What is machine learning in simple terms?
Machine learning is the ability of a machine to learn patterns from data and make decisions or predictions without being explicitly programmed.
How does machine learning differ from traditional programming?
In traditional programming, rules are explicitly written by humans, whereas in machine learning the system learns rules automatically from data.
What does it mean that machine learning predicts “unknown from uncertain information”?
It means the system infers outcomes or patterns even when data is noisy, incomplete, or ambiguous.
Why is machine learning described as “learning from data”?
Because the behavior of the system is derived from examples and observations rather than hand-coded rules.
Why is data so central to modern machine learning?
The world generates massive amounts of data, and learning from this data enables systems to model complex real-world phenomena.
What are the two main goals when learning from data?
Prediction (forecasting future or unknown outcomes) and description (understanding or summarizing the phenomenon).
How does machine learning help humans take better decisions?
It extracts patterns and knowledge from large datasets that humans cannot process manually.
Why are many real-world problems hard to program directly?
They involve pattern recognition tasks where rules are unclear, complex, or constantly changing.
What is the alternative to explicitly programming pattern recognition tasks?
Programming the computer to learn from examples using training data.
Why does standard software engineering methodology fail for some problems?
Because requirements may be unclear, experts may not articulate rules, or conditions may change rapidly.
Why are system requirements sometimes hard to collect?
There may be no human experts, experts may not explain their knowledge well, or user needs may differ or change.
Why are tasks like handwriting recognition or spam detection difficult to write rules for?
Because they involve subtle patterns that are hard to formalize explicitly.
How does machine learning replace guesswork with data?
Decisions are based on statistical patterns learned from real data rather than human assumptions.
How is machine learning related to artificial intelligence?
Machine learning is a subfield of AI focused specifically on enabling systems to learn from experience.
Why is spam email detection a machine learning problem?
Because spam patterns evolve and cannot be reliably captured with fixed rules.
What types of algorithms are commonly used for spam detection?
Naïve Bayes, rule mining, and other probabilistic classifiers.
Why is machine learning useful for recommendation systems?
It can identify user preferences and similarities based on behavior data.
What algorithms are commonly used for recommendation systems?
Collaborative filtering, nearest neighbor methods, and clustering.
Why is feature extraction necessary in machine learning?
Raw data is often large, noisy, and redundant, making learning inefficient and inaccurate.
What is feature extraction?
The process of representing raw data using a smaller set of informative and discriminative attributes.
Why can feeding raw data directly into a model be problematic?
It increases computational load, includes noise, and may obscure relevant patterns.
What is the main tradeoff in feature extraction?
Balancing information preservation with reduced computational and memory requirements.
Why should features be discriminative?
Discriminative features help distinguish between different classes or stimuli.
Why is loudness often removed in speech recognition feature extraction?
Loudness does not affect the linguistic content and introduces unnecessary noise.
What are the two main phases of machine learning?
Training and testing.
What happens during the training phase?
The system learns associations between features and known labels.
What happens during the testing phase?
The trained model classifies or predicts the label of new, unseen data.
What is supervised learning?
A form of learning where training examples are labeled with known output classes.
What is unsupervised learning?
A form of learning where no labels are provided and the system discovers patterns itself.
What is clustering an example of?
Unsupervised learning.
What is a perceptron in machine learning?
A basic neural network model inspired by the structure of the human brain.
How does a perceptron unit compute its output?
Inputs are weighted, summed, and passed through a nonlinear activation function.
Why is a perceptron considered a feed-forward network?
Information flows in one direction from input to output without feedback loops.
What does propagation mean in neural networks?
The process of passing computed outputs from one layer to the next.
What is the role of an error (cost) function in learning?
It measures how far the model’s output is from the desired output.
How does a machine learning model learn?
By adjusting its parameters (weights) to minimize the error function.
What is gradient descent?
An optimization method that iteratively adjusts weights in the direction that reduces error.
Why is gradient descent important in neural networks?
It provides a practical way to train models by minimizing error efficiently.
Last changed25 days ago