Computer — All MCQs
4795 questions available
Q2021.In the A* algorithm, the evaluation function is f(n) = g(n) + h(n), where h(n) is the:
Correct Answer: D
h(n) is the heuristic estimate of cost from node n to the goal.
Q2022.For A* to guarantee an optimal solution, the heuristic must be:
Correct Answer: B
An admissible heuristic never overestimates the true cost, ensuring A* optimality.
Q2023.Which deep learning architecture replaced RNNs in most modern NLP due to parallelization?
Correct Answer: A
Transformers process sequences in parallel via attention, outperforming RNNs in NLP.
Q2024.A model performs 99% on training data but 60% on test data. The most likely issue is:
Correct Answer: D
A large train-test gap indicates overfitting (poor generalization).
Q2025.Which technique is specifically designed to combat the vanishing gradient problem in deep RNNs?
Correct Answer: D
LSTM and GRU use gating mechanisms to preserve gradients across long sequences.
Q2026.In cybersecurity, AI-based 'behavioral analytics' detects threats by:
Correct Answer: D
Behavioral analytics flags anomalies that deviate from established normal patterns.
Q2027.Which of the following is a key risk when LLMs are used in investigations?
Correct Answer: B
LLMs can hallucinate, generating false yet convincing content—risky for evidence.
Q2028.Reinforcement Learning from Human Feedback (RLHF) is mainly used to:
Correct Answer: B
RLHF fine-tunes models using human feedback to align outputs with human values.
Q2029.Which AI approach would best identify a suspect's face across thousands of CCTV frames?
Correct Answer: C
Facial recognition combined with tracking matches identities across video frames.
Q2030.The 'exploration vs exploitation' trade-off is central to:
Correct Answer: D
RL agents must balance exploring new actions and exploiting known rewarding ones.
Q2031.Which method explains individual predictions by attributing importance to each feature?
Correct Answer: B
SHAP and LIME are XAI techniques that attribute feature contributions to predictions.
Q2032.An AI system trained mostly on one ethnicity misidentifies others. This is an example of:
Correct Answer: A
Skewed training data causes algorithmic bias, leading to unfair misidentification.
Q2033.Which technique allows an LLM to use external knowledge at query time without retraining?
Correct Answer: A
RAG retrieves relevant external documents to ground LLM responses without retraining.
Q2034.In NLP, 'stemming' differs from 'lemmatization' because stemming:
Correct Answer: B
Stemming crudely truncates words, while lemmatization uses linguistic/dictionary rules.
Q2035.Which is the correct order of a typical machine learning pipeline?
Correct Answer: C
The standard pipeline is collect, preprocess, train, evaluate, then deploy.
Q2036.An AI tool flags transactions structured just below reporting thresholds. This detects:
Correct Answer: C
Detecting sub-threshold structuring helps uncover money-laundering (smurfing).
Q2037.Which evaluation curve plots true positive rate against false positive rate?
Correct Answer: B
The ROC curve plots TPR vs FPR to evaluate classifier performance.
Q2038.The area under the ROC curve (AUC) of 0.5 indicates a model that is:
Correct Answer: D
AUC of 0.5 means the classifier performs no better than random chance.
Q2039.Which generative model uses a diffusion process to create images?
Correct Answer: D
Stable Diffusion is a diffusion-based generative model for image creation.
Q2040.In cyber crime investigation, AI 'link analysis' is used to:
Correct Answer: D
Link analysis maps and visualizes connections among entities in investigations.