search

carpentries-incubator / ml4bio-workshop

Materials for a workshop introducing machine learning to biologists
https://carpentries-incubator.github.io/ml4bio-workshop/
Other
21 stars 9 forks source link

Defining literacy #49

Open agitter opened 5 years ago

agitter commented 5 years ago

We will work here to define what we mean by "machine learning literacy".

cmilica commented 5 years ago

some important literacy questions could be: How accurate is the model? When is it most accurate? When is it not so accurate? What pipeline did it follow?

agitter commented 5 years ago

One set of literacy concepts would be those taught in machine learning courses. At UW-Madison, that includes CS 540 and CS 760. They are aimed at a different, more technical, audience, but there are some relevant ideas. For instance, the first 760 lecture discusses when a trained model will generalize. It also defines the course goals as:

  1. Understand what a learning system should do
  2. Understand how (and how well) existing systems work

The questions @cmilica posted above seem relevant to me. The part about when a model is or is not accurate relates to generalization and the data distributions under which the model is expected to perform well.

Literacy likely also relates in part to being comfortable with some core vocabulary. We could look for other machine learning glossaries or create our own that links to the definitions in our guides.

dltreu commented 5 years ago

Would "literacy" also include gaining the vocabulary and understanding of machine learning, sufficient to discuss a data-related research problem with a computationalist?

dltreu commented 5 years ago

What about identifying a biologist who has used machine learning or worked with a computationalist and inviting them to give their perspective, review our materials, or somehow join the discussion?

agitter commented 5 years ago

Would "literacy" also include gaining the vocabulary and understanding of machine learning

Yes, I'd say that is a major part of literacy. We should think about ways to bring more attention to vocabulary and assess what terms are being retained. Your prior suggestion to ask participants to explain concepts to a friend will help with this.

What about identifying a biologist who has used machine learning

We should be able to find many examples of this. Maybe I could contact some individuals who may be willing to talk to @cmilica about their experiences?

cmilica commented 5 years ago

I would be definitely ok with that. Before the meeting, it would be good to have specific questions we want to ask and focus on.

agitter commented 5 years ago

In our meeting today we created an initial list of terms. @cmilica can create a new document where we discuss and finalize that list.