Machine Learning (ML)-based System

A Machine Learning (ML)-based System is a automated data-driven system that include learning system which implement machine learning algorithms (to solve a machine learning-based task).

• Context:
  • It can (typically) be created by an ML Engineering Team (that follows a ML system development model).
  • It can (typically) make use of an ML Technology (such as an ML platform or ML libraries).
  • It can range from being an Unsupervised Learning-based System to being a Supervised Learning-based System.
  • It can range from being an Unsupervised Learning System, to being a Supervised Learning System, to being a Semi-Supervised Learning System, to being a Reinforcement Learning System.
  • It can range from being a Partially-Automated Learning System to being a Fully-Automated Learning System.
  • It can range from being a Passive Learning System to being a Active Learning System.
  • It can range from being a Batch Learning System to being an Online Learning System.
  • It can be implemented using an ML Framework (that can include an ML library, or ML tools).
  • It can (typically) be composed of ML Pipelines (such as an ML training pipeline).
  • It can support a Machine Learning-based Application.
  • It can be developed using an ML-System Engineering Development Methodology.
  • …
• Example(s):
  • ML-based Checker System, such as the one that used Samuel's checker program.
  • Personalized Recommender Systems, such as: Netflix's Movie Recommender System and PlayStation's Personalized Store Item Recommendation System.
  • one that includes a ML-based Natural Language Processing System.
  • a Robot Learning System.
  • a Computer Vision System such as: a Face Recognition System, or an OCR-based Mail Sorting System.
  • a Learning Software Package.
  • a Feature Learning System.
  • a Decision Tree Learning System.
  • a Supervised Learning System such as: a Regression Learning System or a Classification Learning System.
  • an Unsupervised Learning System such as: a Clustering System or a Self-Organizing Map.
  • a Reinforcement Learning System such as: a Relational Reinforcement Learning System.
  • a Computer-based Bio-Surveillance System.
  • …
• Counter-Example(s):
  • a Rule-based System, such as a Heuristic-based system or a Rule-based Expert System.
  • a Non-Learning Robot, such as a Roomba robot.
  • a Virtual Software-based Learning System.
  • a Deductive Learning System.
• See: Inductive Learning System, Machine Learning Engineering, Inductive Inference, Inductive Logic Programming-based System, Quantum Machine Learning System.

---

---

References

2017A

• (Sammut & Webb, 2017) ⇒ "Unsupervised Learning". In: Sammut, C., Webb, G.I. (eds) Encyclopedia of Machine Learning and Data Mining. Springer, Boston, MA pp. 1304-1304
  • QUOTE: Unsupervised learning refers to any machine learning process that seeks to learn structure in the absence of either an identified output (cf. supervised learning) or feedback (cf. reinforcement learning). Three typical examples of unsupervised learning are clustering, association rules, and self-organizing maps.

2017B

• (Sammut & Webb, 2017) ⇒ Supervised Learning. In: Sammut, C., Webb, G.I. (eds) Encyclopedia of Machine Learning and Data Mining. Springer, Boston, MA pp. 1213-1214
  • QUOTE: Supervised learning refers to any machine learning process that learns a function from an input type to an output type using data comprising examples that have both input and output values. Two typical examples of supervised learning are classification learning and regression. In these cases, the output types are respectively categorical (the classes) and numeric. Supervised learning stands in contrast to unsupervised learning, which seeks to learn structure in data, and to reinforcement learning in which sequential decision-making policies are learned from reward with no examples of “correct” behavior.

2017C

• (Stone, 2017) ⇒ Stone P. (2017) Reinforcement Learning. In: Sammut, C., Webb, G.I. (eds) Encyclopedia of Machine Learning and Data Mining. Springer, Boston, MA pp. 1088-1090
  • QUOTE: Reinforcement Learning describes a large class of learning problems characteristic of autonomous agents interacting in an environment: sequential decision-making problems with delayed reward. Reinforcement-learning algorithms seek to learn a policy (mapping from states to actions) that maximizes the reward received over time.

    Unlike in supervised learning problems, in reinforcement-learning problems, there are no labeled examples of correct and incorrect behavior. However, unlike unsupervised learning problems, a reward signal can be perceived.

2016

• Martin Zinkevich. (2017). “Rules of Machine Learning: Best Practices for ML Engineering."
  • QUOTE: … Most of the problems you will face are, in fact, engineering problems. … So, the basic approach is: … make sure your pipeline is solid end to end
    • Rule #1: Don’t be afraid to launch a product without machine learning.
    • Rule #2: Make metrics design and implementation a priority.
    • Rule #3: Choose machine learning over a complex heuristic.
    • Rule #4: Keep the first model simple and get the infrastructure right.
    • Rule #5: Test the infrastructure independently from the machine learning.
    • Rule #6: Be careful about dropped data when copying pipelines.
    • Rule #7: Turn heuristics into features, or handle them externally.
    • …

2013

• (IEEE, 2013) ⇒ http://www.ieee-ras.org/robot-learning
  • The robot hardware is progressively becoming more complex, which leads to growing interest in applying machine learning and statistics approaches within the robotics community. At the same time, there has been a growth within the machine learning community in using robots as motivating applications for new algorithms and formalisms. Considerable evidence of this exists in the use of robot learning approaches in high-profile competitions such as RoboCup and the DARPA Challenges, and the growing number of research programs funded by governments around the world. Additionally, the volume of research is increasing, as shown by the number of robot learning papers accepted to IROS and ICRA, and the corresponding number of learning sessions.

---