Medium Level – GK Questions

Hyperparameter tuning: grid search, random search, Bayesian optimization. Critical for maximizing performance.

A/B testing randomly assigns users to model variants, measuring metrics like conversion, engagement.

Self-supervised learning creates supervisory signals from data structure: mask prediction, next sentence prediction.

Section 115BBH taxes crypto income at 30% plus cess, with 1% TDS on transactions.

Micro-segmentation creates granular security zones, containing breaches and limiting attacker movement.

Data minimization limits collection to what's necessary for specified purpose. Reduces privacy risks, storage costs, breach impact.

Defensive distillation trains a student model to mimic a teacher's softened outputs, reducing gradient magnitudes and adversarial vulnerability.

IAM systems manage user identities, authentication, authorization. Critical for implementing least privilege access controls in Zero Trust.

PM-WANI enables public Wi-Fi hotspots via Public Data Offices using satellite/fiber backhaul. Aims to provide affordable internet in rural/remote areas.

Aditya-L1 positioned at Sun-Earth L1 point for uninterrupted solar observation. Studies coronal heating, CMEs, solar flares for space weather forecasting.

India Semiconductor Mission implements ₹76,000 crore incentive scheme for fabs, display fabs, ATMP, design. Partnerships with global players announced.

Leading qubit implementations: superconducting (IBM, Google), trapped ions (IonQ), photonic (Xanadu). Each has trade-offs in coherence, scalability, error rates.

NM-QTA funds quantum computing, communication, sensing, materials research. Implemented by DST with IITs, IISc participation.

Right to Portability (GDPR Article 20) enables individuals to obtain and reuse personal data across services. Promotes competition and user control.

GDPR Article 33 requires breach notification to supervisory authority within 72 hours of awareness. Critical for incident response compliance.

Purpose Limitation prevents function creep by ensuring data is used only as consented. Core principle in GDPR, DPDP Act, and global privacy frameworks.

Membership inference exploits model confidence to infer training data membership. Mitigation: differential privacy, regularization, output perturbation.

Data poisoning injects malicious samples; backdoor attacks embed triggers that activate misclassification. Both compromise model integrity during training.

Runtime security (Falco, Tetragon) monitors system calls, network activity, file access in real-time to detect threats in containerized environments.

Security Context defines pod/container privileges; Pod Security Standards enforce baseline/restricted policies. Both critical for container security.