Doubt about comparison of Machine Learning algorithm

Question

I read this article about the comparison of Machine Learning algorithm.

According to this article, there is some characteristics that define the ML models performance. I know that there are some models define lazy and some define eager.

Looking at the Figure in the article, it is possible to define what characteristics define a models lazy in spite of eager? This characteristic is speed of learning or speed of classification?

Answer 1

Disclaimer: I didn't read the paper.

• This article is apparently a preprint, i.e. it didn't go through the peer reviewing process required to be published in a reputable journal or conference (or it was rejected). The content might be correct, but nobody checked.
• The paper is about a very specific task "DIAGNOSIS OF NETWORK PERFORMANCE ISSUES IN CLIENT-TERMINAL DEVICES", the experiments and conclusions apply only to this task and cannot be generalized to any ML problem.
• The diagram showing the characteristics of different learning algorithms is even more specific, because there's no standard way to measure these characteristics. So the authors used their own definitions, which might or might not be truly measure these characteristics in the specific case of their experiment. It would be very risky to generalize anything from this figure imho.

See this question about eager vs. lazy learning. It is correct that the figure shows two characteristics related to this:

• speed of learning is about the duration of training
• speed of classification is about the duration of testing, i.e. applying the model

As mentioned in the linked question, a lazy learner just stores the training data. This is the case for kNN only in the figure: short training, long classification time. An eager learner (all the others) extracts parameters during the training stage, so it has longer training time and shorter classification time.