YouTube Transcript:
Instance-Based Vs Model-Based Learning | Types of Machine Learning
Skip watching entire videos - get the full transcript, search for keywords, and copy with one click.
Share:
Video Transcript
View:
Hello hello viewers welcome to my YouTube
channel dresses for machine learning and today
sex and unlock today we are going to study an
important topic and its name is in
stone paste learning versus model based
learning so now for three days we have been
studying pipe machine learning first we
studied types of machine learning on the basis of the
amount of supervision required during
training from there another
difference is that sir or production machine
learning and how it is updating it,
so we have typed submachine gun a little and
today also we are going to study types of machine guns
but today our concern on the
basis of which we are deciding is how
our machine learning model launches,
how it learns okay so next and
discussion okay so machinery bottles learn in the previous two
ways just like
we learn human prince teams
ajay award 100g garlic in machine
learning so the first form of planning is a
very simple way of learning and think it's
basically memorizing It is
okay to keep it in, remember the responsibility
first form medical undertaking for mass generalizing
generalizing
or you can ask which one to understand, the
underlying principle, okay so
when we used to be in college or school, there were
two types of people around us or we
ourselves used to learn things from one of these two, either if
we did not understand then
we used to type creation, we would have to do the search India
exam or else we used to
confirm that we found something interesting,
which one behind it we used to learn, we used to
drive a car, right judge you will remember constipation,
there you had to keep something there, whereas
if you talk about education, there used to
type understanding, so
similarly in machinery also there are two types of models,
these two types of algorithms, one is
such a method which captures the data and
keeps it, okay and the second one is a
classical guitar, when we
get the data, we want to work on the linear principles within it, we
want to collect its underlying principles, okay so the
model which is put in the front We
call it instant based model
question based learning and the model which
tries to extrapolate the underlying principle
or tries to extract the fundamental meaning of the pattern is called
called
model based learning. So we will
search and see and by the way, the idea is that this
video has been sent to you. If you have not read about machine learning
algorithms, then the idea of this video is that
in future, when
you read the algorithm on the phone, then you will identify
whether it is doing model based learning or
instant face lifting, that is all you
understand. So let's start with the discussion
with instant based learning. There is no need to write it.
Okay, let's show
what happens in impressing that tree. Simple, but we have
a data set. I think
some relative improvement is seen from practice data.
But why do we have a
column which says CGPA, there is another problem
and we have a third column which says whether the
placement has happened or not.
Yes or no classification problem.
So here we have some data of Suraj. One lock
at at yes. 727 took this opportunity to introduce the school and
what we have to do is we have to do a classification code,
that is, if I have a new suggestion
and if I know his IQ and CGPA, then
I have to calculate whether he will be
placed or not, this
cannot be done by drone. Now if
you are working with an incident based algorithm, then
I will tell you how it will approach this problem. So the
incident of the interface problem is that it will
first block the data. So
yes students who got placed, this is
your IQ and this is our CGPA,
print and these are our Boston city students who did
not get placed. After all,
I have picked up why maximum number of
students should be placed, I
work in this, this is placement,
and in this, if I apply, then placement did not happen,
okay, but all your
blue points, these are both students and
placement, not work and rate, which are the points
from this point who got placed, okay,
and there will be some outliers, for example, here is
a child whose IQ is 0.5 and Boston city students who did not get placed. Both of them did
not happen in the good side and there
could be an enemy here,
right, the placement has been done. Now,
what you do in your entertainment loop is that you
get the training data stored in the training data, so you will learn, you
roll that data to your editor and you
say that when I am asked a question,
I will answer it instantly. Okay,
for example, in Classic, you got this point
whose CGPA is
1.5 and IQ is 103, the area of the
national is coming at this place,
so let's take any of our points and we have to
tell whether it is a remedy or a blue.
Now, what happens in the interview is that
you have not done any model element, you
focus on simple things and that simple
thing is that you have got, okay, you find the
similarity of this point, then
all the other points are found by smooth jasmine similarity
systems, near distance,
you can say that if this point is at this place, then
it will be similar to the nearby points of the
New Year. It will be like where you live
tells something about you that is a
quora idea okay so what will you do you will
find the distance of this point with every point okay and then grab the nearest koi team
points PM koi can do
50 you can grab my skin as if it has
three points this and this now
two points out of these three are rate then
you will say yes there will be placement because
most of you, only those who are close to you got
placed so you are also a
child like that you may get placed sometime in
test second sainik Rampura to impress the logic that
I just told you this is the logic of festival they
say that you will
study in dependent future but if you
want to study right now then I will put the link of that playlist in the
description of this video
but it is a letter but these 10 simple principles
in applied that you have no other way
you put the current point on and your work is done
now here only one thing is to be understood
that till this green point did not come, no
one had come to this point of yours Falguni,
I did not do anything at all, I was sitting there secretly updating you, as
soon as a new point came,
I caught the new point and saw
which were the nearby positions and whether they got placed or
not and declared my result, so
remember that in the spreading that
happens, there is no contact of any trimming or fighting,
you just
hold the point, as soon as a new point comes, you can
use accordingly details, what
model best of foreign studies learning
gas, okay, don't forget subscribe problem
with the help of model base plate wicket, so
I go to the flashlight on
a model based learning, this is your biggest
problem, why is the CGPA
placement done or not, you have a friend at that
here also you have kept a graph plotter
which must be dropped on the other side, why and
CGPA potatoes will not be given those points properly and the
children who did not get placed are amazing and the
children who got placed have 12th class FD certificate,
this is the child whose mind is
not good but got placed for the sweater,
okay, now present the model
What happens in model-based learning is that when
you give your algorithm updated points, it uses a mathematical function to
understand what it wants to do within that data point.
Okay, so
what it will do is draw a boundary like this. Let's say that it draws
a boundary. Now,
what the boundary is telling your model is
that by doing this,
if any point comes in this way, it will be
placed, and if any
point is missing, it will not be placed.
This is how your model will learn, run it on the data and
train it. Okay, now,
machine learning algorithms will
use this approach. There is definitely a mathematical
relationship between input and output, as a
result of which you get a decision
function like this. For classification
problems, you get a decision
function like this, you get a decision boundary.
Using this, what you can do is
classify those missing points.
So far, the good part about this approach is that you
train in this. There is no need for points, but
issues arise now you have the same
training points, you
have just the depression function,
you have the same addition function, now
from the left you have a new point, if anything is
clear, if it is on this side, then you will directly say that it is
placement, it will go, and if it is not placement like this, then you don't have
data points,
what did you do, you went into the data and what was the
concept that this is the line
using which the item is differentiated between the
children who got placement and the
children who did not get placement, so
till the doctor and the idea behind the model is that
now all you have to do in the future, guys, is that
you get some algorithm
and if you are learning that fire, how do
you understand it, you have asked a question that I have
read it, now I am sorry, I
have to press Hussain, I will present this model,
okay now there are many
examples of this, linear regression, logistic
regression, we have distant relatives, many examples.
examples.
Model based learning is of the same nature. There are only 12
gram pulses which is instance learning that
happened in BF rough networks web kar mill
functions. These are some examples that you
come across in your sexual intercourse. All other
algorithms and types of learning are not model based
learning. Okay, I just wanted to
discuss a little thing with you. Now let's
quickly see the
differences between instant and
model based learning. Okay, so here on the left side
you have the model sending and on the right side you have
your confidence. Okay, so I will
tell you about all the key points one by one.
Now you will have to do the data preparation equally in both
and you will have to remove the blocks. You will have to
code the technical data 110.
This time first we will
convert the springs into numbers. All this work has to be done at the
time of both. Here your
work with that
holiday square slot invented as
important cases. You will have to do the data processing
properly. This is the first point. The
second point in model based learning you While
generating the model, the feedback of the model is
this karma interest, okay, set up
parameters, this is what we discussed here,
what is this actually, this will be the Kochi question of Kar and Pushkar,
so one question I had to
find out, the roti question came up, that is called parameters, so it is
written here
that do bottle based learning, so you
get this model in return and the model is
different from some parameters, like lever
network, in it the input parameters are live
send properties, linear equations
log and intercept are parameters in poetry,
okay, arranged, you are doing
just bus learning, there these two
trains models tattoo discovery postponed
anil scoring any tree, in this recipe a
new point comes, you do not
do any pressure, you do not do anything with two tests,
okay, next in come
water being stored water in a suitable
form of journalism, do model based learning,
so if you do the model from somewhere or the other, then
use it in the yearning house to do the operation, where
informed, hence I would be a model Yes, it is
okay to do this if NAAC is
only in model based learning, use
generalized rule in the form of modern weapons
even before scoring instances Singh and
Inspector new points, if
you have to tell the answer about this then it is
written here that your entire focus is to find out the rule, what is the purpose, use the distributor page, that I can differentiate between placement or not, so I am standing behind the rule, not behind the students, okay, so this is written here and here it is written no generalization
report scoring only generalization and key
scoring in skin de villiers and seam, so
what happens here is that the logic of the model
works only when a new point comes
and you decide it according to the same point,
you should generalize it, your rule
depends on the incoming point, it will rise in this
model, send your rule for every
point, if any new point comes
then also it will remain the same, your high
tongue slip actors Credit for Unseen
Scoring Institution Model If you ever have a new
query point, then you
differentiate the model according to the one you have created
and here if you
test it by looking at the daily training data like
I did here, like we came up with an option,
we backed it up with the training data.
Mehraj model, in between you have no
concern with the training data. Okay, next
just come through if the input training data
of the model is training model based
learning. In this, even if you remove the training data, it doesn't make
any difference, but in this space
you always have to keep the training data with you
because your entire logic of
jurisdiction depends on that thing.
Okay, so next differences are
required. This precious volume. Okay,
basically you will have to use a particular function.
Now difference between two
classifications. Okay, smart important
points are points. We have already wasted a lot of
mathematical textbooks on this thing that
I don't want to go into, so ignore it. 10
glass points, this story models and
requires loop storage. So this is very The
important point is that in model based
learning, if you are just extracting celery traffic rules
and using this same rule for
production, then this model does
not have a lot of storage. Development is a
small thing. This question is
how much time will it take to call each situation.
How much will the Williams slip be? Very little tight.
Westminster Abbey swimming. You have to
keep a lot of training data with you at all times.
Now, a margin. If you have 1GB of training
data, then that 1GB of storage will
always be required. So, instance learning is a
bit heavy for you. Okay, so
so
quantum instant face value is also mentioned in the list right now
because in doing these French beans, during
training, your model is not doing
anything silently. 18 edits,
of course, we are less accurate. Okay, tears or
doctor discussion, you should know this much. If you are
ever studying any algorithm, then
you should just come to identify whether
it gives the model or instant state. There is nothing
special about this entire video.
Overall, Jo Jis Thakur is the team, who is
called the best? Yes, what is instant base called,
okay, next we will read the utensils, oil, we will tell you
that this is morning special and all this is okay,
show without or latest video I hope to
unite one so guys please consider
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
Works with YouTube, Coursera, Udemy and more educational platforms
Get Instant Transcripts: Just Edit the Domain in Your Address Bar!
YouTube
←
→
↻
https://www.youtube.com/watch?v=UF8uR6Z6KLc
YoutubeToText
←
→
↻
https://youtubetotext.net/watch?v=UF8uR6Z6KLc