arXiv subjects

Khoa Trinh

Publications and source records attributed to Khoa Trinh.

11 recordsLinked to original sources

arXiv2025

GUIDE: Guided Initialization and Distillation of Embeddings

Algorithmic efficiency techniques such as distillation (\cite{hinton2015distillation}) are useful in improving model quality without increasing serving costs, provided a larger teacher model is available for a smaller student model to learn from during training. Standard distillation methods are limited to only forcing the student to match the teacher's outputs. Given the costs associated with training a large model, we believe we should be extracting more useful information from a teacher model than by just making the student match the teacher's outputs. In this paper, we introduce \guide (Guided Initialization and Distillation of Embeddings). \guide can be considered a distillation technique that forces the student to match the teacher in the parameter space. Using \guide we show 25-26\% reduction in the teacher-student quality gap when using large student models (400M - 1B parameters) trained on $\approx$ 20B tokens. We also present a thorough analysis demonstrating that \guide can be combined with knowledge distillation with near additive improvements. Furthermore, we show that applying \guide alone leads to substantially better model quality than applying knowledge distillation by itself. Most importantly, \guide introduces no training or inference overhead and hence any model quality gains from our method are virtually free.

cs.LG↗

arXiv2024

Dependent randomized rounding for clustering and partition systems with knapsack constraints

Clustering problems are fundamental to unsupervised learning. There is an increased emphasis on fairness in machine learning and AI; one representative notion of fairness is that no single demographic group should be over-represented among the cluster-centers. This, and much more general clustering problems, can be formulated with "knapsack" and "partition" constraints. We develop new randomized algorithms targeting such problems, and study two in particular: multi-knapsack median and multi-knapsack center. Our rounding algorithms give new approximation and pseudo-approximation algorithms for these problems. One key technical tool, which may be of independent interest, is a new tail bound analogous to Feige (2006) for sums of random variables with unbounded variances. Such bounds can be useful in inferring properties of large networks using few samples.

cs.DS↗

arXiv2023

SLaM: Student-Label Mixing for Distillation with Unlabeled Examples

Knowledge distillation with unlabeled examples is a powerful training paradigm for generating compact and lightweight student models in applications where the amount of labeled data is limited but one has access to a large pool of unlabeled data. In this setting, a large teacher model generates ``soft'' pseudo-labels for the unlabeled dataset which are then used for training the student model. Despite its success in a wide variety of applications, a shortcoming of this approach is that the teacher's pseudo-labels are often noisy, leading to impaired student performance. In this paper, we present a principled method for knowledge distillation with unlabeled examples that we call Student-Label Mixing (SLaM) and we show that it consistently improves over prior approaches by evaluating it on several standard benchmarks. Finally, we show that SLaM comes with theoretical guarantees; along the way we give an algorithm improving the best-known sample complexity for learning halfspaces with margin under random classification noise, and provide the first convergence analysis for so-called ``forward loss-adjustment" methods.

cs.LG↗

arXiv2023

Robust Active Distillation

Distilling knowledge from a large teacher model to a lightweight one is a widely successful approach for generating compact, powerful models in the semi-supervised learning setting where a limited amount of labeled data is available. In large-scale applications, however, the teacher tends to provide a large number of incorrect soft-labels that impairs student performance. The sheer size of the teacher additionally constrains the number of soft-labels that can be queried due to prohibitive computational and/or financial costs. The difficulty in achieving simultaneous \emph{efficiency} (i.e., minimizing soft-label queries) and \emph{robustness} (i.e., avoiding student inaccuracies due to incorrect labels) hurts the widespread application of knowledge distillation to many modern tasks. In this paper, we present a parameter-free approach with provable guarantees to query the soft-labels of points that are simultaneously informative and correctly labeled by the teacher. At the core of our work lies a game-theoretic formulation that explicitly considers the inherent trade-off between the informativeness and correctness of input instances. We establish bounds on the expected performance of our approach that hold even in worst-case distillation instances. We present empirical evaluations on popular benchmarks that demonstrate the improved distillation performance enabled by our work relative to that of state-of-the-art active learning and active distillation methods.

cs.LG↗

arXiv2022

Improved Bi-point Rounding Algorithms and a Golden Barrier for $k$-Median

The current best approximation algorithms for $k$-median rely on first obtaining a structured fractional solution known as a bi-point solution, and then rounding it to an integer solution. We improve this second step by unifying and refining previous approaches. We describe a hierarchy of increasingly-complex partitioning schemes for the facilities, along with corresponding sets of algorithms and factor-revealing non-linear programs. We prove that the third layer of this hierarchy is a $2.613$-approximation, improving upon the current best ratio of $2.675$, while no layer can be proved better than $2.588$ under the proposed analysis. On the negative side, we give a family of bi-point solutions which cannot be approximated better than the square root of the golden ratio, even if allowed to open $k+o(k)$ facilities. This gives a barrier to current approaches for obtaining an approximation better than $2 \sqrtϕ \approx 2.544$. Altogether we reduce the approximation gap of bi-point solutions by two thirds.

cs.DS↗

arXiv2022

Weighted Distillation with Unlabeled Examples

Distillation with unlabeled examples is a popular and powerful method for training deep neural networks in settings where the amount of labeled data is limited: A large ''teacher'' neural network is trained on the labeled data available, and then it is used to generate labels on an unlabeled dataset (typically much larger in size). These labels are then utilized to train the smaller ''student'' model which will actually be deployed. Naturally, the success of the approach depends on the quality of the teacher's labels, since the student could be confused if trained on inaccurate data. This paper proposes a principled approach for addressing this issue based on a ''debiasing'' reweighting of the student's loss function tailored to the distillation training paradigm. Our method is hyper-parameter free, data-agnostic, and simple to implement. We demonstrate significant improvements on popular academic datasets and we accompany our results with a theoretical analysis which rigorously justifies the performance of our method in certain settings.

cs.LG↗

arXiv2020

Approximation algorithms for stochastic clustering

We consider stochastic settings for clustering, and develop provably-good approximation algorithms for a number of these notions. These algorithms yield better approximation ratios compared to the usual deterministic clustering setting. Additionally, they offer a number of advantages including clustering which is fairer and has better long-term behavior for each user. In particular, they ensure that *every user* is guaranteed to get good service (on average). We also complement some of these with impossibility results.

cs.DS↗

arXiv2017

A Lottery Model for Center-type Problems With Outliers

In this paper, we give tight approximation algorithms for the $k$-center and matroid center problems with outliers. Unfairness arises naturally in this setting: certain clients could always be considered as outliers. To address this issue, we introduce a lottery model in which each client $j$ is allowed to submit a parameter $p_j \in [0,1]$ and we look for a random solution that covers every client $j$ with probability at least $p_j$. Our techniques include a randomized rounding procedure to round a point inside a matroid intersection polytope to a basis plus at most one extra item such that all marginal probabilities are preserved and such that a certain linear function of the variables does not decrease in the process with probability one.

cs.DS↗

arXiv2017

Fairness in Resource Allocation and Slowed-down Dependent Rounding

We consider an issue of much current concern: could fairness, an issue that is already difficult to guarantee, worsen when algorithms run much of our lives? We consider this in the context of resource-allocation problems, we show that algorithms can guarantee certain types of fairness in a verifiable way. Our conceptual contribution is a simple approach to fairness in this context, which only requires that all users trust some public lottery. Our technical contributions are in ways to address the $k$-center and knapsack-center problems that arise in this context: we develop a novel dependent-rounding technique that, via the new ingredients of "slowing down" and additional randomization, guarantees stronger correlation properties than known before.

cs.DS↗

arXiv2016

An Improved Approximation for $k$-median, and Positive Correlation in Budgeted Optimization

Dependent rounding is a useful technique for optimization problems with hard budget constraints. This framework naturally leads to \emph{negative correlation} properties. However, what if an application naturally calls for dependent rounding on the one hand, and desires \emph{positive} correlation on the other? More generally, we develop algorithms that guarantee the known properties of dependent rounding, but also have nearly best-possible behavior - near-independence, which generalizes positive correlation - on "small" subsets of the variables. The recent breakthrough of Li & Svensson for the classical $k$-median problem has to handle positive correlation in certain dependent-rounding settings, and does so implicitly. We improve upon Li-Svensson's approximation ratio for $k$-median from $2.732 + ε$ to $2.675 + ε$ by developing an algorithm that improves upon various aspects of their work. Our dependent-rounding approach helps us improve the dependence of the runtime on the parameter $ε$ from Li-Svensson's $N^{O(1/ε^2)}$ to $N^{O((1/ε) \log(1/ε))}$.

cs.DS↗

arXiv2014

'Beating the news' with EMBERS: Forecasting Civil Unrest using Open Source Indicators

We describe the design, implementation, and evaluation of EMBERS, an automated, 24x7 continuous system for forecasting civil unrest across 10 countries of Latin America using open source indicators such as tweets, news sources, blogs, economic indicators, and other data sources. Unlike retrospective studies, EMBERS has been making forecasts into the future since Nov 2012 which have been (and continue to be) evaluated by an independent T&E team (MITRE). Of note, EMBERS has successfully forecast the uptick and downtick of incidents during the June 2013 protests in Brazil. We outline the system architecture of EMBERS, individual models that leverage specific data sources, and a fusion and suppression engine that supports trading off specific evaluation criteria. EMBERS also provides an audit trail interface that enables the investigation of why specific predictions were made along with the data utilized for forecasting. Through numerous evaluations, we demonstrate the superiority of EMBERS over baserate methods and its capability to forecast significant societal happenings.

cs.SI↗