Why Machine Learning Mirrors Human Learning: From Features to Reinforcement

Forms of Machine Learning

Many phenomena can be described from multiple perspectives; detailed descriptions capture fine details while abstract names provide concise labels. Translating these varied descriptions into numerical form is essential for computers, which rely on digital computation.

Feature Representation

To enable computation, real‑world attributes are transformed into numerical features . For example, the size of a flower can be represented by the radius of an imagined circle.

However, representing abstract concepts like Chinese words is challenging because numerical distances may not reflect semantic relationships.

One‑Hot Encoding

A simple method for encoding categorical items is One‑Hot encoding. Assuming a finite vocabulary of n words, each word is represented by an n -dimensional vector with a single 1 at the position corresponding to the word and 0s elsewhere.

Learning Paradigms

Machine learning models act as mappings from input features to desired outputs. Three primary learning paradigms exist:

Supervised learning : The model is trained on input‑output pairs, adjusting until its predictions match known targets.

Unsupervised learning : No explicit targets are provided; the model seeks patterns or structures in the data.

Reinforcement learning : The model interacts with an environment, receiving rewards and new states, learning to maximize cumulative reward.

Human Learning Analogies

These paradigms mirror human learning experiences: classroom instruction resembles supervised learning, research exploration aligns with unsupervised learning, and real‑world decision‑making reflects reinforcement learning. Analogous roles such as "top students" (supervised), "researchers" (unsupervised), and "class leaders" (reinforcement) illustrate the connection.