[Tutor] Learning Object Oriented Programming

[Rakesh Biswas]

It was shared to this list and it was a great email but this one was even
better.

best,

rb


On Sun, Jun 12, 2022, 4:14 AM <avi.e.gross at gmail.com> wrote:

> Alan,
>
> My first post was moderated and I have not received a copy here showing me
> it was shared. My main point was not that some aspects could not be
> discussed here but that others would be better done elsewhere or 1:1 with
> someone as this forum is dedicated for other purposes. I am adding or
> expanding on some of what you wrote as I have no disagreement.
>
> The part I found relevant was asking what aspects of Python the language to
> focus on for some purpose, albeit I find such questions hard to answer.
>
> Should you learn all about using sets in Python? Some people never find a
> need to use them and many languages have no such construct. But not
> learning
> them may confuse you later when you encounter what to you looks like a very
> badly formed dictionary that seems to be missing values, or is it keys?
>
> AI is not only a very broad field but a moving target that also keeps
> evolving. If you are interested in the AI I was studying while working on a
> Ph.D. in AI at Rutgers in the early 80's, it is barely recognizable and
> definitely did not use Python. Much of what was seen as new and interesting
> ended up migrating out from the umbrella of AI into other aspects of
> computer science and has become so commonplace that it is not considered
> challenging any more. Other aspects have fallen out of favor or been shown
> to not be a great approach.
>
> I have revisited many aspects of what is now AI in recent years and am
> especially focused on many kinds of statistical methods that overlap AI and
> I Machine Learning.  Much of that is done using Python extensions but I
> have
> done much work you would consider to be aspects of AI in other languages
> such as R or stand-alone applications.
>
> So the question here struck me as having several parts and that the person
> asking was either not very clear in their mind and sort of hoping to find
> some direction to get started. And, of course, sometimes it may simply be
> that they are just not that proficient yet in English.
>
> So if the question is whether learning a bit about Object-Oriented
> techniques in Python if your plan is to study AI, then the answer might be
> YES. But only up to a point. No need to be an expert at first.
>
> Many aspects of AI are mostly done using modules and packages others have
> already written and often tested. They come with pre-made objects that
> often
> are usable as black boxes with details well hidden within and the manuals
> tell you how to access or modify them, albeit good ones often hide much of
> that too and provide you with all the interfaces you normally need. Of
> course if you later want to tweak some things, you may need more, let alone
> if you want to share your own creations.
>
> Consider one of many scenarios in the scikit-learn module. You may want to
> do SOMETHING (and I am being abstract) that has variations.
>
> Typically you ask for a new object to be created that encapsulates what you
> want to do. You call code like:
>
> MyProblemSolver = doSomething(initializers)
>
> You now have an object which has used one of the initializers defined
> within
> that you do not need to see anything unless you feed it something wrong. It
> can set  all kinds of parameters,  maybe check validity, maybe creates
> accessory objects within and who knows what.
>
> You may also tweak it using methods supplied in additional lines of code
> like:
>
> MyProblemSolver.dosomething(...)
>
> I mean objectName.fit(...), objectName.transform(...),
> objectName.predict(...) and so on.
>
> One key part is giving it the data you want to use. For that, you may end
> up
> calling it with a method you hand something to that can range from a python
> list or some exotic object like a numpy Series or pandas DataFrame and so
> on.
>
> There are quite a few such object types  used for various purposes and you
> can make many of them and tweak each and compare results and so on.
>
> But you can use many of these tools without having the slightest idea what
> an object is. A well designed tool may be easy to use by setting reasonable
> defaults so if you want one of many other possible tweaks, like adjusting
> the level of an alpha variable or choosing a different algorithm for some
> part of a transaction by name or even supplying your own function for use
> in
> comparing two things internally, you can tweak it quite a bit and even make
> your own customizations. Beginners rarely need to.
>
> As I mentioned, if you want to do something with some Machine Learning such
> as a sort of neural network, a beginner is often better off skipping
> learning powerful multi-purpose packages like TensorFlow and going straight
> to Keras which is built on top of TensorFlow and pandas and other things
> and
> largely removes you even further from seeing the endless details beneath
> BUT
> may let you dip lower if you wish.
>
> Much depends on whether you are just interested in the topic, are a student
> hoping to major in something, or even looking at specific jobs. Many skills
> are not worthwhile to spend lots of time on except as a hobby.
>
> And, as noted, some of the modules written that encapsulate TensorFlow have
> a rather different but still pythonic aspect as what they do is less
> object-oriented and more function-oriented as you add layers by
> encapsulating functions within functions.  To use those you need to focus
> on
> other aspects of python.
>
> But here is an important point. Generally you do NOT do machine learning or
> other forms of AI alone. You often do lots of things in base python or
> other
> modules to read in the data, clean it up or do calculations on it, perhaps
> present graphics about it, and much more. Then you do one or more analyses
> as described above and often do more calculations on the various parts
> created in the process such as combining several and choosing which model
> made better results on your test cases and perhaps more analyses and graphs
> and ...
>
> So your overall knowledge of other aspects of python and the many modules
> people like for these tasks is important. I personally like to do some
> things using other languages and tools and often do programming using
> environments that create documents (as in a PDF or HTML or WORD documents)
> that include marking up all kinds of text as well as including snippets of
> python and R code that inter-operate with each other on the same data where
> each does what it is good at or provides what I like. This is what happens
> not at the beginning of a learning process but way later.  You sound like
> you need some baby steps headed in the right direction.
>
> So, as others have said, learn enough about python without worrying if you
> have mastered ALL the many ways to create formatted text and so on. I
> suggest that much of AI is data intensive and that learning numpy decently
> (but not expertly) should come next. Depending on your future needs, pandas
> would come next, or might be studied only if it ever is needed. You can
> always come back and learn missing things such as how to use sets. Man/most
> languages have no such thing built in and yet people make do using things
> like vectors of items along with abilities to make the contents unique() or
> do various union() and intersection() and setdiff() operations OR creating
> your own object-like creations or using packages from others.
>
> But you do not need to do this linearly. You can use later stages to help
> guide you by seeing what code they discuss that you have not understood.
> You
> can fill in the gaps. And you would be amazed how much (along with
> misleading nonsense) you can pick up with a web search or two if you
> include
> things like the name of the language and module as in "python numpy remove
> duplicates" or whatever you can phrase well enough.  Frankly, it can be a
> mistake to study python in a multi-volume way when the vast majority of
> what
> you learn is overkill. An intro book or two and maybe something
> intermediate
> but focused such as found using "book machine learning in python" or one
> about graphics and python may let you make progress.
>
> Now if you have specific questions about some code you read about use of
> objects and want to understand, this could be the place, but not if you
> have
> read nothing and want one on one attention to learn as that may require a
> real full-time dedicated person as a tutor.
>
> Again, good luck. Questions deeper than basic python, though, often have a
> better place, or even better, find resources you can read and practice with
> on your own as such learning can be longer-lasting.
>
>
> -----Original Message-----
> From: Tutor <tutor-bounces+avi.e.gross=gmail.com at python.org> On Behalf Of
> Alan Gauld via Tutor
> Sent: Saturday, June 11, 2022 7:56 AM
> To: tutor at python.org
> Subject: Re: [Tutor] Learning Object Oriented Programming
>
> On 11/06/2022 09:40, Aliyan Navaid wrote:
> > To be honest every AI project I've seen amazes me whether it is about
> > finding insights in data, NLP, or Image generation. However, could you
> > list down stuff (in order) that is essential to all these applications?
> >
>
> I see Avi has already posted a reply suggesting you won't get your answes
> here. That is true in terms of the ultimate and specific answers to your
> question however I do think its a fair question for a beginner to ask here.
> The reason is, and your question and follow-up suggest it is true of you
> too, that most beginners see an interesting field but have no idea of the
> complexity involved.
> In your case AI is a whole sub genre of computing with many specialisms
> within it. Until you know which specific sub-branch you want to explore it
> is impossible to give specific advice.
>
> What we can say is a that a good knowledge of general programming
> techniques
> will never go amiss and Functional Programming (which is not the same thing
> as just writing/using functions!) and the use of classes and objects are
> both worth studying. I don't think you need to study the full OOP paradigm
> since most AI projects seem to be more likely to be functional in nature
> with just a few classes or objects used as data carriers. This list and
> other python resources are certainly able to help with that.
>
> Other areas of general programming that may be worth considering include
> data storage(RDBMS, NO-SQL, and other "Big Data" solutions) as well as
> generic algorithms for searching/sorting etc. Wikipedia is probably a good
> starting place for these areas.
>
> Once you get into AI domain specific subjects you will probably find
> domain-specific fora that are better placed to help.
>
>
> --
> Alan G
> Author of the Learn to Program web site
> http://www.alan-g.me.uk/
> http://www.amazon.com/author/alan_gauld
> Follow my photo-blog on Flickr at:
> http://www.flickr.com/photos/alangauldphotos
>
>
> _______________________________________________
> Tutor maillist  -  Tutor at python.org
> To unsubscribe or change subscription options:
> https://mail.python.org/mailman/listinfo/tutor
>
> _______________________________________________
> Tutor maillist  -  Tutor at python.org
> To unsubscribe or change subscription options:
> https://mail.python.org/mailman/listinfo/tutor
>