8
$\begingroup$

It's very common for users to want to find about a specific topic in machine learning, statistics, or probability. Often, these questions start from a place where the user doesn't know what they want to know, so it is challenging to give a good answer.

  • A vague question - a question with little context or explanation of what kind of information is sought - is very likely to be closed as "Needs details or clarity."
  • A question that asks about a broad or general topic (e.g. "computer vision" or "machine learning") is likely a duplicate.

So how should a user who does not know what they want to know ask for references (books, articles, tutorials, websites) about an area with which they are unfamiliar?

$\endgroup$

1 Answer 1

10
$\begingroup$
  1. Search. The first thing to do when asking any question, but especially a question, is to do a site search. At this point, the site has a robust collection of questions providing references for a rich variety of topics. A comprehensive outline of how to search the site can be found in FAQ: Best Practices for Searching CV

  2. Specificity. If you've done a search and have not found anything that is specific to what you want to know, then the next step is to articulate what makes your question distinct from similar reference requests you've already reviewed (if there are any relevant questions).

    1. If you're asking about a specific aspect of a topic, please clearly state what that is. For example, if you're asking for references about "computer vision" or "neural networks" or "machine learning" (without elaboration), then it's very likely to be closed as a duplicate of existing threads. But a question specific to a certain aspect of a topic may not be a duplicate, especially if you can be explicit about what aspect is not covered.

    2. If you're looking for "more recent" or "more current" literature on a specific topic, then you're actually hot on the trail of finding what you need.

      1. If you found a thread containing relevant but older references in (1), then check whether the authors have published a newer edition. It often happens that authors will write a high-quality book and then revise and expand it at a later date for precisely the same reasons that you want a more recent text. The authors are acutely aware that there's more to know about their topic, so they've obligingly expanded and updated their text.

      2. Even if a newer edition of an older reference doesn't exist, it's easy to find books and texts published after a certain date. For example, Amazon has an advanced book search that allows users to specify subject matter, keywords, and publication date (among other fields). Carefully using these filters can turn up a promising crop of books published in the past several years. Reading over the tables of contents gives overview of their contents. Google Scholar also has date filters.

      3. If the older literature doesn't address your question and there are no updated texts, then you need to be specific about what you want to learn from a newer text. Other than the publication date, what distinguishes a more recent resource from an older one? Why is recency important, and how does recency relate to what you want to know about?

  3. Literature Review. If you can answer the questions in (2), then you have all the material you need to write a good question. On the other hand, if find yourself thinking "But I don't know what I don't know!" or if you don't know how to answer (2.1) or (2.2), then you need to do a literature review to understand what topics exist. I've found that the best way to do this is to do a Google Scholar search for "review", "survey" or "overview" articles. Google Scholar has a special filter for review article. Also, Google Scholar supports the usual Google search syntax, so you can do things like exact phrase matching, use boolean operators, and so on. Additionally, Google Scholar lets you filter results by publication date, so you can focus on recent articles. Finally, Google Scholar also lets you search within the resources that cite an article, which can be helpful if you want to know about what people have written about a topic since the publication of the document.

Finding good references on a topic is an iterative process. Once you've reached (3), you should be able to identify the key terms that you want to know more about. Now you can return to (1) and search CV for these topics, to find anything that wasn't turned up the first time around. If you've done (1) and not found what you need, then you can proceed to (2) and write a specific question that is explicit about what you what to know about and explains why any similar threads are not responsive.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .