Filters








11,248 Hits in 10.4 sec

Beyond Fine Tuning: A Modular Approach to Learning on Small Data [article]

Ark Anderson, Kyle Shaffer, Artem Yankov, Court D. Corley, Nathan O. Hodas
2016 arXiv   pre-print
The central impact of using a modular approach comes from adding new representations to a network, as opposed to replacing representations via fine-tuning.  ...  In this paper we present a technique to train neural network models on small amounts of data.  ...  This allows the modular approach to more robustly handle small data sets than naive fine-tuning.  ... 
arXiv:1611.01714v1 fatcat:ag3oajbzu5ddvkkb36z6v3qev4

Beyond Fine-tuning: Few-Sample Sentence Embedding Transfer [article]

Siddhant Garg, Rohit Kumar Sharma, Yingyu Liang
2020 arXiv   pre-print
Fine-tuning (FT) pre-trained sentence embedding models on small datasets has been shown to have limitations.  ...  We perform evaluation on seven small datasets from NLP tasks and show that our approach with end-to-end training outperforms FT with negligible computational overhead.  ...  The authors would also like to acknowledge the support provided by the University of Wisconsin-Madison Office of the Vice Chancellor for Research and Graduate Education with funding from the Wisconsin  ... 
arXiv:2004.05119v2 fatcat:vyt2nlsryjhdnhjgemikmmgehi

Lightweight Adapter Tuning for Multilingual Speech Translation [article]

Hang Le, Juan Pino, Changhan Wang, Jiatao Gu, Didier Schwab, Laurent Besacier
2021 arXiv   pre-print
Starting from different pre-trained models (a multilingual ST trained on parallel data or a multilingual BART (mBART) trained on non-parallel multilingual data), we show that adapters can be used to: (  ...  Adapter modules were recently introduced as an efficient alternative to fine-tuning in NLP.  ...  Acknowledgments This work was supported by a Facebook AI SRA grant, and was granted access to the HPC resources of IDRIS under the allocation 2020-AD011011695 made by GENCI.  ... 
arXiv:2106.01463v2 fatcat:754ss6gtzbhpjbpcn2ijspmvae

Differentially Private Fine-tuning of Language Models [article]

Da Yu, Saurabh Naik, Arturs Backurs, Sivakanth Gopi, Huseyin A. Inan, Gautam Kamath, Janardhan Kulkarni, Yin Tat Lee, Andre Manoel, Lukas Wutschitz, Sergey Yekhanin, Huishuai Zhang
2021 arXiv   pre-print
We propose a meta-framework for this problem, inspired by the recent success of highly parameter-efficient methods for fine-tuning.  ...  All our experiments suggest that larger models are better suited for private fine-tuning: while they are well known to achieve superior accuracy non-privately, we find that they also better maintain their  ...  Janardhan Kulkarni would like to thank Edward Hu for sharing many ideas on fine-tuning.  ... 
arXiv:2110.06500v1 fatcat:p5kk4zuodbdftlx7jf3yhj45dm

Prefix-Tuning: Optimizing Continuous Prompts for Generation [article]

Xiang Lisa Li, Percy Liang
2021 arXiv   pre-print
We find that by learning only 0.1\% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples  ...  In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen, but optimizes a small continuous task-specific  ...  A natural approach to this problem is lightweight fine-tuning, which freezes most of the pretrained parameters and augments the model with small trainable modules.  ... 
arXiv:2101.00190v1 fatcat:bdhj3qnsufcxjml24cndpx43s4

BayesAdapter: Being Bayesian, Inexpensively and Reliably, via Bayesian Fine-tuning [article]

Zhijie Deng, Hao Zhang, Xiao Yang, Yinpeng Dong, Jun Zhu
2021 arXiv   pre-print
In particular, we propose to adapt the pre-trained deterministic NNs to be BNNs via cost-effective Bayesian fine-tuning.  ...  To make BayesAdapter more practical, we technically contribute 1) a modularized, user-friendly implementation for the learning of variational BNNs under two representative variational distributions, 2)  ...  As a solution, we opt to explicitly regularize the variational BNNs to behave uncertainly on a collection of OOD data during Bayesian fine-tuning.  ... 
arXiv:2010.01979v4 fatcat:nhiu3neenjb7johs5hc5husyiq

Enabling Technology for Remote Prosthetic Alignment Tuning

David A Boone, Sarah R Chang
2021 Military medicine  
Target alignments were calculated by the machine learning algorithm in the ProSAT software, based on input of kinetic data samples representing the precondition and where a real prosthetic misalignment  ...  The sensor has been cross-validated against kinetic measurement in a gait laboratory, and bench testing was performed to validate the performance of the tool while adjusting a prosthetic socket based on  ...  We also thank Jonathon Maier, Jung Kim, and Stephen Silverstein for the work on engineering the software foundation that enables this teleprosthetics mobile application.  ... 
doi:10.1093/milmed/usaa453 pmid:33499549 fatcat:uojdvtrv4rbyto7a6njfx6rhxq

Milepost GCC: Machine Learning Enabled Self-tuning Compiler

Grigori Fursin, Yuriy Kashnikov, Abdul Wahid Memon, Zbigniew Chamski, Olivier Temam, Mircea Namolaru, Elad Yom-Tov, Bilha Mendelson, Ayal Zaks, Eric Courtois, Francois Bodin, Phil Barnard (+5 others)
2011 International journal of parallel programming  
Our approach is to develop a modular, extensible, self-tuning optimization infrastructure to automatically learn the best optimizations across multiple programs and architectures based on the correlation  ...  We developed machine learning plugins based on probabilistic and transductive approaches to predict good combinations of optimizations.  ...  Using machine learning to predict good optimization passes The Milepost approach to learning optimizations across programs is based on the observation that programs may exhibit similar behavior for a similar  ... 
doi:10.1007/s10766-010-0161-2 fatcat:r6s7qcgunzf5hcgllcorvkir4m

Scalable Auto-Tuning of Synthesis Parameters for Optimizing High-Performance Processors

Matthew M. Ziegler, Hung-Yi Liu, Luca P. Carloni
2016 Proceedings of the 2016 International Symposium on Low Power Electronics and Design - ISLPED '16  
In this paper we present a novel learning-based algorithm for synthesis parameter optimization.  ...  Architecture of the STS process, which employs a parallel and iterative tuning process to optimize macros [2].  ...  Moreover, to better estimate the cost based on non-trivial contributing scenarios (i.e., scenarios comprising more than one primitive), the Learning algorithm includes a fine-grained cost estimation (see  ... 
doi:10.1145/2934583.2934620 dblp:conf/islped/ZieglerLC16 fatcat:f6kj64s4orbuhfiarlotr5lcze

Hominin interbreeding and language evolution: fine-tuning the details

Antonio Benítez-Burraco, Lluís Barceló-Coblijn
2013 Journal of Anthropological Sciences  
A second key concern raised by Bruner is that we must rely on all evidences if we intend to reach hypotheses that are probable (and eventually, to be able to falsify them).  ...  At present we are in position to try to answer this question by putting together all the archaeological, paleo-neurobiological, genetic, and even molecular data available to date.  ...  In doing so they go beyond Premo's position and explicitly argue for a computational approach to the problem that focuses on the procedural system required for planning and executing the involved motor  ... 
doi:10.4436/jass.91020 pmid:24334493 fatcat:tx7wxbomxffuxhmgeg6l4xy4lm

DBA bandits: Self-driving index tuning under ad-hoc, analytical workloads with safety guarantees [article]

R. Malinga Perera, Bastian Oetomo, Benjamin I. P. Rubinstein, Renata Borovica-Gajic
2020 arXiv   pre-print
Our comprehensive empirical results demonstrate up to 75% speed-up on shifting and ad-hoc workloads and 28% speed-up on static workloads compared against a state-of-the-art commercial tuning tool.  ...  We propose a self-driving approach to online index selection that eschews the DBA and query optimiser, and instead learns the benefits of viable structures through strategic exploration and direct performance  ...  Learning approaches to optimisation and tuning. Recent years have witnessed new machine learning approaches to automate decision-making processes within databases.  ... 
arXiv:2010.09208v2 fatcat:567uvwlv45da3b5a2353bmnvby

Extensive Parameterization And Tuning of Architecture-Sensitive Optimizations

Qing Yi, Jichi Guo
2011 Procedia Computer Science  
We present a framework to support fine-grained parameterization of these optimizations and flexible tuning of their configuration space.  ...  We then use a transformation-aware (TA) search algorithm to support flexible tuning of the parameterized transformation scripts to achieve portable high performance.  ...  In particular, gemm is compute-bound as it reuses every data item a large number of times during evaluation; gemv is memory bound as only a small fraction of data are reused; ger is severely memory-bound  ... 
doi:10.1016/j.procs.2011.04.236 fatcat:bxovcxilibhl5n6ntdh4wsoqdy

Autonomous tuning and charge state detection of gate defined quantum dots [article]

J. Darulová, S.J. Pauka, N. Wiebe, K. W. Chan, G. C. Gardener, M. J. Manfra, M.C. Cassidy, M. Troyer
2019 arXiv   pre-print
With growing device complexity and increasing number of functional devices required for measurements, a manual approach to finding suitable gate voltages to confine electrons electrostatically is impractical  ...  Here, we implement a two-stage device characterization and dot-tuning process which first determines whether devices are functional and then attempts to tune the functional devices to the single or double  ...  Tuning approach Quantum dots are systems confining electrons or holes in regions small enough to make their quantum mechanical energy levels observable.  ... 
arXiv:1911.10709v2 fatcat:zifyol7vqvfejjljowa34thnmm

Fuzzy Controlled Architecture for Performance Tuning of Database Management System

S. F.Rodd, Umakant P. Kulkarni, A. R. Yardi
2012 International Journal of Computer Applications  
In this paper, a new tuning architecture based on fuzzy logic is presented, where in, the control action is expressed in linguistic terms.  ...  Database tuning is complicated due to the fact that several conflicting tuning parameters have to be adjusted simultaneously for a variety of workload types and the highly unpredictable traffic patterns  ...  Our thanks are also due to our esteemed Management for their support. Our Sincere thanks to Computer Center Head Prof. S.R.Mangalwede for providing us with the computing facilities.  ... 
doi:10.5120/4813-7050 fatcat:xesmapqvhzaydixrowmbljk76i

Automatic Database Management System Tuning Through Large-scale Machine Learning

Dana Van Aken, Andrew Pavlo, Geoffrey J. Gordon, Bohan Zhang
2017 Proceedings of the 2017 ACM International Conference on Management of Data - SIGMOD '17  
To overcome these challenges, we present an automated approach that leverages past experience and collects new information to tune DBMS configurations: we use a combination of supervised and unsupervised  ...  Database management system (DBMS) configuration tuning is an essential aspect of any data-intensive application effort.  ...  Similar to MySQL, Postgres has a small number of knobs that have a large impact on the performance.  ... 
doi:10.1145/3035918.3064029 dblp:conf/sigmod/AkenPGZ17 fatcat:skhqczxchnb5zlvxit6dmbbfem
« Previous Showing results 1 — 15 out of 11,248 results