In normal times, the UTCS Neural Networks Research Group meets bi-weekly in one of the GDC seminar rooms. There are four types of meetings: (1) people in the group present and get feedback on their work, (2) we read and discuss recent seminal papers on neural nets, (3) students give practice talks for conferences, PhD proposals, and defenses, and (4) visitors give informal presentations to the group, or more formal departmental talks.
We usually have an organizational meeting in the beginning of each semester where we schedule the meetings for the rest of the semester. Each student is usually responsible for one of the meetings, either presenting his/her own research, or championing a paper. Here is the schedule for then NNRG calendar listing upcoming events for the current semester, and a link to a comprehensive list of past meetings.
Each student should give a presentation about their work at least once a year. It is best to do it at the time when it is most useful to get candid feedback on the research direction to take, or when there's some problem that could use some fresh ideas. You should volunteer to talk when you feel it would be useful.
Prepare your talk like you would any other talk (see the page on giving presentations), but keep it under about 30 minutes so that we will have another 30 minutes for discussion. These talks are very informal. The point is to brainstorm together, so it is best to concentrate on future directions, research problems, and new ideas rather than completed work. You will get more out of it that way, and it is more fun for everyone.
You should also select a paper as background reading for the group. It could be your own paper, or one that inspired your research, or one that otherwise gives a good background for what you will be talking about.
In addition to a presentation of your own work, you can also choose to "champion" a paper (or papers) that you find to be particularly interesting. You should prepare a few slides and give a short summary to get the discussion going. However, unlike in a research talk, a "champion" is not expected to impart the full content of a work, but instead can assume that the participants are familiar with the work. You can be more evaluative: Give your opinion (and rationale) on what ideas from the work are right, wrong, important, and/or instructive, especially as such ideas relate (or might relate) to your work (or the work of anyone in our group).