Machine learning

Arthur Samuel (1959): Field of study that gives computers the ability to learn without being explicitly programmed.

Machine learning algorithms (covered in the class):

• ­Supervised learning
  • Linear Regression
  • Logistic Regression (Classification)
  • Neural Networks
  • Support Vector Machine

• ­Unsupervised learning
  • k-Means clustering
  • Principal Component Analysis

My understanding of machine learning:

Given a set of examples with certain features; the ability of  a computer to approach, and surpass, the ability of a human expert at analysing and extracting meaning out of the given data.

Important points about ML:

1. If a human expert does not find the data sufficient to come up with a conclusion, then the computer is unlikely to perform any better.
2. All machine learning algorithms are based on mathematics, and thus expect all data to be numbers.
3. Usually more data is better, but not if the data is redundant. That is, duplicate examples or features.

More to come in future posts.

Uncertainty, the name of the game

Cricket is the game glorious of uncertainties, and developers’ lives are no less uncertain, just not glorious. Both cricketers and developers need to be flexible and adjust to new situations. Be it change in requirements or the target score. Deal with the uncertainties of the pitch or the legacy code base.

Team work is important

Although there could be solo efforts that can steal the limelight, the fact remains that you need a team to win. You might think that a solo developer can deliver, but I will like to remind that, invariably this solo developer is building on works of thousands of other developers by using other libraries or frameworks.

Different skills make the difference

It takes all type to win the game. You might have all-rounders which add value, but specialists are also needed as there are very few all-rounder that are better than the best specialists. Similarly all kinds of skills are needed for coding, and rarely is there a dev with all skills. Talking of skills, I believe, you are born with the needed skills, but it takes years of practise to master those skills.

With skills you need attitude, that state of mind, which drives your actions. These are the common and opposing personalities that we encounter and each is needed as much as the other.

• The blaster and the wall
• The meticulous and the flamboyant
• The artist and the scientist

The Game Play

There is a definite need to be agile, to take stock of the burn-down or the run-rate after each over, and to revise your strategy accordingly. Breaking the target into sprints or sessions and finally, being time bound.

How to succeed

Like any thing in life, both coding and cricket need passion and commitment to succeed. There are only a few professions, where you do in the weekend, what you do during the week. You need continuously work on your skills even when not in office or a tournament.

There is a need for continuous improvement and to keep abreast with changed rules and strategies. The core tools remain the same, but the way you yield them differ over time.

One more thing

Both are literally back-breaking works but both pay well!

There is a conscious effort to remove the ‘male only’ label and involve the women as well, with good success.

Little background

My daughter will soon be six and she can use the computer very well for the usual tasks and I think she is a bit ahead of the curve, but that might just be a father speaking

The big bottleneck, I see, is her poor English skills. English not being our native tongue and she going to a German kindergarten, has limited her knowledge of English to a few words and phrases. Thus reading and writing code is going to be a challenge.

Visual programming

Given the language ‘handicap’, I started looking out for what the best way to visually teach programming, which does not involve much reading and writing. A little research resulted in 3 options, 2 of which are old and I had come across and one that is pretty new.

Alice

Alice is project from Carnegie Mellon University, with a lot of industry backing. It is quite mature, but does not seem to be actively developed. for example Alice 3.0 is under development for over 4 years and still in a beta state. The recommended documentation is to buy the ‘Learning to Program with Alice’ book.

In addition, it is targeted towards high school students, although and interesting, but unsupported side project, Storytelling Alice is targeted towards younger children.

For now, I have decided to give it a slip and will revisit once Alice 3.0 is more stable.

Scratch

Scratch is a project from MIT. It has a smaller learning curve, but is more limited compared to Alice. I think it only supports 2D compared to 3D support from Alice. This makes it more approachable for younger kids. In addition there are tons of lesson plans, documentation and samples available. Scratch has built-in sharing and there are literally a million user submitted projects.

All in all, I think Scratch will be a good starting point for my daughter. But, there’s more

Kodu

The latest entrant on the scene is Kodu, a Microsoft Research project. originally written for Xbox as a meta game (a game builder), a technical preview for PC has recently being released. It is a very visually engaging and has a polished feel to it. Given that it is a fairly new project, the documentation is lacking, save for the in game help and sample, and a few blog posts. Let’s see how it goes. There is always Scratch to fall back on.

And

Oh yes, there are a few more options, especially Logo and Lego Mindstorms NXT.

She was super excited and did most of the steps herself, including taking the pictures.

Here is a pictorial guide to making your own ‘hand-made’ paper.

Let me be quick in saying that my family and I loved the film and think it is one of the best movie. It is a good take on the state of education in the country, a subject I feel strongly about. This is also not about Chetan Bhagat, although the nit-picking could apply to Five Point Someone as well. But I haven’t read it, so can’t say.

I expected more from Aamir, the perfectionist.

I was born in 1978, the same year that Farhan Qureshi (Madhavan) was born. So I felt more connected to the movie than the current generation. This also means that the college period depicted in the movie was the same period I was in college, which is roughly 1995-1999.

So what?

Well you see, a few of the things shown in the movie are, let’s say, futuristic.

Mobile Phones

Mobile phones were launched in India after 1994. The call charges in a ‘scheme’ were 16 Rs. incoming/outgoing. They were not mainstream until 2000 when charges began to drop. Even then, in 2000, I was the only one in my MBA college to have a mobile phone, and that too was given to me by my employer.

So showing engineering college students carrying a phone, pre 1999, was a bit of a stretch.

Mobile Internet

Even if you think ‘chote’ (Rancho) was rich enough to carry a mobile, the movie shows mobile internet being used in the hospital for video conferencing.

• Internet in 1999 was the good old modem based dialup internet at blazing speeds of 56 kbps. Broadband Internet was unheard of until 2005.
• Mobile Data Cards / USB Modems were launched in 2008/2009.
• Even then, I don’t think the speed that Airtel EDGE gives is insufficient for video conferencing. I don’t know for sure, since I am not using one.

Scooter

The scooter that Pia (Kareena) drives looks most certainly like Kinetic Flyte, which was only launched in 2009.

Bottomline

With a little research, which Aamir Khan is known for, the film could have been more realistic. Anyways, if you haven’t seen the movie, please do, it is really good.

P.S. Happy New Year.

1. User requests a page
2. Request reaches Request Middlewares, which could manipulate or answer the request
3. The URLConffinds the related View using urls.py
4. View Middlewares are called, which could manipulate or answer the request
5. The view function is invoked
6. The view could optionally access data through models
7. All model-to-DB interactions are done via a manager
8. Views could use a special context if needed
9. The context is passed to the Template for rendering

1. Template uses Filters and Tags to render the output
2. Output is returned to the view
3. HTTPResponse is sent to the Response Middlerwares
4. Any of the response middlewares can enrich the response or return a completely new response
5. The response is sent to the user’s browser.

Please leave a comment if I have got something wrong.

Requirements:

• Installed ActivePython 2.6
• Installed Komodo Edit
• Installed Python Win32 Extensions (not sure why, but was recommended in some blog post and the project itself does not say much of what it does. Stuff like this gives me the heebie-jeebies.)
• Installed Pinax, which in turn installed Django.
• Installed PyQt4
• Installed Eric4 (uninstalled after using it for 10 minutes)

Problems faced:

• Pinax installation was a bit flawed, it could not install all dependencies. Worked around by manually installing (pip install else easy_install)
• Windows 7 was not passing command line arguments to .py scripts. Had to hack registry and add %* to the end to make it work. See the Key and value below.

Getting started:

Python: Like everyone else, I followed and recommend ‘Dive Into Python’. But more importantly, this page of titbits is amazing extract from the book and quickly brings a Java dev up to speed. I should blog about ‘Python for Java developers’. Time spent 4 hrs.

Pinax: Stopped at the installation step. Will delve further after understanding Django better.

Django: Followed the tutorial, then the Django Book, and finally the Django Docs for a deep dive. Time spent 8 hrs.

Django is surely one on the best documented project and also very straight forward. It suits my style of writing code and I did not feel like giving up in few hours, like the experience I had with RoR. Will surely blog about Django more.

I was brainstorming on the skills a software developer does not learn at school, which are essential for work and some which are needed for career growth.

Below is an embedded mind map, followed by an embedded public wave. Please feel free to contribute to the wave. If you do not have a wave account, here is the static version.

You can pan / zoom / expand / collapse the map below to get a better understanding.

Empty array vs. Empty list

An empty list is

[]

and an empty map is

{}

And empty array is

(,)

as it would cause ambiguity with a function / closure.

Type declaration vs. Initialisation

Declaration:

_myDict as Dictionary[of string, object]

Initialisation:

_myDict = Dictionary[of string, object]()

Assigning a Type vs. invoking a function

When you are calling a function don’t forget the brackets at the end.

Wrong:

fs = file.OpenRead

Correct:

fs = file.OpenRead()

If you make this mistake you will get errors like:

‘<method/property>’ is not a member of ‘<some type>’. (BCE0019)

Colon is the new semicolon

Remember the times when a missing semicolon used to be causing most compile errors, well, in boo it will be the missing colon which signifies a beginning of a block.

Wrong:

while ((line = filestream.ReadLine()) != null)

m.parse(line)

Correct:

while ((line = filestream.ReadLine()) != null):

m.parse(line)

If you forget the colon, you will get an error like:


Unexpected token: m. (BCE0043)
Unexpected token: <INDENT>. (BCE0043)
expecting "EOF", found '<DEDENT>'. (BCE0044)


Function definition in a script file

Maybe a no-brainer for python gurus, but if you have a script file, i.e. a file with no class, just instructions; functions are to be defined near the top of the script and called after it is def‘ed.

You can’t mix Boo with C# in same project

This is my mistake, but looking at samples having groovy and java files in same project; I blindly assumed that the same applies to Boo as well.

Intellisense / auto-complete does not work well.

Given the dynamic nature of Boo and maybe some bugs in SharpDevelop, it is not a great experience if you rely heavily on intellisense.

Embedding special characters in a string

Boo follows the python way of embedding special characters in string and not c#.

Wrong:

s = @“C:\path\to\dir\”

Correct:

s = “”"C:\path\to\dir\”"”

Embedding values in strings

The syntax is slightly weird, but useful nonetheless. You need to use both $ and enclose the variable in curly braces.

print “processing ${dir.Name}“

Still can’t find a way to pattern match and assign in the same line

I think this is possible, but I can’t figure a way to match a strung and assign the values to 2 variables.

I am trying to do something like, below and a few variations without luck:

key, val = /(?<key>[^=\s]+)\s*=\s*”(?<value>.+)”\s*$/.Match(line)

If the regex would have been simpler I could have used the Split instead of Match and things would have worked.

I ended up with something like:

data = /(?<key>[^=\s]+)\s*=\s*”(?<value>.+)”\s*$/.Match(line)

key, val = data.Groups["key"].Value, data.Groups["value"].Value

Embedding spaces in a regular expression patterns

As seen in the regex sample above, take extra care when you want to match a space. Use \s instead of a literal space.

Here is a list of programming languages available on these runtimes.

1. Can run on CLI using IKVM.NET
2. Can run on JVM using Mainsoft solution
3. Not yet usable
4. Can run on CLR, but is behind the JVM implementation

The main reason for the research was to identify a new language I should pick-up. I looked at Python and Ruby, but both have some sore thumbs that I just can’t stand. I really liked Boo and Groovy; they are similar to C#/Java in syntax and incorporate the good things from Python. Although I like Boo’s syntax and approach more than Groovy, Groovy has a more mature implementation and ecosystem. I will try to use Groovy for some hobby project and get a feel to things.