Google Correlate is an online tool that automatically selects web search queries that are highly correlated with a target temporal or spatial pattern of interest. It uses an approximate nearest neighbor algorithm to quickly identify queries from millions of candidates that best match the pattern. The tool was tested by using it to select queries correlated with influenza activity in the US, and building a model with the selected queries that achieved comparable accuracy to Google Flu Trends' original model, but much faster. Google Correlate was also used to select queries correlated with home refinancing rates in the US, identifying many refinancing-related queries as well as mortgage rate queries.
Moz holy grail of e commerce conversion optimizationBitsytask
This document provides a 91-point checklist for optimizing e-commerce conversion rates. It discusses the importance of optimizing the user experience across all touchpoints on a store's website. This includes understanding customers, optimizing pages like home, navigation, categories, product pages, checkout, and others. It provides tips like showing top products prominently, offering multiple ordering options, using videos to showcase products, localizing the store for different countries, improving the category structure and search, and writing clear unique selling points. The document emphasizes the need to understand customer behavior through surveys, chat interactions, and other tools to inform conversion optimization efforts.
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
This document contains techniques for manipulating online forums and discussions. It describes 6 techniques: 1) Forum Sliding, which involves posting unrelated topics to move critical discussions down the page. 2) Consensus Cracking, which aims to slowly shift public opinion by introducing weak and strong arguments from fake accounts. 3) Topic Dilution, which floods forums with unrelated topics to distract and overwhelm readers. 4) Information Collection, which involves posing questions to gather intelligence on forum members. 5) Anger Trolling, which deliberately posts inflammatory content to identify and track the most aggressive members. 6) Gaining Full Control, which involves obtaining moderator privileges to censor discussions and potentially shut down the forum. The document warns that these techniques only work if
This document summarizes a scientific article that analyzes the performance of Google Flu Trends (GFT), a system created by Google to track influenza levels based on search query data, compared to tracking by the Centers for Disease Control and Prevention (CDC). The summary identifies two key issues that contributed to errors in GFT's estimates: (1) "big data hubris," or overreliance on large datasets without ensuring validity, and (2) "algorithm dynamics," wherein changes to Google's search algorithm affected the data GFT was using. The article concludes by offering lessons on transparency, granularity of data analysis, and combining multiple data sources.
Impact evaluations aim to predict the future, but they are rooted in particular contexts and to what extent they generalize is an open and important question. I founded an organization to systematically collect and synthesize impact evaluation results on a wide variety of interventions in development. These data allow me to answer this and other questions for the rst time using a large data set of studies. I consider several measures of generalizability, discuss the strengths and limitations of each metric, and provide benchmarks based on the data. I use the example of the eect of conditional cash transfers on enrollment rates to show how some of the heterogeneity can be modelled and the eect this can have on the generalizability measures. The predictive power of the model improves over time as more studies are completed. Finally, I show how researchers can
estimate the generalizability of their own study using their own data, even when data from no comparable studies exist.
Read more at: www.hhs.se/site
This document summarizes literature on response rates for Internet surveys. It discusses studies that used single response modes of web or email, as well as dual response modes that allowed respondents to choose web or mail. Studies using preselected samples or probability samples tended to have higher response rates than those using convenience samples. When both web and mail options were provided, the majority of respondents typically chose mail over web. Response rates also tended to be higher when respondents were carefully screened and the survey was designed to be user-friendly.
Running head EVALUATION OFNEIGHBORHOOD WATCH PROGRAM 1 .docxtodd271
Running head: EVALUATION OFNEIGHBORHOOD WATCH PROGRAM 1
EVALUATION OF NEIGHBORHOOD WATCH PROGRAM
Abstract
There has been little work on the evaluation of neighborhood watch programs in the past. However,
scholars and policy makers have increasingly shown interest in identifying whether neighborhood
watch programs are effective. The studies conducted in the past have limitations which make them
unable to precisely measure the effectiveness of neighborhood watch programs. In response to
EVALUATION OF NEIGHBORHOOD WATCH PROGRAM 2
that, this paper proposes an elaborate evaluation technique which can assist in accurately
determining the effectiveness of the existing neighborhood watch programs. The paper does not
ignore the previous studies but instead explores them with the aim of offering theoretical
background of the programs and also explores the proposed techniques with the objective of
presenting a reliable and appropriate method of assessment.
Introduction
Neighborhood watch is also called neighborhood crime watch or crime watch. It refers to the
organization of civilians into a group which is devoted to crime or vandalism prevention within
their respective neighborhood. Neighborhood watch programs aim at achieving a wide range of
objectives related to the security of their neighborhood. Some of the objectives of this program
involve the educating community on security and safety of the neighborhood and establishing
means of maintaining a safety neighborhood. This program usually cooperate with law
enforcement agencies in such a way that if the neighborhood watch group identifies a criminal
they are expected to inform the authorities and not to engage. It is this attribute of the group that
differentiates it from vigilante groups.
In the United States there are numerous neighborhood watch groups and Rosenbaum (1988),
estimated that 20% of American residents were involved in the neighborhood watch activities by
1988. The organization and agenda of these groups vary but they all maintain that their main
objectives are related to the promotion of secure neighborhood and crime reduction. Current
systems of Neighborhood watch programs trace their origin back in 1960. According to
Rasenberger (2006), the shocking events of the murder and rape of Kitty Genovese compelled the
people to organize themselves into crime prevention groups. Ever since the Neighborhood Watch
EVALUATION OF NEIGHBORHOOD WATCH PROGRAM 3
programs have developed nationwide with the assistance of the National Sheriff's Association.
Various scholars have presented various analyses of these programs with the aim of identifying
whether the programs are effective. These evaluations have applied various approaches with some
having limitations which negatively influenced the results obtained. This paper seeks to present a
reliable method of evaluating the programs with the aim of shaping and i.
Forecasting Agri-food Consumption Using Web Search Engine IndicesQUESTJOURNAL
ABSTRACT : In this study, we examine Naver Trend of South Korea (similar to Google Trends), which is a real-time weekly index of the volume of queries that users enter into Naver. We aim to find that these search engine data improves accuracy in forecasting consumption of agri-foods. For a more detail explanation, we classify empirically agri-food items into several specific groups.The results are different by agri-food groups. The agri-food items often used for main or single dishes tend to show significant improvement for both insample estimation and out-of-sample forecast. However, agri-food items used mainly as minor ingredients are not significant. Even if the item is a main or single dish, agri-food items frequently consumed out of the home have no significant relationship. However, agri-food items mainly consumed at home are significant. Meanwhile, the significant difference between common products and brand products are not found. Rather, the more important criteria are macro trends. Agri-food items with gradual growth macro trends or long-term fluctuation have a significant relationship. On the contrary, the steady selling of ramen products or dumpling products is not significant.
This document discusses forecasting household consumption in the Czech Republic using data from Google Trends. It first reviews literature on using sentiment indicators and Google Trends data to predict consumption. It then describes the consumption and sentiment data from the Czech Statistical Office, as well as search data from Google Trends. Finally, it introduces the model that will be used to forecast consumption using these different data sources.
NOVEL MACHINE LEARNING ALGORITHMS FOR CENTRALITY AND CLIQUES DETECTION IN YOU...gerogepatton
The goal of this research project is to analyze the dynamics of social networks using machine learning techniques to locate maximal cliques and to find clusters for the purpose of identifying a target demographic. Unsupervised machine learning techniques are designed and implemented in this project to analyze a dataset from YouTube to discover communities in the social network and find central nodes. Different clustering algorithms are implemented and applied to the YouTube dataset. The well-known Bron-Kerbosch algorithm is used effectively in this research to find maximal cliques. The results obtained from this research could be used for advertising purposes and for building smart recommendation systems. All algorithms were implemented using Python programming language. The experimental results show that we were able to successfully find central nodes through clique-centrality and degree centrality. By utilizing clique detection algorithms, the research shown how machine learning algorithms can detect close knit groups within a larger network.
Novel Machine Learning Algorithms for Centrality and Cliques Detection in You...gerogepatton
This document summarizes a research paper that used machine learning algorithms to analyze social networks on YouTube. The researchers used unsupervised learning techniques like clustering and centrality measures to identify communities and influential users. Specifically, they used Louvain modularity and spectral clustering to detect groups for advertising purposes. Degree centrality and clique centrality were calculated to find central nodes that could be identified as influencers for product sponsorship. The experiments showed the algorithms could successfully find tightly-knit groups and key users within the larger YouTube network.
NOVEL MACHINE LEARNING ALGORITHMS FOR CENTRALITY AND CLIQUES DETECTION IN YOU...ijaia
This document summarizes a research paper that used machine learning algorithms to analyze social networks on YouTube. The researchers used unsupervised learning techniques like clustering and centrality measures to identify communities and influential users. Specifically, they used Louvain modularity and spectral clustering to detect groups for advertising purposes. Degree centrality and clique centrality were calculated to find central nodes that could be targeted for sponsorship deals. The experiments showed the algorithms could successfully find tightly-knit groups and key influencers within the larger YouTube network.
Measuring effectiveness of machine learning systemsAmit Sharma
Many online systems, such as recommender systems or ad systems, are increasingly being used in societally critical domains such as education, healthcare, finance and governance. A natural question to ask is about their effectiveness, which is often measured using observational metrics. However, these metrics hide cause-and-effect processes between these systems, people's behavior and outcomes. I will present a causal framework that allows us to tackle questions about the effects of algorithmic systems and demonstrate its usage through evaluation of Amazon's recommender system and a major search engine. I will also discuss how such evaluations can lead to metrics for designing better systems.
Forecasting covid 19 by states with mobility data Yasas Senarath
COVID-19 is an ongoing pandemic (2020). We provide a state level analysis of COVID-19 spread in USA and also integrating it with the human mobility data. We model the relationship with Human Mobility Data where mobility explains about the difference in the behaviors.
ANALYZING THE EFFECTS OF DIFFERENT POLICIES AND STRICTNESS LEVELS ON MONTHLY ...IJDKP
The corona virus is one of the most unprecedented events of recent decades. Countries struggled to identify
appropriate COVID-19 policies to prevent virus spread effectively. Although much research has been
done, little focused on policy effectiveness and their enforcement levels. As corona virus cases and death
numbers fluctuated among countries, questions of which policies are most effective in preventing corona
virus spread and how strict they should be implemented have yet to be answered. Countries are prone to
making policy and implementation errors that could cost lives. This research identified the most effective
policies and their most effective enforcement levels through data analysis of 12 common coronavirus
policies. A monthly case increase rate prediction model was developed to enable decision makers to
evaluate the effectiveness of COVID-19 policies and their enforcement levels so that they can implement
policies efficiently to save lives, time, and money.
Running head LOGIC MODELLOGIC MODEL 2Logic modelStu.docxwlynn1
Running head: LOGIC MODEL
LOGIC MODEL
2
Logic model
Student’s name
University affiliation
Date
References
Blue-Howells, J., McGuire, J., & Nakashima, J. (2008). Co-location of health care services for homeless veterans: a case study of innovation in program implementation. Social work in health care, 47(3), 219-231.
Output
Integrating patient care
Communication and collaboration between workers hence resulting to communities of practicing clinicians
Attracting new patients to GLA
Funding a two-year pilot grant
Effective process for psychiatric screening for homeless patients
Outcomes
Homeless project were integrated
The issues of homeless veterans were addressed due to institutional barriers
There was creation of coalition and linking the project to legitimate VA-wide goals
Good sustained program maintenance, process evaluation and encouraging development of communities.
Activities
Building a coalition of decision makers
Introduction of a new integrated program
Inputs
The decision to implement
Initial implementation
Sustained maintenance
Termination or transformation
Running head: PROGRAM EVALUATION 1
PROGRAM EVALUATION 2
Program Evaluation
Institutional Affiliation
Insert the student’s name
Instructor’s name
Course
Date
Introduction
Evaluation of the program is usually done to in order to determine the quality of the program, how effective the program is and how the program is performing. This can help to know if the program is making a significant difference among the targeted people. It can also assist to know if the program is functioning or not. This paper therefore seeks to evaluate the program which is assisting the homeless people within the community.
The two program evaluation questions are: what is the reach of the program? And what has been the impact of the program on the homeless people? The answers to these questions would elicit both qualitative and quantitative results. Therefore, the program evaluation will require both quantitative and qualitative data collection plan. This is because the use of mixed-method approach is convenient since the results and findings would be reliable (Creswell, 2017). After identifying the evaluation program questions, the next step will be to come up with plan of evaluating a program. The plan should consist of methods of collecting data, evidences, the person responsible and the duration.
Program Evaluation Question
Evidence
Methods and sources of collecting data
Person in charge
Duration
1. What is the reach of the program?
Number of building materials distributed
Records of the program
Robert
One month
2. What has been the impact of the program on the homeless people?
Number of people resettled
Number of people not yet re.
Using Mobile Data and Airtime Credit Purchases to Estimate Food Security - Pr...UN Global Pulse
This study assessed the potential use of mobile phone data as a proxy for food security and poverty indicators in an East African country. Mobile phone data extracted from airtime credit purchases and activity was compared to a nationwide household survey conducted by the UN World Food Programme at the same time period. Results showed high correlations between airtime credit purchases and survey responses about consumption of several food items commonly purchased in markets. Models based on anonymised mobile phone data were also able to accurately estimate multidimensional poverty indicators. This preliminary research suggested proxies derived from mobile phone data could provide valuable real-time information to fill data gaps between surveys and where timely data is not accessible.
Detecting influenza epidemics using search engine query databenj_2
This document describes a method for detecting influenza epidemics using search engine query data from Google. The researchers:
1) Analyzed search query data from 2003-2008 to identify queries related to influenza-like illness (ILI) without prior knowledge.
2) Developed models correlating the frequency of ILI-related search queries to CDC data on ILI physician visits. The top 45 queries had the best fit.
3) Found search query data could accurately estimate weekly ILI activity in US regions with a 1 day reporting lag, faster than traditional surveillance methods. This approach may enable global influenza surveillance where search engines are widely used.
Accessing The Quality Of Online Classified Websites An Empirical Study Of Th...Jim Webb
This study analyzes the quality of online classified websites for rental properties from the 100 largest US newspapers. The researchers evaluated the websites based on criteria for intrinsic data quality, contextual data quality, accessibility, and representation. The results showed that larger newspapers generally had higher quality websites than smaller newspapers, but that overall improvements were needed across all sites.
An Update of Lot Quality Assurance Sampling (LQAS) Technologies Handout 1CORE Group
This document compares two survey methods - Lot Quality Assurance Sampling (LQAS) and Demographic Health Surveys (DHS) - for measuring health indicators in Uganda. It analyzes data from 24 matched indicators collected independently by LQAS (n=8876) and DHS (n=1200) in southwest Uganda in 2011. On average, the difference between LQAS and DHS estimates was 0.062, with 75.7% agreement. While both methods provide regional estimates, only LQAS can provide data at more granular district levels needed for local health management. LQAS also allows for more frequent, lower cost surveys.
B A S I C L O G I C M O D E L D E V E L O P M E N T Pr.docxcelenarouzie
B A S I C L O G I C M O D E L D E V E L O P M E N T
Produced by The W. K. Kellogg Foundation
53535353
Developing a Basic Logic
Model For Your Program
Drawing a picture of how your program will achieve results
hether you are a grantseeker developing a proposal for start-up funds or a
grantee with a program already in operation, developing a logic model can
strengthen your program. Logic models help identify the factors that will
impact your program and enable you to anticipate the data and resources
you will need to achieve success. As you engage in the process of creating your
program logic model, your organization will systematically address these important
program planning and evaluation issues:
• Cataloguing of the resources and actions you believe you will need to reach intended
results.
• Documentation of connections among your available resources, planned activities and
the results you expect to achieve.
• Description of the results you are aiming for in terms of specific, measurable, action-
oriented, realistic and timed outcomes.
The exercises in this chapter gather the raw material you need to draw a basic logic
model that illustrates how and why your program will work and what it will accomplish.
You can benefit from creating a logic model at any point in the life of any program.
The logic model development process helps people inside and outside your
organization understand and improve the purpose and process of your work.
Chapter 2 is organized into two sections—Program Implementation, and Program
Results. The best recipe for program success is to complete both exercises. (Full-size
masters of each exercise and the checklists are provided in the Forms Appendix at the
back of the guide for you to photocopy and use with stakeholder groups as you design
your program.)
Exercise 1: Program Results. In a series of three steps, you describe the results you
plan to achieve with your program.
Exercise 2: Program Resources and Activities by taking you through three steps
that connect the program’s resources to the actual activities you plan to do.
Chapter
2
W
B A S I C L O G I C M O D E L D E V E L O P M E N T
Produced by The W. K. Kellogg Foundation
54545454
The Mytown Example
Throughout Exercises 1 and 2 we’ll follow an example program to see how the logic
model steps can be applied. In our example, the folks in Mytown, USA are striving to
meet the needs of growing numbers of uninsured residents who are turning to Memorial
Hospital’s Emergency Room for care. Because that care is expensive and not the best
way to offer care, the community is working to create a free clinic. Throughout the
chapters, Mytown’s program information will be dropped into logic model templates for
Program Planning, Implementation, and Evaluation.
Novice Logic modelers may want to have copies of the Basic Logic Model Template in
front of them and follow along. Those read.
Lehman Brothers ALT-A Mortgage Docs, December 18, 2006Bitsytask
This document provides guidelines for several of Aurora's Alt-A loan programs, including maximum loan amounts, LTV/CLTV ratios, eligible property types, borrower documentation types, minimum credit scores, and other requirements. It includes matrices listing the specific guidelines for Alt-A loans, High LTV Alt-A loans, Choice Advantage loans, and No Documentation loans. Additional details are provided around eligible transaction types, interest rate types, prepayment penalties, debt-to-income ratios, mortgage insurance requirements, and other underlying program rules.
This document contains an interest rate sheet for various mortgage loan programs, with rates depending on the borrower's credit score, loan-to-value ratio, and other factors. It lists adjustable interest rates for 2-year and 3-year fixed rate loans, with rates ranging from around 6% to over 12% depending on the borrower's credit profile. Various adjustments are also shown that can increase or decrease the rate depending on the loan amount, property type, income documentation type, and other loan details.
This document provides guidelines for various loan programs, including guidelines for loan-to-value (LTV) ratios, combined loan-to-value (CLTV) ratios, minimum credit scores, maximum loan amounts, debt-to-income ratios, required documentation, reserves requirements, and other underwriting criteria. It includes matrices that specify the requirements for different loan programs depending on the borrower's credit score, including programs for first and second trust deeds. The document also provides guidance on appraisal requirements, credit history and derogatory credit guidelines, bankruptcy and foreclosure requirements, tax and judgment liens, gift and closing cost policies, and state-specific restrictions.
New Century Subprime Mortgage Matrix (Stated Doc / 80%, 550 FICO, 50% DTI) 7-...Bitsytask
This document is a rate sheet from New Century Mortgage Corporation that provides mortgage rates for wholesale loans. It lists rates based on the borrower's credit score, loan-to-value ratio, occupancy type, income documentation, and other loan characteristics. Adjustments to the rates are shown for factors such as loan size, property type, prepayment penalties, and programs. Minimum rates are provided for different loan programs.
Countrywide Option Arm Loans (Negative Amortization) July 26 2006Bitsytask
Countrywide Option Arm loans (Negative Amortization) July 26 2006.
These loans could qualify to 90% loan-to-value and allowed a teaser rate of as low as 1% payment, which would negatively amortize up to 120% of the value of the property. The borrower would make the 1% payment, and go further into debt. Countrywide likely sold these loans as derivatives before having to deal with the inevitable foreclosure.
This document is a client guide for GMAC-RFC that was last updated on September 11, 2006. It outlines GMAC-RFC's objectives, the contractual obligations of clients, and GMAC-RFC's responsibilities as servicer. It also describes client eligibility standards, loan eligibility requirements, property eligibility standards, required loan documentation, and insurance requirements. The guide is organized into multiple chapters covering these various topics.
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
Operation Ajax Declassified PDF Appendix EBitsytask
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
Operation Ajax Declassified PDF Appendix DBitsytask
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
Operation Ajax Declassified PDF Appendix BBitsytask
CIA Declassified Operation Ajax Document (PDF File). This document details the CIA/MI5'a overthrow of Iran's president, Mohammad Mosaddegh, on behalf of British Petroleum (BP).
Basic 'how-to' guide for basic and advanced page speed optimization. Page speed optimization resides in the conversion rate optimization family and is one of the most overlooked items in the web development spectrum. For over a decade, countless organizations have quantified that faster websites result in more conversions, leads & revenue. %3Cb>rawr</b><
Subprime Underwriting Matrix, 100% LTV down to 580 FICOBitsytask
This document provides loan limit and LTV guidelines for different loan programs, including:
- Maximum loan amounts and LTVs for full, fast and lite income documentation for owner-occupied properties depend on credit grade and credit score. Loan limits range from $400,000 to $1,000,000.
- Guidelines for stated income documentation loans include lower loan amounts and LTVs than full documentation loans, with loan limits ranging from $250,000 to $650,000 depending on credit factors.
- Non-owner occupied loans have lower amounts and LTVs compared to owner-occupied, with loan limits between $300,000 and $500,000 based on credit grade and history.
Mitigating the Impact of State Management in Cloud Stream Processing SystemsScyllaDB
Stream processing is a crucial component of modern data infrastructure, but constructing an efficient and scalable stream processing system can be challenging. Decoupling compute and storage architecture has emerged as an effective solution to these challenges, but it can introduce high latency issues, especially when dealing with complex continuous queries that necessitate managing extra-large internal states.
In this talk, we focus on addressing the high latency issues associated with S3 storage in stream processing systems that employ a decoupled compute and storage architecture. We delve into the root causes of latency in this context and explore various techniques to minimize the impact of S3 latency on stream processing performance. Our proposed approach is to implement a tiered storage mechanism that leverages a blend of high-performance and low-cost storage tiers to reduce data movement between the compute and storage layers while maintaining efficient processing.
Throughout the talk, we will present experimental results that demonstrate the effectiveness of our approach in mitigating the impact of S3 latency on stream processing. By the end of the talk, attendees will have gained insights into how to optimize their stream processing systems for reduced latency and improved cost-efficiency.
Quantum Communications Q&A with Gemini LLM. These are based on Shannon's Noisy channel Theorem and offers how the classical theory applies to the quantum world.
Best Practices for Effectively Running dbt in Airflow.pdfTatiana Al-Chueyr
As a popular open-source library for analytics engineering, dbt is often used in combination with Airflow. Orchestrating and executing dbt models as DAGs ensures an additional layer of control over tasks, observability, and provides a reliable, scalable environment to run dbt models.
This webinar will cover a step-by-step guide to Cosmos, an open source package from Astronomer that helps you easily run your dbt Core projects as Airflow DAGs and Task Groups, all with just a few lines of code. We’ll walk through:
- Standard ways of running dbt (and when to utilize other methods)
- How Cosmos can be used to run and visualize your dbt projects in Airflow
- Common challenges and how to address them, including performance, dependency conflicts, and more
- How running dbt projects in Airflow helps with cost optimization
Webinar given on 9 July 2024
論文紹介:A Systematic Survey of Prompt Engineering on Vision-Language Foundation ...Toru Tamaki
Jindong Gu, Zhen Han, Shuo Chen, Ahmad Beirami, Bailan He, Gengyuan Zhang, Ruotong Liao, Yao Qin, Volker Tresp, Philip Torr "A Systematic Survey of Prompt Engineering on Vision-Language Foundation Models" arXiv2023
https://arxiv.org/abs/2307.12980
BT & Neo4j: Knowledge Graphs for Critical Enterprise Systems.pptx.pdfNeo4j
Presented at Gartner Data & Analytics, London Maty 2024. BT Group has used the Neo4j Graph Database to enable impressive digital transformation programs over the last 6 years. By re-imagining their operational support systems to adopt self-serve and data lead principles they have substantially reduced the number of applications and complexity of their operations. The result has been a substantial reduction in risk and costs while improving time to value, innovation, and process automation. Join this session to hear their story, the lessons they learned along the way and how their future innovation plans include the exploration of uses of EKG + Generative AI.
An invited talk given by Mark Billinghurst on Research Directions for Cross Reality Interfaces. This was given on July 2nd 2024 as part of the 2024 Summer School on Cross Reality in Hagenberg, Austria (July 1st - 7th)
Advanced Techniques for Cyber Security Analysis and Anomaly DetectionBert Blevins
Cybersecurity is a major concern in today's connected digital world. Threats to organizations are constantly evolving and have the potential to compromise sensitive information, disrupt operations, and lead to significant financial losses. Traditional cybersecurity techniques often fall short against modern attackers. Therefore, advanced techniques for cyber security analysis and anomaly detection are essential for protecting digital assets. This blog explores these cutting-edge methods, providing a comprehensive overview of their application and importance.
How RPA Help in the Transportation and Logistics Industry.pptxSynapseIndia
Revolutionize your transportation processes with our cutting-edge RPA software. Automate repetitive tasks, reduce costs, and enhance efficiency in the logistics sector with our advanced solutions.
Implementations of Fused Deposition Modeling in real worldEmerging Tech
The presentation showcases the diverse real-world applications of Fused Deposition Modeling (FDM) across multiple industries:
1. **Manufacturing**: FDM is utilized in manufacturing for rapid prototyping, creating custom tools and fixtures, and producing functional end-use parts. Companies leverage its cost-effectiveness and flexibility to streamline production processes.
2. **Medical**: In the medical field, FDM is used to create patient-specific anatomical models, surgical guides, and prosthetics. Its ability to produce precise and biocompatible parts supports advancements in personalized healthcare solutions.
3. **Education**: FDM plays a crucial role in education by enabling students to learn about design and engineering through hands-on 3D printing projects. It promotes innovation and practical skill development in STEM disciplines.
4. **Science**: Researchers use FDM to prototype equipment for scientific experiments, build custom laboratory tools, and create models for visualization and testing purposes. It facilitates rapid iteration and customization in scientific endeavors.
5. **Automotive**: Automotive manufacturers employ FDM for prototyping vehicle components, tooling for assembly lines, and customized parts. It speeds up the design validation process and enhances efficiency in automotive engineering.
6. **Consumer Electronics**: FDM is utilized in consumer electronics for designing and prototyping product enclosures, casings, and internal components. It enables rapid iteration and customization to meet evolving consumer demands.
7. **Robotics**: Robotics engineers leverage FDM to prototype robot parts, create lightweight and durable components, and customize robot designs for specific applications. It supports innovation and optimization in robotic systems.
8. **Aerospace**: In aerospace, FDM is used to manufacture lightweight parts, complex geometries, and prototypes of aircraft components. It contributes to cost reduction, faster production cycles, and weight savings in aerospace engineering.
9. **Architecture**: Architects utilize FDM for creating detailed architectural models, prototypes of building components, and intricate designs. It aids in visualizing concepts, testing structural integrity, and communicating design ideas effectively.
Each industry example demonstrates how FDM enhances innovation, accelerates product development, and addresses specific challenges through advanced manufacturing capabilities.
Paradigm Shifts in User Modeling: A Journey from Historical Foundations to Em...Erasmo Purificato
Slide of the tutorial entitled "Paradigm Shifts in User Modeling: A Journey from Historical Foundations to Emerging Trends" held at UMAP'24: 32nd ACM Conference on User Modeling, Adaptation and Personalization (July 1, 2024 | Cagliari, Italy)
YOUR RELIABLE WEB DESIGN & DEVELOPMENT TEAM — FOR LASTING SUCCESS
WPRiders is a web development company specialized in WordPress and WooCommerce websites and plugins for customers around the world. The company is headquartered in Bucharest, Romania, but our team members are located all over the world. Our customers are primarily from the US and Western Europe, but we have clients from Australia, Canada and other areas as well.
Some facts about WPRiders and why we are one of the best firms around:
More than 700 five-star reviews! You can check them here.
1500 WordPress projects delivered.
We respond 80% faster than other firms! Data provided by Freshdesk.
We’ve been in business since 2015.
We are located in 7 countries and have 22 team members.
With so many projects delivered, our team knows what works and what doesn’t when it comes to WordPress and WooCommerce.
Our team members are:
- highly experienced developers (employees & contractors with 5 -10+ years of experience),
- great designers with an eye for UX/UI with 10+ years of experience
- project managers with development background who speak both tech and non-tech
- QA specialists
- Conversion Rate Optimisation - CRO experts
They are all working together to provide you with the best possible service. We are passionate about WordPress, and we love creating custom solutions that help our clients achieve their goals.
At WPRiders, we are committed to building long-term relationships with our clients. We believe in accountability, in doing the right thing, as well as in transparency and open communication. You can read more about WPRiders on the About us page.
Understanding Insider Security Threats: Types, Examples, Effects, and Mitigat...Bert Blevins
Today’s digitally connected world presents a wide range of security challenges for enterprises. Insider security threats are particularly noteworthy because they have the potential to cause significant harm. Unlike external threats, insider risks originate from within the company, making them more subtle and challenging to identify. This blog aims to provide a comprehensive understanding of insider security threats, including their types, examples, effects, and mitigation techniques.
Choose our Linux Web Hosting for a seamless and successful online presencerajancomputerfbd
Our Linux Web Hosting plans offer unbeatable performance, security, and scalability, ensuring your website runs smoothly and efficiently.
Visit- https://onliveserver.com/linux-web-hosting/
Cookies program to display the information though cookie creation
Google trends correlate
1. Google Correlate Whitepaper 1
Google Correlate Whitepaper
Matt Mohebbi, Dan Vanderkam, Julia Kodysh,
Rob Schonberger, Hyunyoung Choi & Sanjiv Kumar
Draft Date: June 9, 2011
Trends in online web search query data have been shown useful in providing
models of real world phenomena. However, many of these results rely on the
careful choice of queries that prior knowledge suggests should correspond
with the phenomenon. Here, we present an online, automated method for
query selection that does not require such prior knowledge. Instead, given
a temporal or spatial pattern of interest, we determine which queries best
mimic the data. These search queries can then serve to build an estimate of
the true value of the phenomenon. We present the application of this method
to produce accurate models of influenza activity and home refinance rate
in the United States. We additionally show that spatial patterns in real world
activity and temporal patterns in web search query activity can both surface
interesting and useful correlations.
2. Google Correlate Whitepaper 2
Background
Web search activity has previously been shown useful for
providing estimates of real-world activity in a variety of
contexts, with the most common being health and economics.
Examples in health include influenza1,2,3,4,5,6,9
, acute diarrhea6
,
chickenpox6
, listeria7
, and salmonella8
. Examples in economics
include movie box office sales9
, computer game sales9
, music
billboard ranking9
, general retail sales10
, automotive sales10
,
home sales10
, travel10
, investor attention11
, and initial claims for
unemployment12
.
Modeling real-world activity using web search data can
provide a number of benefits. First, it can be more timely,
especially when the alternative is not electronically collected.
Influenza surveillance from the United States Centers for
Disease Control and Prevention (CDC), Influenza Sentinel
Provider Surveillance Network (ILINet) has a delay of one to
two weeks1
. For economic indicators like unemployment, this
delay is measured in months10
. In contrast, search data can
“predict the present” since it is available as the target activity
happens10
. Second, query data has good temporal and spatial
resolution. If an indicator of interest is incomplete (missing
time periods or regions, coarser temporal or spatial resolution,
etc.), query data can sometimes be used to fill in the gaps. For
example, influenza rate data from ILINet is only published by
the CDC at the national and regional level and is not published
for the off season13
, but models based on query data can be
used to provide estimates year-round and at a state and
sometimes even city level, provided there is sufficient search
activity at that level1,14,15
. Third, there can be considerable
expenses incurred in collecting data for traditional indicators.
Finally, while Internet users do not represent a random sample
of the United States population, this population has become
increasingly less biased over time and now represents 77% of
the adult population16
. In the 18-29 subgroup, this number is
almost 90%. This is in contrast to traditional landline phone
surveys which must either under-represent this age group or
blend in cell-phone survey data at considerable difficulty and
expense17
.
Three Google tools have been released previously to enable
access to aggregated online web search query data. Google
Trends and Google Insights for Search are both real-time
systems which provide temporal and spatial activity for a
given query. However, they are both unable to automatically
surface queries which correspond with a particular pattern of
activity. Google Flu Trends provides estimates of Influenza-like
Illness (ILI) activity in the United States, using models based
on query data. These queries are selected from millions of
possible candidates through an automated process1
. Due to
the computational requirements of this process, a batch-
based distributed computing framework18
was employed to
distribute the task across hundreds of machines.
Google Correlate builds on this previous work. Google
Correlate is a generalization of Flu Trends that allows for
automated query selection across millions of candidate
queries for any temporal or spatial pattern of interest. Similar
to Trends and Insights for Search, Google Correlate is an
online system and can surface its results in real time.
Data Summary
Using anonymized logs of Google web search queries
submitted from January 2003 to present, we computed two
different databases for Google Correlate:
us-weekly: temporal only: weekly time series data for the
United States at a national level.
us-states: spatial only: state-by-state series data for the United
States summed across all time.
Each database contains tens of millions of queries. For
additional details, please see the Data section below.
Methods Summary
The objective of Google Correlate is to surface the queries in
the database whose spatial or temporal pattern is most highly
correlated (R2
) with a target pattern. Google Correlate employs
a novel approximate nearest neighbor (ANN) algorithm over
millions of candidate queries in an online search tree to
produce results similar to the batch-based approach
employed by Google Flu Trends but in a fraction of a second.
For additional details, please see the Methods section below.
Flu Trends
Google Flu Trends produces estimates of ILI activity in the
United States using query data. The Flu Trends modeling
process is composed of two steps: variable selection and
model building. Google Correlate can perform the variable
selection and provide the associated time series data as a CSV
download to enable the construction of a model using the
selected queries. In this section we provide a test of the quality
and computational power of Google Correlate, demonstrating
that this automated system can be used to build a new Flu
Trends model for the United States with comparable
performance, but in a fraction of the time used to build the
original Flu Trends model.
The baseline for this comparison the original regional Google
Flu Trends model1
. For these models, query selection was
performed on the regional level, and a single set of queries
was chosen to optimize the results across all regions. The
values of the query time series were summed into a single
input variable per region, and a model was fitted from the data
across all nine regions. This model was built using weekly
training data between 9/28/2003 and 3/11/2007 inclusive,
and evaluated by computing the correlation between the
resulting predictive estimates and the corresponding regional
weekly truth data over the holdout period between 3/18/2007
3. Google Correlate Whitepaper 3
to 5/11/2008.
While we sought to make a close comparison between the
results of the Google Flu Trends methodology and modeling of
ILI activity using Google Correlate, there are several
differences between the methods employed. First, we worked
with a different resolution for query selection. Since Google
Correlate provides only national query time series data, we can
only perform query selection on the national level. After the
national-level query selection, we sum the query time series
into a single explanatory variable and fit a linear model to the
nine census regions. Second, we used a different cross-
validation technique for variable selection in Google Correlate
from the one used in Flu Trends.
We used Google Correlate to perform query selection by
uploading ILI activity data from the CDC over the training time
period. This weekly time series is at the national level and
represents the rate of ILI-related doctors office visits per
100,000 visits. We summed the time series of all 100 queries
returned by Google Correlate into a single explanatory variable.
We then fit a linear model to the nine census regions and
generated regional estimates for the holdout time period.
Training window correlation (R2
)
Mean Min Max
Google Flu Trends 0.90 0.80 0.96
Google Correlate 0.87 0.70 0.97
n = 9 regions
Holdout window correlation (R2
)
Mean Min Max
Google Flu Trends 0.97 0.92 0.99
Google Correlate 0.96 0.88 0.98
n = 9 regions
We see that the Google Correlate-based model slightly
underperforms the Flu Trends model for the hold out time, with
average correlation across all nine regions of 0.97 for Flu
Trends and 0.96 for Correlate. This difference could be due, in
part, to the difference in resolution of the query selection
process. The time required to create the model with Google
Correlate was a fraction of that required for the original Flu
Trends model.
Refinance
Every week, Mortgage Bankers Association of America (MBA)
compiles all mortgage application to refinance an existing
mortgage into a refinance index. The MBA’s loan application
survey covers more than half of all United States residential
mortgage loan applications and is considered by many to be
the best gauge of mortgage refinancing activity.
Consumers refinance a home for a number of reasons,
including to switch to a lower mortgage interest rate, to
change the mortgage length, to tap into their home equity and
to switch mortgage type. In 2003, the refinancing activity
peaked due to record low interest rate and the real estate
boom. Despite the lower mortgage interest rate in 2010, the
level of refinancing was not as high as in 2003 due to the
housing recession and the subprime credit crisis.
We examined the top 100 most correlated queries with the
refinance index time series from January 2003 to August
2010 and extended the window week by week until the end of
March 2011. Fifty percent of the selected queries were
refinance-related, including refinancing calculator, refinancing
closing costs, and refinance comparison. Mortgage rate related
queries such as lowest mortgage rates and no cost mortgage
accounted for about 35% of queries selected. Even though
queries for mortgage rates are related to refinancing, it is not
always about refinancing and thus the signal could be mixed.
Refi Index vs. Mortgage Rate
Refi Index vs. Search Volume of refinancing calculator
Using these queries, we applied the same method from Choi
and Varian10
and compared two alternative models with
baseline model with a moving window from August 2010 to
March 2011. Let yt
be the time series of the refinance index,
Refit
be the summed query time series for queries returned by
Google Correlate containing “refinance” or “refinancing”, and
Financet
be the summed query time series for all 100 queries
returned by Google Correlate.
4. Google Correlate Whitepaper 4
Baseline Model: yt
= α + φyt−1
+ et
Alternative Model 1: yt
= α + φyt−1
+ βRefit
+ et
Alternative Model 2: yt
= α + φyt−1
+ βFinancet
+ et
The model fit is significantly improved and prediction error
is decreased for the two alternatives. The out of sample
mean absolute error (MAE) with rolling window for the 31
weeks is decreased by 7.04% for Alternative Model 1 and
the MAE for Alternative Model 2 is increased by 9.12%.
Ribosome
A ribosome is a component inside living cells. Using the
us-weekly database, the query ribosome surfaces the following
highly-correlated (R2
> 0.96) queries:
1. mitochondria
2. cell wall
3. chloroplasts
4. chromatin
5. plant cells
6. vacuole
7. chloroplast
8. nuclear membrane
9. reticulum
10. cell function
The time series for these queries feature upticks in the Fall and
Spring, sharp drops during Thanksgiving and Christmas and a
long trough in the summer. This mirrors the school year in the
United States and suggests that the queries are being driven
by biology classes.
It is worth noting that all of these top terms relate to biology.
Other school topics (e.g. the Canterbury Tales) are also studied
early in the school semester and yet this time series is not
correlate nearly as well. It’s both surprising and impressive
that the phenomenon of biology study appears to be uniquely
characterized by its temporal pattern. This can be seen with
other queries, for example eigenvector, but to a smaller extent.
Latitude
Using a us-states data series containing the latitude for each
state in the United States, we find the following highly-
correlated queries were surfaced (R2
> 0.84):
1. sad light therapy
2. defroster
3. seasonal affective disorder lights
4. 10000 lux
5. sun lamp
6. track length
7. floor heating
8. fleece hat
9. irish water spaniel
10. hydronic
The “sad” in sad light therapy is likely the acronym for seasonal
affective disorder, which also seems to describe the
relationship between queries sad light therapy, seasonal
affective disorder lights, 10000 lux and sun lamp. These top
results surfaced by Google Correlate imply that latitude in the
United States can be modeled using the spatial patterns in
SAD-related queries. This is consistent with studies on the
correlation of SAD prevalence and latitude in North America19
.
Disclaimers
This system is not intended to serve as a replacement for
traditional data collection mechanisms. While the queries
selected by Google Correlate for a specific target series exhibit
strong correlations with the target series over many years, this
correspondence may not hold in the future due to changes in
user behavior which are unrelated to the target behavior. For
example, the correlation of a drug whose time series
historically tracked well the activity of a disease, could
significantly be changed by a recall of the drug.
Additionally, the underlying cause of search behavior can
never be known. Users submitting influenza-like illness (ILI)
queries are not necessarily experiencing ILI-symptoms. And
similarly, non-ILI related queries which are highly correlated
with an ILI series do not necessarily increase or decrease the
likelihood of contracting influenza.
Query data does not represent a random sample of the
population. While over three quarters of United States adults
use the Internet, several subgroups are underrepresented. This
could lead to sampling error depending on the modeling
performed.
Google Correlate requires indicators with unique spatial or
temporal patterns. Indicators with little variation or with very
regular variation are unlikely to surface meaningful results.
Indicators with unique variation may still not surface results
due to a lack of information-seeking behavior for the indicator.
Acknowledgements
The authors would like to thank Doug Beeferman and Jeremy
Ginsberg for providing early inspiration for Google Correlate.
We’d also like to thank Hal Varian for his valuable feedback on
Google Correlate and Jean-Baptiste Michel for his useful
comments on this manuscript. Finally, we’d like to thank Craig
5. Google Correlate Whitepaper 5
Nevill-Manning and Corinna Cortes for their guidance and
support.
Privacy
At Google, we recognize that privacy is important. None of the
data in Google Correlate can be associated with a particular
individual. The data contains no information about the identity,
IP address, or specific physical location of any user.
Furthermore, any original web search logs older than nine
months are anonymized in accordance with Google’s Privacy
Policy20
.
Data
Google Correlate contains two different databases of Google
web search queries. The first contains contains weekly time
series for the United States at a national resolution (us-weekly).
The second contains state-by-state series for the United
States summed across all time (us-states). Both datasets are
one-dimensional, with us-weekly having a time dimension but
no space dimension and us-states having a space dimension
but no time dimension. Both dataset contain tens of millions of
series.
To help smooth query data across similar underlying user
behavior, n-grams of the queries are used as series identifiers.
This approach is similar to Google Trends and Insights for
Search but is in contrast to Flu Trends where only lowercasing
was performed on the queries.
The following example illustrates how n-grams are extracted
from the query ‘cold and flu symptoms’.
cold *
cold and
cold and flu *
cold and flu symptoms *
and *
and flu
and flu symptoms
flu
flu symptoms *
symptoms *
This list is filtered to contain only n-grams which appear often
and in many states. The n-grams marked with an asterisk are
kept when this filter is applied using the us-weekly dataset.
Each of these filtered n-grams has a corresponding time
series stored in the database, and for each instance of ‘cold
and flu symptoms’ in the web search logs, each resulting
n-gram receives a count. Filtering is done for privacy reasons
but since rare queries are sporadic in nature, they are unlikely
to be useful for modeling of long term phenomena. Distracting
queries such as misspellings and those containing adult
sexual content are also excluded.
The series in both datasets are normalized by dividing by the
total count for all queries in that week (us-weekly) or state
(us-states). The normalization controls for the year over year
growth in all Internet search use (us-weekly) and state-by-
state variation in Internet usage (us-states). Finally, each time
series is standardized to have a mean value of zero and a
variance of one, so that queries can be easily compared.
Methods
In our Approximate Nearest Neighbor (ANN) system, we
achieve a good balance of precision and speed by using a
two-pass hash-based system. In the first pass, we compute an
approximate distance from the target series to a hash of each
series in our database. In the second pass, we compute the
exact distance function on the top results returned from the
first pass.
Each query is described as a series in a high-dimensional
space. For instance, for us-weekly, we use normalized weekly
counts from January 2003 to present to represent each query
in a 400+ dimensional space. For us-states, each query is
represented as a 51-dimensional vector (50 states and the
District of Columbia). Since the number of queries in the
database is in the tens of millions, computing the exact
correlation between the target series and each database
series is costly. To make search feasible at a large scale, we
employ an ANN system that allows fast and efficient search in
high-dimensional spaces.
Traditional tree-based nearest neighbors search methods are
not appropriate for Google Correlate due to the high
dimensionality which results in sparsenes. Most of these
methods reduce to brute force linear search with such data.
For Google Correlate, we used a novel asymmetric hashing
technique which uses the concept of projected quantization21
to reduce the search complexity. The core idea behind
projected quantization is to exploit the clustered nature of the
data, typically observed with various real-world applications.
At the training time, the database query series are projected in
to a set of lower dimensional spaces.
Each set of projections is further quantized using a clustering
method such as K-means. K-means is appropriate when the
distance between two series is given by Euclidean distance.
Since Pearson correlation can be easily converted into
Euclidean distance by normalizing each series to be a
standard Gaussian (mean of zero, variance of one) followed by
a simple scaling (for details, see appendix), K-means
clustering gives good quantization performance with the
Google Correlate data. Next, each series in the database is
represented by the center of the corresponding cluster.
6. Google Correlate Whitepaper 6
This gives a very compact representation of the query series.
For instance, if 256 clusters are generated, each query series
can be represented via a unique ID from 0 to 255. This requires
only 8 bits to represent a vector. This process is repeated for
each set of projections. In the above example, if there are m
sets of projections, it yields an 8m bit representation for each
vector.
During the online search, given the target series, the most
correlated database series are retrieved by asymmetric
matching. The key concept in asymmetric matching is that
the target query is not quantized but kept as the original
series. It is compared against the quantized version of each
database series. For instance, in our example, each database
series is represented as an 8m bit code. While matching,
this code is expanded by replacing each of the 8 bits by the
corresponding K-means center obtained at training time, and
Euclidean distance is computed between the target series
and the expanded database series. The sum of the Euclidean
distances between the target series and the database series
in m subspaces represents the approximate distance between
the two. Approximate distance between target series and the
database series is used to rank all the database series. Since
the number of centers is usually small, matching of the target
series against all the database series can be done very quickly.
To further improve the precision, we take the top one thousand
series from the database returned by our approximate search
system (the first pass) and reorder those by doing exact
correlation computation (the second pass). By combining
asymmetric hashes and reordering, the system is able to
achieve more than 99% precision for the top result at about
100 requests per second on O(100) machines, which is orders
of magnitude faster than exact search.
References
1. Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski
MS, et al. (2009) Detecting influenza epidemics using search
engine query data. Nature 457: 1012-1014.
2. Eysenbach G (2006) Infodemiology: tracking flu-related
searches on the web for syndromic surveillance. AMIA Annu
Symp Proc: 244-248.
3. Hulth A, Rydevik G, Linde A (2009) Web queries as a source
for syndromic surveillance. PLoS One 4: e4378-e4378.
4. Johnson HA, Wagner MM, Hogan WR, Chapman W,
Olszewski RT, et al. (2004) Analysis of Web access logs for
surveillance of influenza. Stud Health Technol Inform 107:
1202-1206.
5. Polgreen PM, Chen Y, Pennock DM, Nelson FD (2008) Using
internet searches for influenza surveillance. Clin Infect Dis 47:
1443-1448.
6. Pelat C, Turbelin Cm, Bar-Hen A, Flahault A, Valleron A-J
(2009) More diseases tracked by using Google Trends. Emerg
Infect Dis 15: 1327-1328.
7. http://ecmaj.ca/cgi/content/full/180/8/829
8. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2917042/
9. http://www.pnas.org/content/107/41/17486.full.pdf
10. http://www.google.com/googleblogs/pdfs/google_
predicting_the_present.pdf
11. http://www.nd.edu/~zda/Google.pdf
12. http://static.googleusercontent.com/external_content/
untrusted_dlcp/research.google.com/en/us/archive/papers/
initialclaimsUS.pdf
13. http://www.cdc.gov/flu/weekly
14. http://www.eht-journal.net/index.php/ehtj/article/
view/7183/8094
15. http://www.nature.com/nature/journal/v457/n7232/extref/
nature07634-s1.pdf
16. http://www.pewinternet.org/Static-Pages/Trend-Data/
Whos-Online.aspx
17. http://pewresearch.org/pubs/515/polling-cell-only-
problem
18. Dean, J. & Ghemawat, S. Mapreduce: Simplified data
processing on large clusters. OSDI: Sixth Symposium on
Operating System Design and Implementation (2004)
19. http://cbn.eldoc.ub.rug.nl/FILES/root/1999/JAffectDisordM
ersch/1999JAffectDisordMersch.pdf
20. http://www.google.com/privacypolicy.html
21. A. Gersho and R. M. Gray, Vector Quantization and Signal
Compression, Springer, 1991.