1] Fixed Weight Competitive Nets :

Fixed Weight Competitive Nets are a type of neural network often used in unsupervised learning to
perform tasks like clustering or pattern recognition. They are designed to group similar data points
together by having nodes compete to "win" representation of the input, meaning that only one node
becomes active (or "wins") for a given input, while others are suppressed.

An example of a fixed-weight competitive network is the Kohonen Self-Organizing Map (SOM),

which organizes inputs into a grid-like structure and is often used for clustering and data
visualization.

2] Principal Component Analysis

Principal Component Analysis (PCA) is a dimensionality reduction technique used in machine

learning and data analysis to simplify complex datasets.

The first principal component captures the direction of the most variance in the data.

The second principal component captures the next highest variance direction, orthogonal to the
first, and so on.

PCA is based on the eigenvalues and eigenvectors of the data's covariance matrix. The eigenvectors
represent the principal components' directions, while the eigenvalues indicate the amount of
variance along each component.

3] Logistic Regression

Logistic Regression is a statistical method used for binary classification tasks in machine learning,
where the goal is to predict one of two possible outcomes (e.g., yes/no, 0/1, true/false) based on
input features. Despite its name, logistic regression is used for classification, not regression.

In a medical setting, logistic regression might be used to predict the probability of a patient having a
disease (1) or not (0) based on features like age, blood pressure, and test results. The output
probability can then be used to classify patients as high or low risk.

4] Neural Networks Basics

Neural Networks are computing systems inspired by the structure and function of the human brain,
designed to recognize patterns, make decisions, and learn from data.

perceptron is the simplest type of neural network unit (or neuron) that can be used for binary
classification.

 Structure: It takes multiple inputs, multiplies each input by a weight, sums them up, and
passes this sum through an activation function (like the step or sigmoid function) to produce
an output.

mathematical form of perceptron

Learning: During training, the perceptron adjusts its weights to minimize classification errors,
learning to classify inputs into one of two classes.

Multi-Layer Perceptron (MLP), capable of learning more complex patterns. An MLP consists of an
input layer, one or more hidden layers, and an output layer, where each neuron in a layer is
connected to all neurons in the next layer.

Backpropagation: This is the process by which an MLP learns from errors by adjusting weights in the
opposite direction of the error gradient. During backpropagation:

1. The error between predicted and actual output is calculated.

2. Using gradient descent, the weights are updated to minimize this error, starting from the
output layer and moving backward through the network.

3. This iterative process allows the network to learn by minimizing errors over multiple training
examples.

Same for AND and NOT Gate where -0.5 is the bias
Based on the syllabus topics from the image, here is a brief description of each topic, suitable for a 2-
mark answer. Let me know if you need further details on any specific topic!

Unit III: Learning

1. Machine Learning: Field of study focused on using algorithms and statistical models to
enable computers to learn patterns from data and make decisions without explicit
programming.

2. Classification: A type of supervised learning that assigns labels to input data, such as
predicting whether an email is spam or not.
 3. Regression: A supervised learning technique that predicts continuous values, like predicting
house prices based on features like size and location.

4. Approaches of Machine Learning Models: Includes supervised, unsupervised, semi-

supervised, and reinforcement learning approaches to train models based on labeled or
unlabeled data.

5. Types of Learning: Refers to supervised, unsupervised, and reinforcement learning, which

are different methods for training machine learning models based on data types and goals.

6. Probability: A branch of mathematics dealing with the likelihood of occurrences, crucial in

machine learning for handling uncertainty in predictions.

7. Basics of Linear Algebra: Fundamental mathematics for machine learning involving vectors,
matrices, and transformations, important for handling multidimensional data.

8. Hypothesis Space and Inductive Bias: Hypothesis space is the set of all possible models;
inductive bias is the set of assumptions made by the model to learn effectively.

9. Evaluation: Measuring a model's performance using metrics like accuracy, precision, recall,
and F1 score.

10. Training and Test Sets: Data is split into training (to learn patterns) and test sets (to evaluate
performance) to prevent overfitting.

11. Cross-Validation: A technique to evaluate model performance by splitting data into several
subsets and training on different combinations of these subsets.

12. Concept of Overfitting and Underfitting: Overfitting occurs when a model learns noise;
underfitting happens when it fails to capture patterns in the data.

13. Bias and Variance: Bias is the error due to oversimplified models; variance is the error from
sensitivity to small fluctuations in training data.

14. Linear Regression: A regression algorithm that models the relationship between variables by
fitting a straight line.

15. Logistic Regression: A classification algorithm that predicts probabilities for binary classes,
using the sigmoid function.

Unit IV: Supervised Learning

1. Neural Network: Computational models inspired by the brain, used to learn patterns from
data, consisting of layers of interconnected "neurons."

2. Perceptron Networks: A simple neural network with a single layer used for binary
classification.

3. Adaline (Adaptive Linear Neuron): A single-layer neural network that uses a linear activation
function for classification tasks.

4. Backpropagation Networks: A method for training multi-layer neural networks by minimizing

the error between predicted and actual values.
 5. Decision Tree: A model that makes decisions by splitting data into branches based on feature
values, resembling a tree structure.

6. Entropy: A measure of uncertainty or impurity in a dataset, used to decide splits in decision

trees.

7. Information Gain: A metric used in decision trees to choose splits that maximize reduction in
entropy.

8. Gini Impurity: A metric for evaluating splits in decision trees based on how well classes are
separated.

9. Classification Algorithm: Algorithms used to categorize data into classes, such as SVM,
decision trees, and k-nearest neighbors.

10. Rule-Based Classification: A method of classification based on "if-then" rules derived from
training data.

11. Naïve Bayesian Classification: A probabilistic classifier based on Bayes' theorem, assuming
features are independent.

12. Support Vector Machines (SVM): A classification technique that finds the optimal boundary
(hyperplane) to separate data points from different classes.

Unit V: Unsupervised Learning

1. Unsupervised Learning: Learning patterns from unlabeled data, with applications in

clustering and association.

2. Principal Component Analysis (PCA): A dimensionality reduction technique that identifies

the main axes of variance in data.

3. Neural Network: Fixed Weight Competitive Nets: Networks where weights are fixed, and
neurons compete to represent input patterns.

4. Kohonen Self-Organizing Feature Maps: A type of unsupervised neural network that clusters
data by mapping it into a lower-dimensional grid.

5. Clustering: A method in unsupervised learning where data is grouped based on similarity

without predefined labels.

6. Types of Clustering: Includes k-means, hierarchical clustering, and DBSCAN, each using
different approaches to group data.

7. Hierarchical Clustering Algorithms: A clustering technique that builds a hierarchy of clusters

using agglomerative or divisive approaches.

8. k-means Algorithm: A clustering algorithm that partitions data into k clusters by minimizing
the distance between data points and cluster centroids.