthesizing, and common-sense reasoning, etc. Such a major leap-forward in general AI capacity will change the pattern of how personalization is conducted. For one thing, it will reform the way of interaction between humans and personalization systems. Instead of being a passive medium of information filtering, large language models present the foundation for active user engagement. On top of such a new foundation, user requests can be proactively explored, and user's required information can be delivered in a natural and explainable way. For another thing, it will also considerably expand the scope of personalization, making it grow from the sole function of collecting personalized information to the compound function of providing personalized services. By leveraging large language models as general-purpose interface, the personalization systems may compile user requests into plans, calls the functions of external tools to execute the plans, and integrate the tools' outputs to complete the end-to-end personalization tasks. Today, large language models are still being developed, whereas the application in personalization is largely unexplored. Therefore, we consider it to be the right time to review the challenges in personalization and the opportunities to address them with LLMs. In particular, we dedicate this perspective paper to the discussion of the following aspects: the development and challenges for the existing personalization system, the newly emerged capabilities of large language models, and the potential ways of making use of large language models for personalization.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191