1) _____________________
Collaborative Filtering (CF) is most widely used techniques in recommendation
systems.
2) Common similarity metrics include Cosine Similarity andand
_____________________ Pearson Correlation Coefficient
_______________________
𝐵
𝑃 𝑃(𝐴)
3) As per Bayes Rule P(A/B) = 𝐴
_____________________
𝑃(𝐵)
1) ________________________________
Collaborative Filtering (CF) works on the principle: “Users who agreed in the past
tend to agree again.”
2) Collaborative Filtering suffers from the cold start problem
_____________________
3) A similarity score close to 1___ indicates very similar users in cosine
similarity.
1) There are two main types of Collaborative Filtering
User-based Collaborative Filtering,
________________________________ Item-based Collaborative
and _____________________________ Filtering
2) P(Spam ∣ Offer): Probability an email is spam,
____ given it contains the word
Offer
____
3) In Bayesian classification, we choose the class that has the highest
posterior
_________ probability given the input features.
ELECTRONICS AND COMMUNICATION ENGINEERING
8TH SEMESTER
1) Introduction: Basic definitions, types of learning, hypothesis space and inductive bias, evaluation, cross-validation. Linear
regression, Decision trees, over fitting
8 Hrs.
2) Instance based learning, Feature reduction, Collaborative filtering based recommendation. Probability and Bayes learning.
8 Hrs.
3) Logistic Regression, Support Vector Machine, Kernel function and Kernel SVM
8 Hrs.
4) Neural Network: Perceptron, multilayer network, back propagation, introduction to deep neural network
8 Hrs.
5) Computational learning theory, PAC learning model, Sample complexity, VC Dimension, Ensemble learning. Clustering: k-
means, adaptive hierarchical clustering, Gaussian mixture model.
8 Hrs.
LOGISTIC
REGRESSION
LOGISTIC REGRESSION
Logistic Regression is a supervised learning algorithm used for classification
problems, especially binary classification
Despite its name, logistic regression is not used for regression tasks—it’s used for
predicting discrete outcomes, typically probabilities that map to class labels.
LOGISTIC REGRESSION
We are modeling the probability that Y=1, given x:
1
𝑃 𝑌=1𝑥 =
1 + 𝑒 −(𝑚𝑥+𝑏)
𝑃 𝑌=1𝑥
log = −(𝑚𝑥 + 𝑏)
1−𝑃 𝑌 = 1 𝑥
This is why it’s called logistic regression—we’re modeling the log of odds as a linear function.
LOGISTIC REGRESSION
How do we find “m” and “b”?
Cross-entropy loss (also called log loss) measures the difference between the actual
label and the predicted probability.
LOGISTIC REGRESSION
LOGISTIC REGRESSION
ELECTRONICS AND COMMUNICATION ENGINEERING
8TH SEMESTER
1) Introduction: Basic definitions, types of learning, hypothesis space and inductive bias, evaluation, cross-validation. Linear
regression, Decision trees, over fitting
8 Hrs.
2) Instance based learning, Feature reduction, Collaborative filtering based recommendation. Probability and Bayes learning.
8 Hrs.
3) Logistic Regression, Support Vector Machine, Kernel function and Kernel SVM
8 Hrs.
4) Neural Network: Perceptron, multilayer network, back propagation, introduction to deep neural network
8 Hrs.
5) Computational learning theory, PAC learning model, Sample complexity, VC Dimension, Ensemble learning. Clustering: k-
means, adaptive hierarchical clustering, Gaussian mixture model.
8 Hrs.
SUPPORT VECTOR
MACHINE
SUPPORT VECTOR MACHINE
MARGIN
SUPPORT VECTOR MACHINE
1. Identifies the data points closest to the boundary — these are called support vectors.
2. Draws the boundary so that it is as far as possible from these closest points.
3. Maximizing this margin increases the model’s confidence and generalization ability.
HANDLING NON-SEPARABLE DATA WITH SOFT MARGINS
In real life, data is messy. Some points:
•Might overlap between classes
•Might be outliers
•May not be separable by a clear margin
Hard-margin SVMs require perfect separation. If even
one point is on the wrong side, the optimization fails.
Instead of demanding perfect classification, SVM with soft margins allows:
•Some violations of the margin boundary
•A trade-off between margin width and misclassification
HANDLING NON-SEPARABLE DATA WITH SOFT MARGINS
HANDLING NON-SEPARABLE DATA WITH SOFT MARGINS
VERY LOW VALUE OF C VERY HIGH VALUE OF C
VERY LOW VALUE OF C VERY HIGH VALUE OF C
NON-LINEAR SVM & THE KERNEL FUNCTION
Feature transformation to higher dimensions is often computationally expensive,
especially for large datasets.
How can we get the benefit of transformation without the cost?
NON-LINEAR SVM & THE KERNEL FUNCTION
NON-LINEAR SVM & THE KERNEL FUNCTION
NON-LINEAR SVM & THE KERNEL FUNCTION
NON-LINEAR SVM & THE KERNEL FUNCTION
NON-LINEAR SVM & THE KERNEL FUNCTION