Machine Learning is the study of algorithms that enable computers to learn patterns from data and make predictions or decisions without being explicitly programmed.

• Classification (spam detection, image recognition)
• Regression (predicting prices, temperature forecasting)
• Clustering (customer segmentation, grouping similar items)
• Anomaly detection (fraud detection, predictive maintenance)

A labeled training set is a dataset where each example comes with the correct output (label), which is used to train supervised learning algorithms.

• Classification
• Regression

• Clustering
• Dimensionality reduction
• Anomaly detection
• Association rule learning

Reinforcement Learning, as it allows the robot to learn optimal actions through trial-and-error interaction with the environment.

Unsupervised learning, specifically clustering algorithms (e.g., K-Means, hierarchical clustering).

Supervised learning, because labeled examples of spam and non-spam emails are available for training.

An online learning system updates the model incrementally as new data arrives, rather than retraining from scratch on the full dataset.

Out-of-core learning refers to training a model on datasets too large to fit in memory, by loading and processing the data in batches.

Instance-based or lazy learning algorithms, such as k-Nearest Neighbors (k-NN).

Model parameters are learned from training data (e.g., weights in a neural network), while hyperparameters are set before training (e.g., learning rate, number of layers).

They search for an internal representation of the data (model) that explains the relationships. They commonly use optimization strategies to minimize error and make predictions via the learned model.

• Insufficient or poor-quality data
• Overfitting / underfitting
• Feature selection and representation
• Computational complexity and scalability

This is overfitting. Possible solutions:

• Use more training data
• Regularization
• Simplify the model

A test set is a dataset held out from training used to evaluate the model’s generalization to new, unseen data.

Used to tune hyperparameters and select the best model without touching the test set.

It leads to overfitting to the test set, giving overly optimistic estimates of model performance.

Cross-validation splits data into multiple folds to train/test iteratively, giving a more reliable estimate of model performance than a single validation set.

The Big Picture phase defines what to build, why it matters, and how it fits into the ecosystem. It bridges human-centered design, data reasoning, and systems engineering.

• Discover — Divergent exploration of the problem
• Define — Converge to a clear problem
• Develop — Divergent exploration of solutions
• Deliver — Converge via prototyping and testing

Design Thinking uses empathy-driven observation, ideation, prototyping, and testing to create user-centered solutions while integrating iterative learning and feedback loops.

BizML is a business-driven ML methodology that ensures every ML project delivers measurable business or operational value, aligning data science, engineering, and strategy.

1. Value — Define business impact
2. Target — Specify prediction objectives
3. Performance — Set KPIs and metrics
4. Fuel — Prepare data
5. Algorithm — Train predictive model
6. Launch — Deploy and monitor model

Systems Thinking maps interactions among humans, devices, and cloud components, identifies leverage points and feedback loops, and integrates safety, trust, and reliability considerations.

Design Swigle is a hybrid agile–human-centered framework for continuous improvement in Edge AI. It enables rapid persona creation, prototyping, and feedback-driven refinement of ML systems.

Conduct ethnographic studies and sensor audits to understand users and devices, identify friction points like latency or unreliable connectivity, and collect human stories showing where edge intelligence adds value.

• User feedback refines design.
• Telemetry informs system architecture adjustments.
• Model drift triggers retraining.
• Business KPIs realign objectives.

It ensures that solutions are human-centered, technically feasible, aligned with business goals, and maintainable at the edge with continuous learning and monitoring loops.