1. How can supervised learning be differentiated from unsupervised learning?
Ans:
Supervised learning relies on labeled data to help models understand how inputs correspond to correct outputs, enabling accurate predictions and classifications. Unsupervised learning analyzes unlabeled datasets to uncover hidden structures, clusters or natural groupings. Each method addresses different problem types and plays a crucial role in data-driven decision-making.
2. In what ways does transfer learning enhance machine learning models?
Ans:
Transfer learning strengthens model performance by adapting knowledge gained from a large pre-trained model to a smaller, domain-specific task. This approach reduces training time, lowers resource usage and delivers strong accuracy even when only limited data is available. Utilizing previously learned patterns helps models perform efficiently in specialized applications.
3. What does overfitting mean in machine learning and why is it problematic?
Ans:
When a model learns both meaningful patterns and unimportant noise from its training data, it is said to be overfitting, which results in subpar performance on fresh or untested datasets. This imbalance is avoided and generalization is enhanced by methods like regularization, pruning and cross-validation. Reducing overfitting guarantees that models are dependable in a variety of real-world scenarios.
4. What are GANs and how do they operate?
Ans:
Generative adversarial networks employ two models a generator that generates false data and a discriminator that evaluates the perceived realism of the data. The competition between these two networks forces the generator to provide outputs that are more convincing during training. GANs are widely utilized in tasks including image generation, creative media and data expansion.
5. How do attention mechanisms contribute to the effectiveness of transformer models?
Ans:
Attention mechanisms enable transformer models to concentrate on the most relevant portions of an input sequence. Through self-attention, each token assesses its relationship with every other token, capturing context and dependencies with high precision. This design significantly improves tasks involving text processing, translation and generative AI.
6. Why is feature engineering a vital step in machine learning?
Ans:
By creating, honing or choosing variables that emphasize crucial patterns and relationships, feature engineering improves raw data. Algorithms can learn more effectively with well-structured information, increasing prediction accuracy and generalization. On the other hand, regardless of the intricacy of the model, poor feature design might impair performance.
7. How is missing data typically handled within datasets?
Ans:
Imputing values using averages like mean, median or mode or utilizing models that can naturally handle missing entries are common ways to deal with missing data. Removing impacted rows or columns may be required when the amount of missing data is considerable. The size of the dataset and the significance of the missing values determine the strategy that is selected.
8. What distinguishes Random Forest from XGBoost in their modeling approaches?
Ans:
Random Forest builds a collection of independent decision trees and averages their outputs, lowering variance and reducing overfitting. XGBoost, on the other hand, constructs trees sequentially, with each new tree correcting the shortcomings of previous ones, delivering high predictive strength. Both techniques are powerful but rely on different learning strategies.
9. How is a machine learning model’s performance evaluated?
Ans:
The model's evaluation depends on the type of problem being solved. While classification tasks demand metrics like accuracy, precision, recall, F1-score and ROC AUC, regression tasks rely on statistics like mean squared error, mean absolute error and R-squared. These measures help evaluate the reliability and viability of a model's deployment.
10. What ethical principles should be followed when developing AI systems?
Ans:
Fairness is the main goal of ethical AI development which emphasizes robust privacy protection, decision-making openness and a reduction in bias. To avoid abuse and foster user confidence, systems should continue to be transparent and responsible. Following ethical standards guarantees that AI solutions function ethically and in line with society expectations.
