281 Hits in 1.2 sec

SOL: Safe On-Node Learning in Cloud Platforms [article]

Yawen Wang, Daniel Crankshaw, Neeraja J. Yadwadkar, Daniel Berger, Christos Kozyrakis, Ricardo Bianchini
2022 arXiv   pre-print
Cloud platforms run many software agents on each server node. These agents manage all aspects of node operation, and in some cases frequently collect data and make decisions. Unfortunately, their behavior is typically based on pre-defined static heuristics or offline analysis; they do not leverage on-node machine learning (ML). In this paper, we first characterize the spectrum of node agents in Azure, and identify the classes of agents that are most likely to benefit from on-node ML. We then
more » ... pose SOL, an extensible framework for designing ML-based agents that are safe and robust to the range of failure conditions that occur in production. SOL provides a simple API to agent developers and manages the scheduling and running of the agent-specific functions they write. We illustrate the use of SOL by implementing three ML-based agents that manage CPU cores, node power, and memory placement. Our experiments show that (1) ML substantially improves our agents, and (2) SOL ensures that agents operate safely under a variety of failure conditions. We conclude that ML-based agents show significant potential and that SOL can help build them.
arXiv:2201.10477v1 fatcat:tnz34xuuu5fernsqyy3x4scgpe

GraphX: Unifying Data-Parallel and Graph-Parallel Analytics [article]

Reynold S. Xin, Daniel Crankshaw, Ankur Dave, Joseph E. Gonzalez, Michael J. Franklin, Ion Stoica
2014 arXiv   pre-print
From social networks to language modeling, the growing scale and importance of graph data has driven the development of numerous new graph-parallel systems (e.g., Pregel, GraphLab). By restricting the computation that can be expressed and introducing new techniques to partition and distribute the graph, these systems can efficiently execute iterative graph algorithms orders of magnitude faster than more general data-parallel systems. However, the same restrictions that enable the performance
more » ... ns also make it difficult to express many of the important stages in a typical graph-analytics pipeline: constructing the graph, modifying its structure, or expressing computation that spans multiple graphs. As a consequence, existing graph analytics pipelines compose graph-parallel and data-parallel systems using external storage systems, leading to extensive data movement and complicated programming model. To address these challenges we introduce GraphX, a distributed graph computation framework that unifies graph-parallel and data-parallel computation. GraphX provides a small, core set of graph-parallel operators expressive enough to implement the Pregel and PowerGraph abstractions, yet simple enough to be cast in relational algebra. GraphX uses a collection of query optimization techniques such as automatic join rewrites to efficiently implement these graph-parallel operators. We evaluate GraphX on real-world graphs and workloads and demonstrate that GraphX achieves comparable performance as specialized graph computation systems, while outperforming them in end-to-end graph pipelines. Moreover, GraphX achieves a balance between expressiveness, performance, and ease of use.
arXiv:1402.2394v1 fatcat:xxx2uvx6arbgdnjqiw7igqztnm

Composing Meta-Policies for Autonomous Driving Using Hierarchical Deep Reinforcement Learning [article]

Richard Liaw, Sanjay Krishnan, Animesh Garg, Daniel Crankshaw, Joseph E. Gonzalez, Ken Goldberg
2017 arXiv   pre-print
, 2 EECS, UC Berkeley, CA USA; 3 CS, Stanford University, Stanford CA US; The AUTOLAB at UC Berkeley (; {rliaw, sanjaykrishnan, animesh.garg, jegonzal, goldberg}, crankshaw  ... 
arXiv:1711.01503v1 fatcat:nk4nre7j65falaofacwiljr2ne

Clipper: A Low-Latency Online Prediction Serving System [article]

Daniel Crankshaw, Xin Wang, Giulio Zhou, Michael J. Franklin, Joseph E. Gonzalez, Ion Stoica
2017 arXiv   pre-print
Machine learning is being deployed in a growing number of applications which demand real-time, accurate, and robust predictions under heavy query load. However, most machine learning frameworks and systems only address model training and not deployment. In this paper, we introduce Clipper, a general-purpose low-latency prediction serving system. Interposing between end-user applications and a wide range of machine learning frameworks, Clipper introduces a modular architecture to simplify model
more » ... eployment across frameworks and applications. Furthermore, by introducing caching, batching, and adaptive model selection techniques, Clipper reduces prediction latency and improves prediction throughput, accuracy, and robustness without modifying the underlying machine learning frameworks. We evaluate Clipper on four common machine learning benchmark datasets and demonstrate its ability to meet the latency, accuracy, and throughput demands of online serving applications. Finally, we compare Clipper to the TensorFlow Serving system and demonstrate that we are able to achieve comparable throughput and latency while enabling model composition and online learning to improve accuracy and render more robust predictions.
arXiv:1612.03079v2 fatcat:fe2w5dhxsnazhd62gxcx4ielpu

InferLine: ML Prediction Pipeline Provisioning and Management for Tight Latency Objectives [article]

Daniel Crankshaw, Gur-Eyal Sela, Corey Zumar, Xiangxi Mo, Joseph E. Gonzalez, Ion Stoica, Alexey Tumanov
2020 arXiv   pre-print
Serving ML prediction pipelines spanning multiple models and hardware accelerators is a key challenge in production machine learning. Optimally configuring these pipelines to meet tight end-to-end latency goals is complicated by the interaction between model batch size, the choice of hardware accelerator, and variation in the query arrival process. In this paper we introduce InferLine, a system which provisions and manages the individual stages of prediction pipelines to meet end-to-end tail
more » ... ency constraints while minimizing cost. InferLine consists of a low-frequency combinatorial planner and a high-frequency auto-scaling tuner. The low-frequency planner leverages stage-wise profiling, discrete event simulation, and constrained combinatorial search to automatically select hardware type, replication, and batching parameters for each stage in the pipeline. The high-frequency tuner uses network calculus to auto-scale each stage to meet tail latency goals in response to changes in the query arrival process. We demonstrate that InferLine outperforms existing approaches by up to 7.6x in cost while achieving up to 34.5x lower latency SLO miss rate on realistic workloads and generalizes across state-of-the-art model serving frameworks.
arXiv:1812.01776v2 fatcat:kmnfgv5byffepdkxwhy5ektosi

402 'TAP it': trainee led initiative to reduce unnecessary blood testing in a paediatric emergency department

Eliza Magnusen, Kene Maduemem, Ayesha Shafiq, Daniel Crankshaw, Habab Mekki, Arani Sridhar
2020 Emergency Medicine Journal  
Aims/Objectives/BackgroundUnnecessary blood testing in the paediatric emergency department (PED) is a potential starting point for diagnostic dilemma, anxiety to families and increased healthcare costs. We hypothesized that a significant number of blood tests are performed instinctively rather than clinically indicated. This stimulated a quality improvement initiative to enlighten trainees on the utility of blood tests while aiming to enhance clinical decision making.Methods/DesignChildren
more » ... nting to the acute care team in a tertiary PED who had blood tests over a 2-week period in April 2019 were enrolled. Blood tests requested were interpreted in line with presenting features and clinical impression. Following implementation of changes (posters, QI champions, educational sessions), a repeat analysis was done over a 2-week period in October 2019.Results/ConclusionsOne hundred and one children in the first cycle were enrolled. Blood testing analysis revealed that 70%, 47%, and 32% had liver bloods, bone profile, and clotting testing done, respectively. Over half of these blood tests had no clear clinical indication. The yield of the tests performed without clinical indication was 0%. Case vignettes were attempted by trainees and ANPs to evaluate their attitudes to blood test requests. Number of requested blood tests not clinically indicated was lower than anticipated; probably explained by self-thought processes. Blood tests performed on 100 children in the second cycle demonstrated a significant reduction in tests done without clear clinical indication. Liver bloods, clotting screen, bone profile tests were performed on 40%, 24% and 27% respectively in the second cycle.The utility of blood tests may be an under recognised subject in paediatric training which needs addressing. This project typified positive impact of culture change via QI champions and educational sessions. Implementation of such changes is sustainable with an estimated savings of at least £500/month.
doi:10.1136/emj-2020-rcemabstracts.53 fatcat:meww2alulbbermc2htk24qxgly

Synthesis and properties of [2-[3,5-3H2]-tyrosine,4-glutamic acid]deamino-1-carba-oxytocin

Michal Lebl, Tomislav Barth, Denis J. Crankshaw, Bohuslav Černý, Edwin E. Daniel, A. K. Grover, Karel Jošt
1984 Collection of Czechoslovak Chemical Communications  
[Vol. 49J [19841 Lebl, Barth, Crankshaw, Cerny, Daniel, Grover, Jost tion of an aliquot.  ... 
doi:10.1135/cccc19841921 fatcat:7pxm3qikorezvna3rkpnyfwgiy

The Missing Piece in Complex Analytics: Low Latency, Scalable Model Management and Serving with Velox [article]

Daniel Crankshaw, Peter Bailis, Joseph E. Gonzalez, Haoyuan Li, Zhao Zhang, Michael J. Franklin, Ali Ghodsi, Michael I. Jordan
2014 arXiv   pre-print
To support complex data-intensive applications such as personalized recommendations, targeted advertising, and intelligent services, the data management community has focused heavily on the design of systems to support training complex models on large datasets. Unfortunately, the design of these systems largely ignores a critical component of the overall analytics process: the deployment and serving of models at scale. In this work, we present Velox, a new component of the Berkeley Data
more » ... s Stack. Velox is a data management system for facilitating the next steps in real-world, large-scale analytics pipelines: online model management, maintenance, and serving. Velox provides end-user applications and services with a low-latency, intuitive interface to models, transforming the raw statistical models currently trained using existing offline large-scale compute frameworks into full-blown, end-to-end data products capable of recommending products, targeting advertisements, and personalizing web content. To provide up-to-date results for these complex models, Velox also facilitates lightweight online model maintenance and selection (i.e., dynamic weighting). In this paper, we describe the challenges and architectural considerations required to achieve this functionality, including the abilities to span online and offline systems, to adaptively adjust model materialization strategies, and to exploit inherent statistical properties such as model error tolerance, all while operating at "Big Data" scale.
arXiv:1409.3809v2 fatcat:33a5muyjlbhmrjxn2zkwxlxpxq

Inhalant use initiation among U.S. adolescents: Evidence from the National Survey of Parents and Youth using discrete-time survival analysis

James M. Nonnemaker, Erik C. Crankshaw, Daniel R. Shive, Altijani H. Hussin, Matthew C. Farrelly
2011 Addictive Behaviours  
The purpose of this paper is to identify factors associated with initiation to inhalant use among adolescents ages 9 to 18. The data are from the National Survey of Parents and Youth, a longitudinal household survey. Baseline surveys for adolescents and parents were conducted between November 1999 and June 2001 and then annually for three subsequent rounds. The outcome measure is an indicator of a respondent's first use of inhalants. Discrete-time survival analysis was used to model the hazard
more » ... f initiation. The hazard of inhalant initiation peaks at about 14 years of age (slightly younger than smoking and marijuana initiation). African Americans were less likely than Whites to initiate inhalant use, and higher family income was protective against inhalant initiation. The findings suggest that parenting is associated with initiation of inhalant use: parental drug use was a risk factor for inhalant initiation, and a measure of parental monitoring was protective. The study results also suggest a strong relationship between inhalant use and other problem behaviors and sensation seeking. These results highlight the need to intervene early for youth at risk of or just beginning to engage in risky behaviors including inhalant use.
doi:10.1016/j.addbeh.2011.03.009 pmid:21481544 pmcid:PMC3104053 fatcat:dtzo5bvyp5eytki3l64w376nyy

IDK Cascades: Fast Deep Learning by Learning not to Overthink [article]

Xin Wang, Yujia Luo, Daniel Crankshaw, Alexey Tumanov, Fisher Yu, Joseph E. Gonzalez
2018 arXiv   pre-print
Advances in deep learning have led to substantial increases in prediction accuracy but have been accompanied by increases in the cost of rendering predictions. We conjecture that fora majority of real-world inputs, the recent advances in deep learning have created models that effectively "overthink" on simple inputs. In this paper, we revisit the classic question of building model cascades that primarily leverage class asymmetry to reduce cost. We introduce the "I Don't Know"(IDK) prediction
more » ... cades framework, a general framework to systematically compose a set of pre-trained models to accelerate inference without a loss in prediction accuracy. We propose two search based methods for constructing cascades as well as a new cost-aware objective within this framework. The proposed IDK cascade framework can be easily adopted in the existing model serving systems without additional model re-training. We evaluate the proposed techniques on a range of benchmarks to demonstrate the effectiveness of the proposed framework.
arXiv:1706.00885v4 fatcat:d34s54tttred7ps4wtwhctgsx4

Page 4 of None Vol. 11, Issue 3 [page]

1982 None  
DANIEL. Canine tra- chealis membrane fractionation and characterization. Cell Calcium. 1: 135-146, 1980. . GROVER, A. K., C. Y. Kwan, J. CRANKSHAW, D. J. CRANKSHAW, R. E. GARFIELD; AND E. E. DANIEL.  ...  CRANKSHAW, AND E. E. DANIEL. Effect of digitonin on rat myometrium subcellular membrane frac- tions. Can. J. Physiol. Pharmacol. 59: 1128-1133, 1981. . GROVER, A. K., C. Y. KWAN, AND E. E. DANIEL.  ... 

Page 911 of Cellular and Molecular Life Sciences Vol. 41, Issue 7 [page]

1985 Cellular and Molecular Life Sciences  
Grover, A.K., Crankshaw, J., Garfield, R.E., and Daniel, E.E., Smooth muscle membrane orientation. Can. J. Physiol. Pharmac. 58 (1980) 1202-1211. Grover, A.K., Crankshaw, J., Triggle, C.  ...  Lee, R.M.K.W., and Daniel, E. E., Bovine aorta membrane fractionation and characterization. Fed. Proc. (abstr.) 40 (1981) 551. Grover, A. K., Kwan, C. Y., Crankshaw, J., and Daniel, E.  ... 

Page 19 of None Vol. 8, Issue 3 [page]

1980 None  
CRANKsHAW, D. J., L. A. BRANDA, M. A. MarTLis, AND E. E. DANIEL. Localization of the oxytocin receptor in the plasma mem- brane of rat myometrium. Eur. J. Biochem. 86: 481-486, 1978. . Csapo, A. I.  ...  Crankshaw, and E. E. Daniel for valuable discussion. This study was supported by grants from the Medical Research Council (Canada). R. E. Garfield is a Scholar of the Canadian Heart Foundation.  ... 

Page 5 of None Vol. 19, Issue 1 [page]

1986 None  
CRANKSHAW, D. J. CRANKSHAW, R. E. GARFIELD, AND E. E. DANIEL. Characteristics of calcium transport and binding to rat myometrium plasma membrane subfractions. Am. J.  ...  CRANKSHAW, R. E. GARFIELD, D. J. CRANK- SHAW, C. Y. KWAN, L. A. BRANDA, AND E. E. DANIEL. Character- ization of membrane fractions and isolation of purified plasma membranes from rat myometrium. J.  ... 

Page 130 of Metal Progress Vol. 65, Issue 5 [page]

1954 Metal Progress  
John Hamilton Crankshaw @ has been promoted to vice-president in charge of engineering at the J. A. Zurn Mfg.  ...  The author of numerous papers and articles, Pro- fessor Daniels is a member of the A. I. M.  ... 
« Previous Showing results 1 — 15 out of 281 results