Poe
Sure! Here’s a detailed response for Question 1:
1. Explain different types of Machine Learning techniques. [5]
Machine Learning techniques can be broadly categorized into three main types:
Supervised Learning: This technique involves training a model on a labeled dataset, where the
input data is paired with the correct output. The model learns to map inputs to outputs.
Common algorithms include:
Linear Regression
Logistic Regression
Decision Trees
Support Vector Machines (SVM)
Neural Networks
Unsupervised Learning: In this approach, the model is trained on data without explicit labels. The
goal is to find patterns or groupings within the data. Common techniques include:
Clustering (e.g., K-Means, Hierarchical Clustering)
Dimensionality Reduction (e.g., PCA, t-SNE)
Association Learning (e.g., Apriori Algorithm)
Reinforcement Learning: This technique focuses on training agents to make decisions by
interacting with an environment. The agent learns through rewards and penalties, optimizing its
actions to maximize cumulative reward. Examples include:
Q-Learning
Deep Q-Networks (DQN)
Policy Gradient Methods
2. Differentiate overfitting and underfitting problems encountered during Machine Learning. [5]
Overfitting: This occurs when a model learns not only the underlying patterns in the training data
but also the noise. As a result, the model performs well on training data but poorly on unseen
data. Indicators include:
High accuracy on training data but low accuracy on validation/test data.
Complexity of the model is too high relative to the amount of data.
Underfitting: This happens when a model is too simple to capture the underlying trends in the
data. It fails to perform well on both training and unseen data. Indicators include:
Poor performance on both training and validation/test datasets.
The model's assumptions are too simplistic for the given problem.
3. Explain Logistic Regression and its applications. [5]
Logistic Regression is a statistical method for predicting binary classes. It models the probability
that a given input belongs to a particular category using the logistic function. The output is a
value between 0 and 1, which can be interpreted as the probability of the positive class.
Applications:
Medical Diagnosis: Predicting the presence or absence of a disease based on patient data.
Credit Scoring: Evaluating the likelihood of a borrower defaulting on a loan.
Marketing: Analyzing whether a customer will purchase a product based on their behavior and
demographics.
4. Write down the applications of Neural Networks. [5]
Neural Networks have a wide range of applications, including:
Image Recognition: Used in facial recognition, object detection, and image classification tasks.
Natural Language Processing (NLP): Applied in language translation, sentiment analysis, and
chatbots.
Healthcare: Assisting in medical diagnosis, such as detecting tumors in medical images.
Finance: Fraud detection, algorithmic trading, and credit scoring.
Autonomous Vehicles: Enabling perception tasks such as object detection and lane detection.
5. Explain Support Vector Machine. [6]
Support Vector Machine (SVM) is a supervised learning algorithm used for classification and
regression tasks. It works by finding the hyperplane that best separates different classes in the
feature space. The key components include:
Hyperplane: A decision boundary that separates classes.
Support Vectors: Data points that are closest to the hyperplane and influence its position. They
are critical in defining the hyperplane.
Kernel Trick: SVM can use various kernel functions to transform input data into higher-
dimensional spaces, allowing for non-linear separation.
SVM is effective in high-dimensional spaces and is robust against overfitting, especially in cases
where the number of dimensions exceeds the number of samples.
6. Define the terms Hyperplane, Support Vectors, Kernel, Hard and Soft Margin. [5]
Hyperplane: A flat affine subspace that divides the space into two half-spaces. In two dimensions,
it is a line; in three dimensions, it is a plane.
Support Vectors: The data points that lie closest to the hyperplane. They are critical for defining
the hyperplane and the decision boundary.
Kernel: A function that enables SVM to operate in a high-dimensional space without explicitly
transforming the data, allowing for non-linear classification.
Hard Margin: A type of SVM that requires all data points to be correctly classified with a clear
margin. It works well when the classes are linearly separable.
Soft Margin: Allows some misclassifications to achieve better generalization. It introduces a
regularization parameter to balance the trade-off between maximizing the margin and
minimizing classification errors.
7. What are Ensemble Learning Models? Explain bagging and boosting in detail. [6]
Ensemble Learning Models combine multiple learners to improve the overall performance. The
idea is to create a model that performs better than any single model alone.
Bagging (Bootstrap Aggregating): This technique involves training multiple models (usually of the
same type) on different subsets of the training data, sampled with replacement. The final
prediction is made by averaging the predictions (for regression) or voting (for classification).
Bagging reduces variance and helps prevent overfitting. An example is Random Forest.
Boosting: This technique sequentially trains models, where each model focuses on correcting the
errors of its predecessor. The models are combined to create a strong learner. Boosting reduces
both bias and variance. Examples include AdaBoost and Gradient Boosting.
8. What is the role of the Activation Functions in Neural Networks? List down the names of some
popular Activation Functions used in Neural Networks. [6]
Role of Activation Functions: Activation functions introduce non-linearity into the model,
enabling neural networks to learn complex patterns. They determine whether a neuron should be
activated based on the input it receives.
Popular Activation Functions:
Sigmoid: Outputs values between 0 and 1, often used in binary classification.
Tanh: Outputs values between -1 and 1, centered around zero.
ReLU (Rectified Linear Unit): Outputs the input directly if positive; otherwise, it outputs zero. It
helps mitigate the vanishing gradient problem.
Leaky ReLU: Similar to ReLU but allows a small, non-zero gradient when the input is negative.
Softmax: Used in multi-class classification to output a probability distribution over classes.
9. Explain the architecture of the Multilayer Feed-Forward Neural Network. [6]
A Multilayer Feed-Forward Neural Network consists of:
Input Layer: The first layer that receives input data. Each node represents a feature of the input.
Hidden Layers: One or more layers where the actual processing occurs. Each neuron in these
layers applies a weighted sum of the inputs followed by an activation function.
Output Layer: The final layer that produces the output. The number of neurons corresponds to
the number of classes in classification tasks or a single neuron for regression.
The architecture allows for complex relationships between inputs and outputs through multiple
layers and non-linear activation functions.
10. Explain Gradient Descent and its types. What are the different steps used in the Gradient
Descent algorithm? [6]
Gradient Descent is an optimization algorithm used to minimize the loss function in machine
learning models. It iteratively adjusts model parameters in the direction of the steepest descent
of the loss function.
Types of Gradient Descent:
Batch Gradient Descent: Uses the entire dataset to compute the gradient at each step. It is stable
but can be slow for large datasets.
Stochastic Gradient Descent (SGD): Updates the parameters using one training example at a time.
It is faster but introduces more noise into the updates.
Mini-Batch Gradient Descent: A compromise between batch and stochastic methods, it uses a
small subset of the data to compute updates.
Steps in Gradient Descent:
Initialize parameters (weights).
Compute the predicted output using the current parameters.
Calculate the loss (error) between predicted and actual output.
Compute the gradient of the loss function with respect to each parameter.
Update the parameters by moving in the opposite direction of the gradient, scaled by a learning
rate.
Repeat steps 2-5 until convergence or a specified number of iterations is reached.
11. What is a perceptron and what are its basic components? How does a perceptron work? [6]
A Perceptron is the simplest form of a neural network, serving as a binary classifier. It consists of:
Input Features: The input values fed into the perceptron.
Weights: Each input feature has an associated weight that adjusts its influence on the output.
Bias: An additional parameter that allows shifting the activation function.
Activation Function: Typically a step function that outputs either 0 or 1 based on whether the
weighted sum exceeds a certain threshold.
How it Works:
Compute the weighted sum of the inputs plus the bias.
Apply the activation function to the weighted sum.
Output the result (0 or 1) based on the activation function.
This process allows the perceptron to classify inputs into two distinct classes.
