Resolved: Minimise or maximise the objective function
In video Machine Learning(ML) Techniques it says objective function is a measure how far it is from the target. One would thought minimise the objective function should be good as it will reduce the distance from the target. However, in video Machine Learning(ML) Types of Machine Learning, when talked about reinforcement learning, it says to maximise the objective function. So should we minimise or maximise the objective function?
Hi Josh!
Thanks for reaching out.
In fact, you've guessed it. Depending on the type of machine learning applied, the objective function should be minimised or optimized.
When we are doing supervised or unsupervised learning, we are trying to be closer to the target; in other words, we'd like to minimize the distance away from the target. That's when we'd like to minimize the objective function.
When we are referring to reinforcement learning, we'd like to reinforce our algorithm, so we'd like to maximize the related reward. Then, we are maximizing the objective function.
Hope this helps.
Kind regards,
Martin
The question regarding the primary motive relating the reward/reinforcement learning with objective function (O.F.) is not in accordance with the module video. As per my beginner knowledge, the arrow example given relates to supervised/unsupervised learning, the O.F. must be minimized which increaes incase of reinforcement learning. Please check into this, as most answers will be wrong like I did.
Hi UTKARSH!
Thanks for reaching out.
In reinforcement learning, we use the same equation structure, but the model is different. In that sense, the objective function is intended to be maximized in reinforcement learning, not in (un)supervised learning.
Does this clarify your question or did you mean something else, please? Thank you.
Hope this helps.
Kind regards,
Martin
Well, mine was wrong too! However, it's important to read these comments and the explanation to have further understanding. I'm also "new" to all this, doing a career transtion path, and noticed some quizzes from this optional course in particular have been slightly misleading, in my perspective.
Hi Erick!
Thanks for reaching out and sharing your experience and opinion!
We are following the evolution of this course closely since we regard it actually as fundamental more than as an optional one. Meaning, the explanations provided in it are essential for anyone who wants to work in the data science field.
Then, we agree that many of the terms are confusing, particularly because today people use them with slighlty different applications or meanings. Nevertheless, we have tried to provide challenging questions that will ensure your good understanding of the matter in general. Then, I hope that as you advance in our Program (regardless of the career track chosen), the topics discussed in this course will be gradually becoming clearer.
Hope this helps but please feel free to get back to us should you need further assistance. Thank you.
Kind regards,
Martin
