Need for Machine Learning

okay so in this video I will talk about the need for machine learning do we really need machine learning and you know if so what are the applications why do we care about machine learning really right so we're necessity is the mother of all invention so essentially if machine learning is so popular there must be some applications that really really really need machine learning okay so ability to mimic humans and automate certain monotonous tasks is the thing that actually led to the devil to the to the rise of machine learning okay so for example one of those tasks is recognizing handwritten characters okay so this is a snapshot of a popular data set called as the M missed numbers data set right so these are handwritten digits and the idea is you need to write or rather write machine learning code or be able to design a machine learning model such that it can actually automatically identify whether the number is five or six or seven and so on okay now of course this has lots of applications all across various sectors for example when you're filling insurance forms maybe you know a handwritten text that needs to be read by automated mechanisms so that it can be processed by downstream systems and so on okay so similarly when you write cheques basically you know in the bank there is some guy who actually tries to match the number with what you have written in words match the number with what you've written in words okay now can this be automated right so this can be automated if you can actually recognize digits or you can actually recognize a kind of friend writing like characters themselves right that's one application but there's so many other applications for example you know from the supermarkets you can actually get a whole bunch of transactional data where a transaction where every transaction refers to the real bill that you get that a customer gets when they buy something from from these supermarkets like Walmart or Moore store or Vegeta or any of those popular supermarkets right so the idea is to discover new knowledge from these large databases so once you have these transactional databases can you discover some new knowledge from them okay now this is where data mining which is more or less very related to machine learning is very useful okay so what you could analyze is these things that were bought together and try to understand which things were frequently Cobar okay and if they are frequently cobarde there must be this is a very important pattern which could be used to come up with better store layout designs for for designing discount bundles or for cross selling products when you want when some between you know when a customer buys product X can you recommend product Y and so on based on frequently cobalt patterns okay so that's yet another application of machine learning so in fact machine learning is also necessary for developing systems that can automatically enact and customize themselves to individual users okay so in fact such systems are called personalization based systems now these systems could include personalized search personalized recommendation systems that you see here or even personalized ads okay so so these are recommendations on a job portal recommendations on Amazon recommendations on Netflix about movies right jobs products movies right so recommended in a very personalized way based on your history or based on other users just similar to you what did they buy and so on all that requires machine learning right so machine learning is of course also useful for machine translation so you know coming up with with with documents of this form and then automatically translating them to other languages of course this helps not just for not just for for for official document purposes but it's also helpful for tourists when they are visiting different countries and they want to figure out what's going on right in a new country where they don't understand the actual language right so recently people have started using machine learning in in novel ways in a very new way right for example given an image producing a sentence to describe its contents also called as image captioning okay for example here you want to be able to say the dog is hiding okay so in people have been using machine learning to do tasks of that kind as well okay so a very simple and a concrete example of how machine learning can be done is the spam classifier that sits in all our inboxes right so if you use Gmail or you use any other email provider you know all of those email providers they have a module called as spam classifier whenever a email comes in to your to your email address the spam classifier runs on that email tries to figure out whether it is spam or normal or a useful email okay if it is spam it really puts it in the spam folder the junk folder as we call it okay so what is found all emails that user does not want to receive or has not asked to be delivered right to receive okay so the problem is to identify spam emails and what machine learning models do is to really use you know some reasonable amount of training data also called as labeled training data right and be able to learn this classifier such that the classifier accurately predicts whether this email spam or not okay so the accuracy of the classifier is in fact measured in terms of percent of filtered spam emails and percent of non-spam emails that were incorrectly filtered out okay so machine learning classifiers do need a machine and most of the superest machine learning models do need data a database of emails that were actually labeled by users so maybe you need like thousand emails which were not spam and thousand emails which was spam labeled so that this machine learning classifier can do a case based learning and therefore come up with this nice model okay so in fact machine learning revolution is is you know it has been around in stages so in fact in 1960s work on machine translation already started okay so promise of machine translation failed to move beyond basic machine translation it was too expensive and failed to deliver on its promise at that point was too expensive did not have enough label data and so on okay so in you know over time this this field has grown right and evolved and now you know there is in 2010 the field has grown so much that now it is known as deep learning a specific part of machine learning which which thanks to the you know new compute power and the availability of large data sets has blossomed like anything right so deep learning has been successful in in a variety of places so overall you know to summarize you know the idea is that there are a whole bunch of use cases which require machine learning or which significantly benefit from machine learning so machine learning has become popular now because of the advantage of the internet huge data okay so I mean the lots of use data is available on the internet now okay lot of compute power is also available earlier you know ten years back there were no GPUs or no reasonable GPUs on which you could learn from large amount of data but they are available now new optimized algorithms as well yeah and you theory is developed by the researchers new algorithms are coming up which can actually do wonders which can actually figure out lots of insights from your data yeah industry support has also increased people have started understanding the power of machine learning and therefore the industry funding industry support into this into these fields have significantly increased and that is why you know machine learning is is very popular right now not just that there is a need but actually it has become an integral part of many code processes yeah

18 thoughts on “Need for Machine Learning”

  1. I have 100% dataset I trained 80% data and 20% testing so that means I applied 20% testing on my labeled dataset plz correct me I m wrong

  2. Sir how to get a jobs in Data Science field as a fresher bcz my friends have done courses but they are struggling to get a job.

  3. #Thankyou sir for all your efforts u are putting to make these videos aware us about ML ……
    Really appreciate..
    #love from Kolkata

  4. Would have bought this course if Ravindra sir was speaker ! No offence, love Ravindra sir’s way of teaching.
    For nor andrew ng ✌🏻

Leave a Reply

Your email address will not be published. Required fields are marked *