Tag Archives: Data Science

Data Science Tools and Cloud Usage

15 Thursday Jan 2015

Posted by in Cloud, Data Science

≈ Leave a comment

Tags

, ,

At the end of 2014 O’Reilly published a Data Science Salary Survey report. Two areas of the report that caught my attention not because of the Salary side, but because of the other data collected and the trends it shows.

The first of these is the popularity of Tools that help enable Data Scientists. R and Excel seem to be on a par which is interesting to see as R is typically seen as being more powerful than Excel (I’m sure there is a bigger debate around that but wont get into it here!) , although Excel is more graphically pleasing to the user in manipulation of the data. However the data does not show where someone is using both or has a preference between one and the other.

common+tools

The respondents fall into several roles, which is most probably the swing between a Windows and Linux type environment and the tools used:

  • Analyst – includes coding
  • Statistician
  • Software developer
  • Technical lead
  • Manager
  • Product developer
  • Non-coding Analyst
  • Database administrator
  • UI/UX developer

Interesting that there is no one single role for a Data Scientist listed in the roles.

The report also shows the use of amount of cloud computing that is used by Data Scientists that responded to the survey. Approx a third still not moving to cloud, however two thirds are using it or experimenting with it in some way.  As the common tools are now being altered for the cloud, such as R cluster computing which is now available, there will be more shift to a cloud experience for data manipulation. The one thing that lets R down is the use of memory to hold and load data. The bigger the data set the more memory you need. This may change over time as a limitation and R Cluster is one way around this.

common+cloud

Of course this is only a report based on a number of respondents showing a sample of what is being carried out in the field of Data Science.  The trends may be different if run with a bigger data set and different roles responded.

Source: http://www.oreilly.com/data/free/2014-data-science-salary-survey.csp

Machine Intelligence Landscape

10 Saturday Jan 2015

Posted by in Data Science, Machine Learning

≈ Leave a comment

Tags

,

A great info sheet from http://www.shivonzilis.com/

Machine_Intelligence

Larger version of this can be found at: https://www.dropbox.com/s/qobt07e9skpk1z2/Machine_Intelligence_Landscape_12-10-2014.png?dl=0

Data Scientist Skill Set

05 Monday Jan 2015

Posted by in Data Science

≈ 1 Comment

Tags

,

O’Reilly released a free downloadable report a while back that presents the results of a survey of Data Scientists across the industry – circa 250 respondents. The report looks at a list of skills and classifies Data Scientists into 4 main categories:

  • Data Businessperson
  • Data Creative
  • Data Developer
  • Data Researcher

Under each of these headings the roles are defined as:

DS+Types

As an Architect I can see a fit to the “Jack of All Trades” box, however I think that there is a reach across the Researcher, Creative and Businessperson categories if we were to be classed. However as an Architect it is important to understand the skills that a Data Scientist needs across these areas as going forward there will be more opportunities to work side by side with Data Scientists in solutions and architectures.

The report gives a list of skills that a Data Scientist has under each classification of Data Scientist

  • Algorithms (ex: computational complexity, CS theory)
  • Back-End Programming (ex: JAVA/Rails/Objective C)
  • Bayesian/Monte-Carlo Statistics (ex: MCMC, BUGS)
  • Big and Distributed Data (ex: Hadoop, Map/Reduce)
  • Business (ex: management, business development, budgeting)
  • Classical Statistics (ex: general linear model, ANOVA)
  • Data Manipulation (ex: regexes, R, SAS, web scraping)
  • Front-End Programming (ex: JavaScript, HTML, CSS)
  • Graphical Models (ex: social networks, Bayes networks)
  • Machine Learning (ex: decision trees, neural nets, SVM, clustering)
  • Math (ex: linear algebra, real analysis, calculus)
  • Optimization (ex: linear, integer, convex, global)
  • Product Development (ex: design, project management)
  • Science (ex: experimental design, technical writing/publishing)
  • Simulation (ex: discrete, agent-based, continuous)
  • Spatial Statistics (ex: geographic covariates, GIS)
  • Structured Data (ex: SQL, JSON, XML)
  • Surveys and Marketing (ex: multinomial modeling)
  • Systems Administration (ex: *nix, DBA, cloud tech.)
  • Temporal Statistics (ex: forecasting, time-series analysis)
  • Unstructured Data (ex: noSQL, text mining)
  • Visualization

ML = Machine Learning

OR = Operations Research

From reading other reports this is by no means a full list of skills but provides a good insight into what a Data Scientist needs in their skills bag.

The report then looks at typical tasks that would be covered by each category and splits these into 22 core tasks across 5 main tasks.

Data+Science+Skills+2

The visualisation below illustrates the results showing the skills and tasks across each Data Scientist type to show a percentage of skill that is needed.

Data+Science+Skills

Overall a good report giving a highlight of the business areas and skills of a Data Scientist

Report Source

Analyzing the Analyzers

An Introspective Survey of Data Scientists and Their Work

http://www.oreilly.com/data/free/analyzing-the-analyzers.csp

The Data Scientists Toolbox Visualisation

31 Wednesday Dec 2014

Posted by in Data Science

≈ 2 Comments

Tags

,

Having found myself with a bit of spare time over the Christmas Holidays, I drew the below visualisation based on the Coursera course on “The Data Scientist’s Toolbox” based on some of the key takeaways I noted down from the course.

Its a bit like the show “Catchphrase” Say what you see…..

Click the drawing to get an enlarged view.

The Data Scientists Toolbox Visualisation

The Data Scientists Toolbox Visualisation

In-depth Introduction to Machine Learning

16 Tuesday Dec 2014

Posted by in Data Science

≈ 3 Comments

Tags

,

There is an excellent set of videos covering an In-depth introduction to Machine Learning spread over a number of videos and PDFs links below.

Chapter 1: Introduction (slides, playlist)

Chapter 2: Statistical Learning (slides, playlist)

 Chapter 3: Linear Regression (slides, playlist)

 Chapter 4: Classification (slides, playlist)

 Chapter 5: Resampling Methods (slides, playlist)

 Chapter 6: Linear Model Selection and Regularization (slides, playlist)

 Chapter 7: Moving Beyond Linearity (slides, playlist)

 Chapter 8: Tree-Based Methods (slides, playlist)

 Chapter 9: Support Vector Machines (slides, playlist)

 Chapter 10: Unsupervised Learning (slides, playlist)

 Interviews (playlist)

The links have come from the following Source: http://www.r-bloggers.com/in-depth-introduction-to-machine-learning-in-15-hours-of-expert-videos/

It is worth looking at this page as there are other links worth reading

Data Science and Information Theory

14 Sunday Dec 2014

Posted by in Data Science

≈ 1 Comment

Tags

,

I have recently been asked a question by a colleague “what’s the difference between Data Science and Information Theory?”

Here is my viewpoint on this.

Information Theory
Wikipedia states “Information theory is a branch of applied mathematics, electrical engineering, and computer science involving the quantification of information.

Information theory was developed by Claude E. Shannon to find fundamental limits on signal processing operations such as compressing data and on reliably storing and communicating data”[1]

The influencial article by mathematician Claude E Shannon in 1948 is called “A Mathematical Theory of Communciation”. [2]

Shannon states in his paper that a communication system essentially consists of 5 parts[2]

1. An information source which produces a message or sequence of messages to be communicated to the receiving terminal
2. A transmitter which operates on the message in some way to produce a signal suitable for transmission over the channel.
3. The channel is merely the medium used to transmit the signal from transmitter to receiver.
4. The receiver ordinarily performs the inverse operation of that done by the transmitter, reconstructing the message from the signal.
5. The destination is the person (or thing) for whom the message is intended.

Information Theory

He then goes on to discuss that communications are classified in three main categories of discrete, continuous and mixed. Then applies his mathematical computations and theory for each:

PART I: DISCRETE NOISELESS SYSTEMS
1. THE DISCRETE NOISELESS CHANNEL
2. THE DISCRETE SOURCE OF INFORMATION
3. THE SERIES OF APPROXIMATIONS TO ENGLISH
4. GRAPHICAL REPRESENTATION OF A MARKOFF PROCESS
5. ERGODIC AND MIXED SOURCES
6. CHOICE, UNCERTAINTY AND ENTROPY
7. THE ENTROPY OF AN INFORMATION SOURCE
8. REPRESENTATION OF THE ENCODING AND DECODING OPERATIONS
9. THE FUNDAMENTAL THEOREM FOR A NOISELESS CHANNEL
10. DISCUSSION AND EXAMPLES
PART II: THE DISCRETE CHANNEL WITH NOISE
11. REPRESENTATION OF A NOISY DISCRETE CHANNEL
12. EQUIVOCATION AND CHANNEL CAPACITY
13. THE FUNDAMENTAL THEOREM FOR A DISCRETE CHANNEL WITH NOISE
14. DISCUSSION
15. EXAMPLE OF A DISCRETE CHANNEL AND ITS CAPACITY
16. THE CHANNEL CAPACITY IN CERTAIN SPECIAL CASES
17. AN EXAMPLE OF EFFICIENT CODING
PART III: MATHEMATICAL PRELIMINARIES
18. SETS AND ENSEMBLES OF FUNCTIONS
19. BAND LIMITED ENSEMBLES OF FUNCTIONS
20. ENTROPY OF A CONTINUOUS DISTRIBUTION
21. ENTROPY OF AN ENSEMBLE OF FUNCTIONS
22. ENTROPY LOSS IN LINEAR FILTERS
23. ENTROPY OF A SUM OF TWO ENSEMBLES
PART IV: THE CONTINUOUS CHANNEL
24. THE CAPACITY OF A CONTINUOUS CHANNEL
25. CHANNEL CAPACITY WITH AN AVERAGE POWER LIMITATION
26. THE CHANNEL CAPACITY WITH A PEAK POWER LIMITATION
PART V: THE RATE FOR A CONTINUOUS SOURCE
27. FIDELITY EVALUATION FUNCTIONS
28. THE RATE FOR A SOURCE RELATIVE TO A FIDELITY EVALUATION
29. THE CALCULATION OF RATES

Data Science
Wikipedia states “Data science is, in general terms, the extraction of knowledge from data. The key word in this job title is “science,” with the main goals being to extract meaning from data and to produce data products.”[3]

A Data Scientist practitioner will use various tools and methodologies to extract the information to a question they are set to produce a set of answers. The most important part of Data Science is the question.

Some of the tools/methodologies used are (*this list is by no means complete):
• Computer Programming
• Data Engineering
• Data Warehousing
• Discrete Optimisation
• Geometric Methods
• Graphical Representation
• Information Theory
• Machine Learning
• Modelling
• Pattern recognition and learning
• Probability Models
• Statistical Learning

A great start to learning some of these are covered in my previous blog post: https://maxhemingway.com/2014/12/12/learning-data-science/

In Conclusion
In answer to the question Data Science can use the foundations that Shannon set out for Information Theory in 1948 and others as the theorem has progressed over time, as well as other tools/methodologies to answer the question set. Information Theory is part of the core syllabus on some Data Science courses [4]

Sources:
[1] http://en.wikipedia.org/wiki/Information_theory
[2] http://cm.bell-labs.com/cm/ms/what/shannonday/shannon1948.pdf
[3] http://en.wikipedia.org/wiki/Data_science
[4] http://www.ecp.fr/data_sciences_program

Learning Data Science

12 Friday Dec 2014

Posted by in Architecture, Data Science

≈ 5 Comments

Tags

,

As a Architect I am always looking into ways of working with new technologies, innovations and fields. One of this is Data Science and as such I am currently undertaking a set of courses to get an understanding of the field of Data Science.
Having a level of understanding will allow me to work more closely with data scientists in helpling them with suitable solutions as well as increase my skills in manipulating data.

The John Hopkins University are currently running a set of courses on Data Science consisting of 9 modules. Below is a dependency chart for anyone wanting to take these
Coursera Johns Hopkins Specialization in Data Science course dependency information

There are 9 modules to take which can be done free of charge or you can pay about $30.00 per course to get a certificate and take a final capstone project to test your understanding.

Each course lasts around 4 week and consists of video based lectures, forums, projects and knowledge test quizes.

John Hopkins consider two forms of dependencis for these courses:

Hard dependency: Students will be required to know material from the prerequisite course. Taking the dependent course simultaneously will be challenging and only possible for highly motivated students willing to work ahead of the course schedule for the prerequisite. Taking hard dependent courses out of order is not possible unless the student already knows the material covered in the prerequisite course.

Soft dependency: Knowledge of material from the prerequisite course is recommended and useful. Concurrently taking the prerequisite course and the dependent course is possible. It is not recommended to take them out of order, but would be possible for highly motivated students willing to self teach components of the prerequisite course as needed.

The courses are listed below in order that they should be taken in with links to the courses.
The Data Scientist’s Toolbox
https://www.coursera.org/course/datascitoolbox

This is the primary introductory course for the specialization. It should be taken first and has no prerequisite courses. Students should be computer literate, have programmed in at least one computer language and be motivated self learners.

R Programming
https://www.coursera.org/course/rprog

This is the most crucial course for the remainder of the specialization. It is softly dependent on The Data Scientist’s Toolbox. It should be taken before the remaining courses in the series.

Getting and Cleaning Data
https://www.coursera.org/course/getdata

This course has hard dependencies on R Programming and The Data Scientist’s Toolbox.

Exploratory Data Analysis
https://www.coursera.org/course/exdata

This course has hard dependencies on R Programming and The Data Scientist’s Toolbox.

Reproducible Research
https://www.coursera.org/course/repdata

This course has hard dependencies on R Programming and The Data Scientist’s Toolbox.

Statistical Inference
https://www.coursera.org/course/statinference

This course has hard dependencies on R Programming and The Data Scientist’s Toolbox. In addition, students will need basic (non calculus) mathematics skills.

Regression Models
https://www.coursera.org/course/regmods

This course has hard dependencies on R Programming, The Data Scientist’s Toolbox and Statistical Inference.

Practical Machine Learning
https://www.coursera.org/course/predmachlearn

This course has hard dependencies on R Programming, The Data Scientist’s Toolbox and Regression Models. It has a soft dependency on Exploratory Data Analysis.

Developing Data Products
https://www.coursera.org/course/devdataprod

This course has hard dependencies on R Programming, The Data Scientist’s Toolbox and Reproducible Research. It has a soft dependency of Exploratory Data Analysis.

*material from Coursera John Hopkins University
https://www.coursera.org/jhu