The Feynman Technique with examples

By Jorge Lison. All rights reserved.

What is the Feynman Technique?

The "Feynman Technique" is a methodology commonly used to understand concepts, remember complex ideas, and test or solidify our own knowledge.

freynman-technique-diagram-h-jorgelisonReview: break downexplanations.4Practical examples3Definition and explanation2Concept1

At jorgelison.com we will be using this methodology to create a series of blog posts and articles relating to artificial intelligence, data mining techniques, web development, and the Internet of Things (IoT).

1. Pick and learn about a concept

example

For example: "what is machine learning?"

Dedicate some time to go through the didactic resources that you consider convenient. Good learning materials are:

  • Accurate
  • Useful
  • Concise
  • Relevant

The most effective method of presentation depends on the content.

2. Define the concept


Explain the idea to yourself as if you are the student.

If you get stuck during your explanation, stop and review your learning material. Try to explain the concept to yourself again, without using any educational resources.


example

Example:
Machine learning is the art of implicitly programming machines by using data.


Original resource:
"Machine learning is a subset of artificial intelligence in the field of computer science that often uses statistical techniques to give computers the ability to learn (i.e., progressively improve performance on a specific task) with data, without being explicitly programmed. [...]"

Source: "Machine Learning." Wikipedia, Wikimedia Foundation.

3. Work through practical examples

For example, create a table or a chart that classifies the different types of machine learning techniques.

Machine Learning

The art of implicitly programming machines by using data.

Type Description Examples
... ...
  • ...
  • ...

Click (arrow down) for a full example...

example

Type of Machine Learning Systems (1/2)


Type Description Examples
Supervised The data sets provided to the machine include the desired solution(s), a.k.a "labeled training data." Systems for classification tasks and
predictions, such as:
  • Linear regresion
  • Logistic regresion
  • K-nearest neighbors
  • Decision trees and random forests
  • Support Vector Machines (SVMs)
Unsupervised The data sets provided to the machine does NOT include solution(s). The machine produces its own solution to the problem (task), a.k.a "unlabeled training data." Clustering:
  • K-means
  • Hierarchical Cluster Analysis (HCA)
  • Expectation Maximization
Visualization and dimensionality reduction:
  • Principal Component Analysis (PCA)
  • Kernel PCA
  • Locally-Linear Embedding (LLE)
  • t-distributed Stochastic Neighbor Embedding (t-SNE)
Association rule learning:
  • Appriori
  • Eclat

Click (arrow down) to continue...

example

Type of Machine Learning Systems (2/2)


Type Description Examples
Semisupervised The data sets provided to the machine contains a mix of labeled and unlabeled data. A combination of supervised and unsupervised systems, such as:
  • Deep Belief Networks (DBNs)
Reinforcement Machine actions have rewards or penalties in return. The machine must find the best strategy to maximize rewards over time.
  • Markov Decision Process

4. Review your explanations


Simplify complex terminology and break down complicated terms into even simpler explanations. Imagine that you are trying to explain that new concept to a five year old that keeps asking “why? why? why?...”


example

Example:
Machine learning is the art of implicitly (indirectly) programming machines by using data (examples).



Click (arrow down) to see revisited "Type of Machine Learning Systems" table...

example

Revision: Type of Machine Learning Systems (1/2)

Type Description Examples
Supervised The data set of examples provided to the machine includes the solutions desired outcome, a.k.a "labeled training data." Systems for classification tasks and
predictions, such as:
  • Linear regresion
  • Logistic regresion
  • K-nearest neighbors
  • Decision trees and random forests
  • Support Vector Machines (SVMs)

What are these about? Delve into the examples

Unsupervised The data set of examples provided to the machine does NOT include the solutions desired outcome. The machine produces its own solution to the problem task, a.k.a "unlabeled training data." Clustering:
  • K-means
  • Hierarchical Cluster Analysis (HCA)
  • Expectation Maximization
Visualization and dimensionality reduction:
  • Principal Component Analysis (PCA)
  • Kernel PCA
  • Locally-Linear Embedding (LLE)
  • t-distributed Stochastic Neighbor Embedding (t-SNE)
Association rule learning:
  • Appriori
  • Eclat

What are these about? Delve into the examples


Type of Machine Learning Systems (2/2)


Type Description Examples
Semisupervised The set of data examples provided to the machine contains a mix of labeled and unlabeled data examples with solution, and examples with no solutions. A combination of supervised and unsupervised systems, such as:
  • Deep Belief Networks (DBNs)

How does a "Deep Belief Network" function? Research and understand.

Reinforcement Machine actions have rewards or penalties in return. The machine must find the best strategy to maximize rewards over time get the maxium amount of rewards.
  • Markov Decision Process

How does a "Markov Decision Process" work? Research and understand.