Learning Problem and Techniques
In the Machine Learning teatre, we can think of two entity:
- The teacher who has domain knowledge
- The learner is the entity who has to learn the knowledge to perform the task
We can distinguish learning strategies by the amount of inference the learner performs on the information provided by the teacher
Taxonomy of Machine learning
Rote Learning
Knowledge is directly implanted in the learner, either by directly programming or by memorizing given facts and data with no inferences (like from a database). There is no inference and all the work is made by the teacher.
Learning from instructions
Acquiring knowledge from a teacher or other organized source, such as textbook. This requires that the learner transform the knowledge input from the input language to an internal representation.
The learner is required to perform some inference, but a large fraction of the cognitive burden remains with the teacher, who is required to presente and organize knowledge in a way that incrementally increases the learner’s actual knwoledge.
Learning by analogy
Consists of acquiring new facts or skills by transforming and increasing existing knowledge that bears strong similarity to the desired new concept or skill into a form effectively used in the new situation.
Requires more inference on the part of the learner that does rote learning or learning from instruction.
Learning from examples
Given a set of concrete examples, the learner induces a general concept description that describe the examples.
The amount of inference performed by the learner is much grater than in learning from instruction and in learning by analogy. This is the most popular machine learning technique.
What is the learning problem?
Can be described as finding a general rule that explains data given only a sample of limited size.
The difficulty of this task is similar to the problem of children learning to speak from the sounds emited by the grown-up people.
Learning Techniques
There are four big families: