SECTION A – What These Questions Are About
Section A contains 10 short conceptual questions.
Each question targets a specific core idea from neural networks, fuzzy logic, evolutionary algorithms, and hybrid soft-computing techniques.
What topics these questions belong to:
• Multilayer vs single-layer perceptron → From Artificial Neural Networks (ANN); deals with capability to learn non-linear patterns, hidden layers, and backpropagation.
• Hopfield network purpose → From associative memory models; used for pattern storage and retrieval using energy minimization.
• Membership function in fuzzy sets → From fuzzy logic; defines how strongly an element belongs to a fuzzy set (0–1 value).
• Fuzzy vs crisp relations → Fuzzy relations allow graded associations, crisp relations allow only yes/no connections.
• Data clustering algorithms → From unsupervised learning; K-means, hierarchical clustering, SOM, etc.
• Neuro-fuzzy controls → Hybrid systems combining fuzzy logic rules with ANN learning ability.
• Survival of the fittest → Evolutionary algorithms concept referring to selection of best individuals for next generations.
• Mutation → Introduces random genetic changes to maintain diversity and avoid premature convergence.
• Hybrid fuzzy controller → Combines fuzzy inference with other AI methods like ANN or GA.
• Traveling Salesman Problem (TSP) → Classic NP-hard optimization problem; find shortest route visiting cities once.
Purpose of Section A:
To check understanding of fundamental theory and soft computing concepts.
SECTION B – What These Questions Are About
Section B contains theoretical explanation questions requiring deeper understanding (any 3).
What topics these questions belong to:
• Backpropagation in neural networks → Algorithm to adjust weights by minimizing error through gradient descent.
• Fuzzy logic for uncertain decision-making → Uses membership functions and fuzzy inference to handle vague or imprecise data.
• Evolutionary algorithms and natural evolution → Mimic natural selection, crossover, mutation, and fitness evaluation.
• Rank method in evolutionary computation → Assigns fitness based on ordering rather than absolute fitness values.
• Soft computing techniques in MATLAB → Using ANN toolbox, FIS editor, clustering tools, GA toolbox, etc.
Purpose of Section B:
To evaluate analytical explanation ability regarding training, decision-making, optimization, and implementation.
SECTION C – What These Questions Are About
This section focuses on self-learning networks, fuzzy automata, fuzzy uncertainty handling, clustering, optimization, etc.
Part 3 Topics
• Kohonen Self-Organizing Map (SOM) → Competitive learning, winner-neuron update, neighborhood adjustment.
• Hopfield network storage & retrieval → Storing patterns as stable states using symmetric weight matrices and recalling via energy minimization.
Part 4 Topics
• Fuzzy languages in fuzzy automata → Strings processed with fuzzy transitions, fuzzy states, degrees of acceptance.
• Handling uncertainty using fuzzy functions → Fuzzy sets represent uncertain inputs/outputs via membership functions.
Part 5 Topics
• Clustering algorithms for grouping data → Find similarity-based groups using K-means, SOM, FCM, etc.
• Simulated annealing optimization → Energy-based probabilistic search inspired by metal cooling process.
Purpose of Section C:
To test logical understanding of learning behavior, fuzzy processing, optimization and clustering.
SECTION D – Evolutionary Computation & GA Applications
This part focuses on fitness computation, rank-space search, and genetic algorithms.
What these questions are about:
• Role of fitness computation → Determines which individuals survive, mate, mutate; guides evolutionary direction.
• Rank-space method → Converts rank into selection probability controlling selective pressure.
• Genetic Algorithm for TSP → Represents cities as chromosome sequence; uses crossover, mutation, and selection to find shortest route.
• GA vs traditional search techniques → GA is population-based, stochastic, parallel, robust; traditional search is sequential and deterministic.
Purpose:
To evaluate understanding of optimization, genetic operators, and search techniques in evolutionary computation.
