NPTEL WEEK 1
Q1.Write short note on machine learning and deep learning.(3 marks*)
Machine learning : It is a artificial intelligence, that computers being able to think and act
with less human intervation.It requires less computing powers.Computers are fed structured
data and learn to become better at evaluating and acting on that over data.In machine learning
there is supervised,semi-supervised,unsupervised and reinforcement learning.For ex.image
recognition,chatbot.
Deep learning : It is type of machine learning which subset of AI.Its machine learning to
think using structures modeled on the human brain.It needs less ongoing human intervation.It
analyzes images,videos and unstructured data.It calssifies two models that is discriminative
learning and generative learning. For ex.image pressing
Q2.What is feature descriptor in deep learning?(3 marks)
A feature descriptor is an algorithm which takes an image and output features
descriptors/features vectors.Feature discriptor encode interesting information into a series of
number and act as a short numerical "fingerprint" that can be used to differentiate one feature
from another.
Ideally,this information would be invarient under image transformation,so we can find the
feature again even if the image is transformed is some way.After detecting interest point we
go on to compute a descriptor for every one of them.Descriptor can be categorized into two
calss-local descriptor and global descriptor.
Q3.How Bayesian deep learning works.explain with example.(6 marks)
Bayesian deep learning is a discipline at the crossing between deep elaning architectures and
bayesian probability theory.The primary attraction of BDL is that it offers principled
uncertainty estimates from deep learning architectures.BDL models typically derive
estimations of uncertainty by either placing probability distributions over model weights or
by learning a direct mapping to probabilistic output.Epistimic uncertainty is modeled by
placing prior distribution overa models weights and then capturing how much these weights
varry over the data,on the other hand,uncertainty is modeled by placing a distribution over
output of the model.There are two primary types of uncertainty:
1. Aleatoric uncertainty : This kind of uncertainty can be reduced even with more data
collected.
2. Epistemic uncertainty : This kind of uncertainty can be reduced by collecting more
data.
* here are real world application some great eg , Bayesian inference can identify patterns or
clues for credit cad fraud by analyzing the data.It allows for indetify of spam message by
using bayes theorem,that can tell if an email is likely to be pam or not.It applied in medical
diagonoses to use data from previous cases and determine the probability of patient having a
certain disease.It used in robotics to help robots make decisions.
Q4.what is difference between Discriminative learning and generative learning.(6
Marks*)
Discriminative learning : Discriminative models, also called conditional models, tend to
learn the boundary between classes/labels in a dataset.discriminative models separate classes
by rather modeling the conditional probability and do not make any assumptions about the
data point. They are also not capable of generating new data instances.Discriminative models
have the advantage of being more robust to outliers.Discriminative model focuses on
predicting labels of the data.
* Examples of discriminative models in deep learning are: Logistic regression,Support vector
machine,Decision tree,Random forest
Generative learning: Generative models are models where the focus is the distribution of
individual classes in a dataset and the learning algorithms tend to model the underlying
patterns/distribution of the data points. Generative models use probability estimates and
likelihood to model data points and distinguish between different class labels in a dataset.
These models are capable of generating new data instances
* Examples of deep learning generative models : Naive Bayes,Hidden Markov model,Linear
discriminant analysis (LDA), dimensionality reduction technique.
Q5.What is Deep learning? Why do we need explain brief with example with their
application. (12 marks**)
Deep learning is a class of machine learning algorithms that uses multiple layers to
progressively extract higher-level features from the raw input.Deep learning techniques are
outperforming current machine learning technique.It is an artificial intelligence work that
mirrors the activities of the human bawin in preparing information.
WHY DO WE NEED:
1.The Need of deep learning in being able to Solve complex problems that requires
discovering hidden pattems in the data and/or a deep understanding of intricate relationships
between a large number of interdependent variables.
2.it is a subset of machine leaming, which is essentially a neural network with thate or more
layers.
3.These neural network attempt to simulate the behaviour of the human brain-albert for
From matching its ability - allowing it to "learn" from large amount data
4.The deep leaming architecture is flexible to be fitted to new troubles in the future.
5.It's work on the concept of repeated teaching. it trains the computer so that it can
understand a particular pattern and also identifies a picture or voice
6.Deep leaming is very much important as it makes Our task. accurate and fast. The capacity
to process large number of details makes deep leaming very strong when handling with
undeveloped dala
The Practical example of deep learning is:
1. Virtual assistants
2. Vision for driverless car
3. Money laundering
4. Face recognition
5. Image processing
Application of Deep leaming -
a. Self driving cars it is a one of the hottest areas of Study & business for teach demon
& is the power that is rejuvenating self- ruling driving.
b. Virtual Assistants is are one of the very popular application of deep learning, we all
use virtual assistants like. Alexa, Sixl, google Assistant in our day to day life
c. face recognition that help in arranging pictures dependent an areas identified in
photos, face, a mix of individuals, dates & so forth
d. Fraud Detection it's important application of deep teaming that is,
fraud detection. it is very useful for the banking of financial sector then frauds. saving
billions of dollars of cast in recuperation of Protection for monetary organizations.
