arXiv subjects

Yeow Meng Chee

Publications and source records attributed to Yeow Meng Chee.

At least 37 records · Page 2Linked to original sources

arXiv2024

On the Asymptotic Nonnegative Rank of Matrices and its Applications in Information Theory

In this paper, we study the asymptotic nonnegative rank of matrices, which characterizes the asymptotic growth of the nonnegative rank of fixed nonnegative matrices under the Kronecker product. This quantity is important since it governs several notions in information theory such as the so-called exact Rényi common information and the amortized communication complexity. By using the theory of asymptotic spectra of V. Strassen (J. Reine Angew. Math. 1988), we define formally the asymptotic spectrum of nonnegative matrices and give a dual characterization of the asymptotic nonnegative rank. As a complementary of the nonnegative rank, we introduce the notion of the subrank of a nonnegative matrix and show that it is exactly equal to the size of the maximum induced matching of the bipartite graph defined on the support of the matrix (therefore, independent of the value of entries). Finally, we show that two matrix parameters, namely rank and fractional cover number, belong to the asymptotic spectrum of nonnegative matrices.

cs.IT↗

arXiv2024

Fixed-Budget Differentially Private Best Arm Identification

We study best arm identification (BAI) in linear bandits in the fixed-budget regime under differential privacy constraints, when the arm rewards are supported on the unit interval. Given a finite budget $T$ and a privacy parameter $\varepsilon>0$, the goal is to minimise the error probability in finding the arm with the largest mean after $T$ sampling rounds, subject to the constraint that the policy of the decision maker satisfies a certain {\em $\varepsilon$-differential privacy} ($\varepsilon$-DP) constraint. We construct a policy satisfying the $\varepsilon$-DP constraint (called {\sc DP-BAI}) by proposing the principle of {\em maximum absolute determinants}, and derive an upper bound on its error probability. Furthermore, we derive a minimax lower bound on the error probability, and demonstrate that the lower and the upper bounds decay exponentially in $T$, with exponents in the two bounds matching order-wise in (a) the sub-optimality gaps of the arms, (b) $\varepsilon$, and (c) the problem complexity that is expressible as the sum of two terms, one characterising the complexity of standard fixed-budget BAI (without privacy constraints), and the other accounting for the $\varepsilon$-DP constraint. Additionally, we present some auxiliary results that contribute to the derivation of the lower bound on the error probability. These results, we posit, may be of independent interest and could prove instrumental in proving lower bounds on error probabilities in several other bandit problems. Whereas prior works provide results for BAI in the fixed-budget regime without privacy constraints or in the fixed-confidence regime with privacy constraints, our work fills the gap in the literature by providing the results for BAI in the fixed-budget regime under the $\varepsilon$-DP constraint.

cs.LG↗

arXiv2023

Federated Best Arm Identification with Heterogeneous Clients

We study best arm identification in a federated multi-armed bandit setting with a central server and multiple clients, when each client has access to a {\em subset} of arms and each arm yields independent Gaussian observations. The goal is to identify the best arm of each client subject to an upper bound on the error probability; here, the best arm is one that has the largest {\em average} value of the means averaged across all clients having access to the arm. Our interest is in the asymptotics as the error probability vanishes. We provide an asymptotic lower bound on the growth rate of the expected stopping time of any algorithm. Furthermore, we show that for any algorithm whose upper bound on the expected stopping time matches with the lower bound up to a multiplicative constant ({\em almost-optimal} algorithm), the ratio of any two consecutive communication time instants must be {\em bounded}, a result that is of independent interest. We thereby infer that an algorithm can communicate no more sparsely than at exponential time instants in order to be almost-optimal. For the class of almost-optimal algorithms, we present the first-of-its-kind asymptotic lower bound on the expected number of {\em communication rounds} until stoppage. We propose a novel algorithm that communicates at exponential time instants, and demonstrate that it is asymptotically almost-optimal.

cs.LG↗

arXiv2023

Learning to Search Feasible and Infeasible Regions of Routing Problems with Flexible Neural k-Opt

In this paper, we present Neural k-Opt (NeuOpt), a novel learning-to-search (L2S) solver for routing problems. It learns to perform flexible k-opt exchanges based on a tailored action factorization method and a customized recurrent dual-stream decoder. As a pioneering work to circumvent the pure feasibility masking scheme and enable the autonomous exploration of both feasible and infeasible regions, we then propose the Guided Infeasible Region Exploration (GIRE) scheme, which supplements the NeuOpt policy network with feasibility-related features and leverages reward shaping to steer reinforcement learning more effectively. Additionally, we equip NeuOpt with Dynamic Data Augmentation (D2A) for more diverse searches during inference. Extensive experiments on the Traveling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP) demonstrate that our NeuOpt not only significantly outstrips existing (masking-based) L2S solvers, but also showcases superiority over the learning-to-construct (L2C) and learning-to-predict (L2P) solvers. Notably, we offer fresh perspectives on how neural solvers can handle VRP constraints. Our code is available: https://github.com/yining043/NeuOpt.

cs.LG↗

arXiv2023

FLAC: A Robust Failure-Aware Atomic Commit Protocol for Distributed Transactions

In distributed transaction processing, atomic commit protocol (ACP) is used to ensure database consistency. With the use of commodity compute nodes and networks, failures such as system crashes and network partitioning are common. It is therefore important for ACP to dynamically adapt to the operating condition for efficiency while ensuring the consistency of the database. Existing ACPs often assume stable operating conditions, hence, they are either non-generalizable to different environments or slow in practice. In this paper, we propose a novel and practical ACP, called Failure-Aware Atomic Commit (FLAC). In essence, FLAC includes three protocols, which are specifically designed for three different environments: (i) no failure occurs, (ii) participant nodes might crash but there is no delayed connection, or (iii) both crashed nodes and delayed connection can occur. It models these environments as the failure-free, crash-failure, and network-failure robustness levels. During its operation, FLAC can monitor if any failure occurs and dynamically switch to operate the most suitable protocol, using a robustness level state machine, whose parameters are fine-tuned by reinforcement learning. Consequently, it improves both the response time and throughput, and effectively handles nodes distributed across the Internet where crash and network failures might occur. We implement FLAC in a distributed transactional key-value storage system based on Google Percolator and evaluate its performance with both a micro benchmark and a macro benchmark of real workload. The results show that FLAC achieves up to 2.22x throughput improvement and 2.82x latency speedup, compared to existing ACPs for high-contention workloads.

cs.DC↗

arXiv2023

Learning Generalizable Models for Vehicle Routing Problems via Knowledge Distillation

Recent neural methods for vehicle routing problems always train and test the deep models on the same instance distribution (i.e., uniform). To tackle the consequent cross-distribution generalization concerns, we bring the knowledge distillation to this field and propose an Adaptive Multi-Distribution Knowledge Distillation (AMDKD) scheme for learning more generalizable deep models. Particularly, our AMDKD leverages various knowledge from multiple teachers trained on exemplar distributions to yield a light-weight yet generalist student model. Meanwhile, we equip AMDKD with an adaptive strategy that allows the student to concentrate on difficult distributions, so as to absorb hard-to-master knowledge more effectively. Extensive experimental results show that, compared with the baseline neural methods, our AMDKD is able to achieve competitive results on both unseen in-distribution and out-of-distribution instances, which are either randomly synthesized or adopted from benchmark datasets (i.e., TSPLIB and CVRPLIB). Notably, our AMDKD is generic, and consumes less computational resources for inference.

cs.LG↗

arXiv2022

Efficient Neural Neighborhood Search for Pickup and Delivery Problems

We present an efficient Neural Neighborhood Search (N2S) approach for pickup and delivery problems (PDPs). In specific, we design a powerful Synthesis Attention that allows the vanilla self-attention to synthesize various types of features regarding a route solution. We also exploit two customized decoders that automatically learn to perform removal and reinsertion of a pickup-delivery node pair to tackle the precedence constraint. Additionally, a diversity enhancement scheme is leveraged to further ameliorate the performance. Our N2S is generic, and extensive experiments on two canonical PDP variants show that it can produce state-of-the-art results among existing neural methods. Moreover, it even outstrips the well-known LKH3 solver on the more constrained PDP variant. Our implementation for N2S is available online.

cs.LG↗

arXiv2022

Two dimensional RC/Subarray Constrained Codes: Bounded Weight and Almost Balanced Weight

In this work, we study two types of constraints on two-dimensional binary arrays. In particular, given $p,ε>0$, we study (i) The $p$-bounded constraint: a binary vector of size $m$ is said to be $p$-bounded if its weight is at most $pm$, and (ii) The $ε$-balanced constraint: a binary vector of size $m$ is said to be $ε$-balanced if its weight is within $[(0.5-ε)*m,(0.5+ε)*m]$. Such constraints are crucial in several data storage systems, those regard the information data as two-dimensional (2D) instead of one-dimensional (1D), such as the crossbar resistive memory arrays and the holographic data storage. In this work, efficient encoding/decoding algorithms are presented for binary arrays so that the weight constraint (either $p$-bounded constraint or $ε$-balanced constraint) is enforced over every row and every column, regarded as 2D row-column (RC) constrained codes; or over every subarray, regarded as 2D subarray constrained codes. While low-complexity designs have been proposed in the literature, mostly focusing on 2D RC constrained codes where $p = 1/2$ and $ε= 0$, this work provides efficient coding methods that work for both 2D RC constrained codes and 2D subarray constrained codes, and more importantly, the methods are applicable for arbitrary values of $p$ and $ε$. Furthermore, for certain values of $p$ and $ε$, we show that, for sufficiently large array size, there exists linear-time encoding/decoding algorithm that incurs at most one redundant bit.

cs.IT↗

arXiv2022

Neural Network Decoders for Permutation Codes Correcting Different Errors

Permutation codes were extensively studied in order to correct different types of errors for the applications on power line communication and rank modulation for flash memory. In this paper, we introduce the neural network decoders for permutation codes to correct these errors with one-shot decoding, which treat the decoding as $n$ classification tasks for non-binary symbols for a code of length $n$. These are actually the first general decoders introduced to deal with any error type for these two applications. The performance of the decoders is evaluated by simulations with different error models.

cs.IT↗

arXiv2022

Primitive3D: 3D Object Dataset Synthesis from Randomly Assembled Primitives

Numerous advancements in deep learning can be attributed to the access to large-scale and well-annotated datasets. However, such a dataset is prohibitively expensive in 3D computer vision due to the substantial collection cost. To alleviate this issue, we propose a cost-effective method for automatically generating a large amount of 3D objects with annotations. In particular, we synthesize objects simply by assembling multiple random primitives. These objects are thus auto-annotated with part labels originating from primitives. This allows us to perform multi-task learning by combining the supervised segmentation with unsupervised reconstruction. Considering the large overhead of learning on the generated dataset, we further propose a dataset distillation strategy to remove redundant samples regarding a target dataset. We conduct extensive experiments for the downstream tasks of 3D object classification. The results indicate that our dataset, together with multi-task pretraining on its annotations, achieves the best performance compared to other commonly used datasets. Further study suggests that our strategy can improve the model performance by pretraining and fine-tuning scheme, especially for the dataset with a small scale. In addition, pretraining with the proposed dataset distillation method can save 86\% of the pretraining time with negligible performance degradation. We expect that our attempt provides a new data-centric perspective for training 3D deep models.

cs.CV↗

arXiv2022

Serverless Data Science -- Are We There Yet? A Case Study of Model Serving

Machine learning (ML) is an important part of modern data science applications. Data scientists today have to manage the end-to-end ML life cycle that includes both model training and model serving, the latter of which is essential, as it makes their works available to end-users. Systems of model serving require high performance, low cost, and ease of management. Cloud providers are already offering model serving choices, including managed services and self-rented servers. Recently, serverless computing, whose advantages include high elasticity and a fine-grained cost model, brings another option for model serving. Our goal in this paper is to examine the viability of serverless as a mainstream model serving platform. To this end, we first conduct a comprehensive evaluation of the performance and cost of serverless against other model serving systems on Amazon Web Service and Google Cloud Platform. We find that serverless outperforms many cloud-based alternatives. Further, there are settings under which it even achieves better performance than GPU-based systems. Next, we present the design space of serverless model serving, which comprises multiple dimensions, including cloud platforms, serving runtimes, and other function-specific parameters. For each dimension, we analyze the impact of different choices and provide suggestions for data scientists to better utilize serverless model serving. Finally, we discuss challenges and opportunities in building a more practical serverless model serving system.

cs.DC↗

arXiv2022

Cost-Effective Algorithms for Average-Case Interactive Graph Search

Interactive graph search (IGS) uses human intelligence to locate the target node in hierarchy, which can be applied for image classification, product categorization and searching a database. Specifically, IGS aims to categorize an object from a given category hierarchy via several rounds of interactive queries. In each round of query, the search algorithm picks a category and receives a boolean answer on whether the object is under the chosen category. The main efficiency goal asks for the minimum number of queries to identify the correct hierarchical category for the object. In this paper, we study the average-case interactive graph search (AIGS) problem that aims to minimize the expected number of queries when the objects follow a probability distribution. We propose a greedy search policy that splits the candidate categories as evenly as possible with respect to the probability weights, which offers an approximation guarantee of $O(\log n)$ for AIGS given the category hierarchy is a directed acyclic graph (DAG), where $n$ is the total number of categories. Meanwhile, if the input hierarchy is a tree, we show that a constant approximation factor of $(1+\sqrt{5})/2$ can be achieved. Furthermore, we present efficient implementations of the greedy policy, namely GreedyTree and GreedyDAG, that can quickly categorize the object in practice. Extensive experiments in real-world scenarios are carried out to demonstrate the superiority of our proposed methods.

cs.DB↗

arXiv2021

Endurance-Limited Memories: Capacity and Codes

\emph{Resistive memories}, such as \emph{phase change memories} and \emph{resistive random access memories} have attracted significant attention in recent years due to their better scalability, speed, rewritability, and yet non-volatility. However, their \emph{limited endurance} is still a major drawback that has to be improved before they can be widely adapted in large-scale systems. In this work, in order to reduce the wear out of the cells, we propose a new coding scheme, called \emph{endurance-limited memories} (\emph{ELM}) codes, that increases the endurance of these memories by limiting the number of cell programming operations. Namely, an \emph{$\ell$-change $t$-write ELM code} is a coding scheme that allows to write $t$ messages into some $n$ binary cells while guaranteeing that each cell is programmed at most $\ell$ times. In case $\ell=1$, these codes coincide with the well-studied \emph{write-once memory} (\emph{WOM}) codes. We study some models of these codes which depend upon whether the encoder knows on each write the number of times each cell was programmed, knows only the memory state, or even does not know anything. For the decoder, we consider these similar three cases. We fully characterize the capacity regions and the maximum sum-rates of three models where the encoder knows on each write the number of times each cell was programmed. In particular, it is shown that in these models the maximum sum-rate is $\log \sum_{i=0}^{\ell} {t \choose i}$. We also study and expose the capacity regions of the models where the decoder is informed with the number of times each cell was programmed. Finally we present the most practical model where the encoder read the memory before encoding new data and the decoder has no information about the previous states of the memory.

cs.IT↗

arXiv2020

Explicit Baranyai Partitions for Quadruples, Part I: Quadrupling Constructions

It is well known that, whenever $k$ divides $n$, the complete $k$-uniform hypergraph on $n$ vertices can be partitioned into disjoint perfect matchings. Equivalently, the set of $k$-subsets of an $n$-set can be partitioned into parallel classes so that each parallel class is a partition of the $n$-set. This result is known as Baranyai's theorem, which guarantees the existence of \emph{Baranyai partitions}. Unfortunately, the proof of Baranyai's theorem uses network flow arguments, making this result non-explicit. In particular, there is no known method to produce Baranyai partitions in time and space that scale linearly with the number of hyperedges in the hypergraph. It is desirable for certain applications to have an explicit construction that generates Baranyai partitions in linear time. Such an efficient construction is known for $k=2$ and $k=3$. In this paper, we present an explicit recursive quadrupling construction for $k=4$ and $n=4t$, where $t \equiv 0,3,4,6,8,9 ~(\text{mod}~12)$. In a follow-up paper (Part II), the other values of~$t$, namely $t \equiv 1,2,5,7,10,11 ~(\text{mod}~12)$, will be considered.

math.CO↗

arXiv2020

Constrained de Bruijn Codes: Properties, Enumeration, Constructions, and Applications

The de Bruijn graph, its sequences, and their various generalizations, have found many applications in information theory, including many new ones in the last decade. In this paper, motivated by a coding problem for emerging memory technologies, a set of sequences which generalize sequences in the de Bruijn graph are defined. These sequences can be also defined and viewed as constrained sequences. Hence, they will be called constrained de Bruijn sequences and a set of such sequences will be called a constrained de Bruijn code. Several properties and alternative definitions for such codes are examined and they are analyzed as generalized sequences in the de Bruijn graph (and its generalization) and as constrained sequences. Various enumeration techniques are used to compute the total number of sequences for any given set of parameters. A construction method of such codes from the theory of shift-register sequences is proposed. Finally, we show how these constrained de Bruijn sequences and codes can be applied in constructions of codes for correcting synchronization errors in the $\ell$-symbol read channel and in the racetrack memory channel. For this purpose, these codes are superior in their size on previously known codes.

cs.IT↗

arXiv2019

Robust Positioning Patterns with Low Redundancy

A robust positioning pattern is a large array that allows a mobile device to locate its position by reading a possibly corrupted small window around it. In this paper, we provide constructions of binary positioning patterns, equipped with efficient locating algorithms, that are robust to a constant number of errors and have redundancy within a constant factor of optimality. Furthermore, we modify our constructions to correct rank errors and obtain binary positioning patterns robust to any errors of rank less than a constant number. Additionally, we construct $q$-ary robust positioning sequences robust to a large number of errors, some of which have length attaining the upper bound. Our construction of binary positioning sequences that are robust to a constant number of errors has the least known redundancy amongst those explicit constructions with efficient locating algorithms. On the other hand, for binary robust positioning arrays, our construction is the first explicit construction whose redundancy is within a constant factor of optimality. The locating algorithms accompanying both constructions run in time cubic in sequence length or array dimension.

cs.IT↗

arXiv2019

Efficient Algorithm for the Linear Complexity of Sequences and Some Related Consequences

The linear complexity of a sequence $s$ is one of the measures of its predictability. It represents the smallest degree of a linear recursion which the sequence satisfies. There are several algorithms to find the linear complexity of a periodic sequence $s$ of length $N$ (where $N$ is of some given form) over a finite field $F_q$ in $O(N)$ symbol field operations. The first such algorithm is The Games-Chan Algorithm which considers binary sequences of period $2^n$, and is known for its extreme simplicity. We generalize this algorithm and apply it efficiently for several families of binary sequences. Our algorithm is very simple, it requires $βN$ bit operations for a small constant $β$, where $N$ is the period of the sequence. We make an analysis on the number of bit operations required by the algorithm and compare it with previous algorithms. In the process, the algorithm also finds the recursion for the shortest linear feedback shift-register which generates the sequence. Some other interesting properties related to shift-register sequences, which might not be too surprising but generally unnoted, are also consequences of our exposition.

cs.CR↗

arXiv2019

Optimal Codes Correcting a Single Indel / Edit for DNA-Based Data Storage

An indel refers to a single insertion or deletion, while an edit refers to a single insertion, deletion or substitution. In this paper, we investigate codes that combat either a single indel or a single edit and provide linear-time algorithms that encode binary messages into these codes of length n. Over the quaternary alphabet, we provide two linear-time encoders. One corrects a single edit with log n + O(log log n) redundancy bits, while the other corrects a single indel with log n + 2 redundant bits. These two encoders are order-optimal. The former encoder is the first known order-optimal encoder that corrects a single edit, while the latter encoder (that corrects a single indel) reduces the redundancy of the best known encoder of Tenengolts (1984) by at least four bits. Over the DNA alphabet, we impose an additional constraint: the GC-balanced constraint and require that exactly half of the symbols of any DNA codeword to be either C or G. In particular, via a modification of Knuth's balancing technique, we provide a linear-time map that translates binary messages into GC-balanced codewords and the resulting codebook is able to correct a single indel or a single edit. These are the first known constructions of GC-balanced codes that correct a single indel or a single edit.

cs.IT↗