arXiv subjects

Yuzhen Huang

Publications and source records attributed to Yuzhen Huang.

27 records · Page 2Linked to original sources

arXiv2024

Movable Antenna Array Aided Ultra Reliable Covert Communications

In this paper, we construct a framework of the movable antenna (MA) aided covert communication shielded by the general noise uncertainty for the first time. According to the analysis performance on the derived closed-form expressions of the sum of the probabilities of the detection errors and the communication outage probability, the perfect covertness and the ultra reliability can be achieved by adjusting the antenna position in the MA array. Then, we formulate the communication covertness maximization problem with the constraints of the ultra reliability and the independent discrete movable position to optimize the transmitter's parameter. With the maximal ratio transmitting (MRT) design for the beamforming, we solve the closed-form optimal information transmit power and design a lightweight discrete projected gradient descent (DPGD) algorithm to determine the optimal antenna position. The numerical results show that the optimal achievable covertness and the feasible region of the steering angle with the MA array is significant larger than the one with the fixed-position antenna (FPA) array.

cs.IT↗

arXiv2024

Compression Represents Intelligence Linearly

There is a belief that learning to compress well will lead to intelligence. Recently, language modeling has been shown to be equivalent to compression, which offers a compelling rationale for the success of large language models (LLMs): the development of more advanced language models is essentially enhancing compression which facilitates intelligence. Despite such appealing discussions, little empirical evidence is present for the interplay between compression and intelligence. In this work, we examine their relationship in the context of LLMs, treating LLMs as data compressors. Given the abstract concept of "intelligence", we adopt the average downstream benchmark scores as a surrogate, specifically targeting intelligence related to knowledge and commonsense, coding, and mathematical reasoning. Across 12 benchmarks, our study brings together 31 public LLMs that originate from diverse organizations. Remarkably, we find that LLMs' intelligence -- reflected by average benchmark scores -- almost linearly correlates with their ability to compress external text corpora. These results provide concrete evidence supporting the belief that superior compression indicates greater intelligence. Furthermore, our findings suggest that compression efficiency, as an unsupervised metric derived from raw text corpora, serves as a reliable evaluation measure that is linearly associated with the model capabilities. We open-source our compression datasets as well as our data collection pipelines to facilitate future researchers to assess compression properly.

cs.CL↗

arXiv2023

C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models

New NLP benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present C-Eval, the first comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context. C-Eval comprises multiple-choice questions across four difficulty levels: middle school, high school, college, and professional. The questions span 52 diverse disciplines, ranging from humanities to science and engineering. C-Eval is accompanied by C-Eval Hard, a subset of very challenging subjects in C-Eval that requires advanced reasoning abilities to solve. We conduct a comprehensive evaluation of the most advanced LLMs on C-Eval, including both English- and Chinese-oriented models. Results indicate that only GPT-4 could achieve an average accuracy of over 60%, suggesting that there is still significant room for improvement for current LLMs. We anticipate C-Eval will help analyze important strengths and shortcomings of foundation models, and foster their development and growth for Chinese users.

cs.CL↗

arXiv2023

Pre-train and Search: Efficient Embedding Table Sharding with Pre-trained Neural Cost Models

Sharding a large machine learning model across multiple devices to balance the costs is important in distributed training. This is challenging because partitioning is NP-hard, and estimating the costs accurately and efficiently is difficult. In this work, we explore a "pre-train, and search" paradigm for efficient sharding. The idea is to pre-train a universal and once-for-all neural network to predict the costs of all the possible shards, which serves as an efficient sharding simulator. Built upon this pre-trained cost model, we then perform an online search to identify the best sharding plans given any specific sharding task. We instantiate this idea in deep learning recommendation models (DLRMs) and propose NeuroShard for embedding table sharding. NeuroShard pre-trains neural cost models on augmented tables to cover various sharding scenarios. Then it identifies the best column-wise and table-wise sharding plans with beam search and greedy grid search, respectively. Experiments show that NeuroShard significantly and consistently outperforms the state-of-the-art on the benchmark sharding dataset, achieving up to 23.8% improvement. When deployed in an ultra-large production DLRM with multi-terabyte embedding tables, NeuroShard achieves 11.6% improvement in embedding costs over the state-of-the-art, which translates to 6.6% end-to-end training throughput improvement. To facilitate future research of the "pre-train, and search" paradigm in ML for Systems, we open-source our code at https://github.com/daochenzha/neuroshard

cs.LG↗

arXiv2021

Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism

In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communication Thresholding (DCT) for communication-efficient hybrid training. DCT filters the entities to be communicated across the network through a simple hard-thresholding function, allowing only the most relevant information to pass through. For communication efficient DP, DCT compresses the parameter gradients sent to the parameter server during model synchronization. The threshold is updated only once every few thousand iterations to reduce the computational overhead of compression. For communication efficient MP, DCT incorporates a novel technique to compress the activations and gradients sent across the network during the forward and backward propagation, respectively. This is done by identifying and updating only the most relevant neurons of the neural network for each training sample in the data. We evaluate DCT on publicly available natural language processing and recommender models and datasets, as well as recommendation systems used in production at Facebook. DCT reduces communication by at least $100\times$ and $20\times$ during DP and MP, respectively. The algorithm has been deployed in production, and it improves end-to-end training time for a state-of-the-art industrial recommender model by 37\%, without any loss in performance.

cs.LG↗

arXiv2020

TensorOpt: Exploring the Tradeoffs in Distributed DNN Training with Auto-Parallelism

A good parallelization strategy can significantly improve the efficiency or reduce the cost for the distributed training of deep neural networks (DNNs). Recently, several methods have been proposed to find efficient parallelization strategies but they all optimize a single objective (e.g., execution time, memory consumption) and produce only one strategy. We propose FT, an efficient algorithm that searches for an optimal set of parallelization strategies to allow the trade-off among different objectives. FT can adapt to different scenarios by minimizing the memory consumption when the number of devices is limited and fully utilize additional resources to reduce the execution time. For popular DNN models (e.g., vision, language), an in-depth analysis is conducted to understand the trade-offs among different objectives and their influence on the parallelization strategies. We also develop a user-friendly system, called TensorOpt, which allows users to run their distributed DNN training jobs without caring the details of parallelization strategies. Experimental results show that FT runs efficiently and provides accurate estimation of runtime costs, and TensorOpt is more flexible in adapting to resource availability compared with existing frameworks.

cs.DC↗

arXiv2019

Elastic deep learning in multi-tenant GPU cluster

We study how to support elasticity, i.e., the ability to dynamically adjust the parallelism (number of GPUs), for deep neural network (DNN) training. Elasticity can benefit multi-tenant GPU cluster management in many ways, e.g., achieving various scheduling objectives (e.g., job throughput, job completion time, GPU efficiency) according to cluster load variations, maximizing the use of transient idle resources, performance profiling, job migration, and straggler mitigation. However, existing parallelism adjustment strategies incur high overheads, which hinder many applications from making effective use of elasticity. We propose EDL to enable low-overhead elastic deep learning with a simple API. We present techniques that are necessary to reduce the overhead of parallelism adjustments, such as stop-free scaling and dynamic data pipeline. We also demonstrate that EDL can indeed bring significant benefits to the above-listed applications in GPU cluster management.

cs.DC↗

arXiv2016

Efficient Processing of Reachability and Time-Based Path Queries in a Temporal Graph

A temporal graph is a graph in which vertices communicate with each other at specific time, e.g., $A$ calls $B$ at 11 a.m. and talks for 7 minutes, which is modeled by an edge from $A$ to $B$ with starting time "11 a.m." and duration "7 mins". Temporal graphs can be used to model many networks with time-related activities, but efficient algorithms for analyzing temporal graphs are severely inadequate. We study fundamental problems such as answering reachability and time-based path queries in a temporal graph, and propose an efficient indexing technique specifically designed for processing these queries in a temporal graph. Our results show that our method is efficient and scalable in both index construction and query processing.

cs.DB↗

arXiv2016

Efficient Processing of Very Large Graphs in a Small Cluster

Inspired by the success of Google's Pregel, many systems have been developed recently for iterative computation over big graphs. These systems provide a user-friendly vertex-centric programming interface, where a programmer only needs to specify the behavior of one generic vertex when developing a parallel graph algorithm. However, most existing systems require the input graph to reside in memories of the machines in a cluster, and the few out-of-core systems suffer from problems such as poor efficiency for sparse computation workload, high demand on network bandwidth, and expensive cost incurred by external-memory join and group-by. In this paper, we introduce the GraphD system for a user to process very large graphs with ordinary computing resources. GraphD fully overlaps computation with communication, by streaming edges and messages on local disks, while transmitting messages in parallel. For a broad class of Pregel algorithms where message combiner is applicable, GraphD eliminates the need of any expensive external-memory join or group-by. These key techniques allow GraphD to achieve comparable performance to in-memory Pregel-like systems without keeping edges and messages in memories. We prove that to process a graph G=(V, E) with n machines using GraphD, each machine only requires O(|V|/n) memory space, allowing GraphD to scale to very large graphs with a small cluster. Extensive experiments show that GraphD beats existing out-of-core systems by orders of magnitude, and achieves comparable performance to in-memory systems running with enough memories.

cs.DC↗