Introduction to Level 4 by Tony Eng
Video Transcript: When I designed this course, I wanted to expose you to a range of different audiences. And so this week, we’re going to look at a learner audience– for example, a student in an undergraduate class that you might teaching. You’re going to pick something technical to explain to them. Don’t fret so much over what it is that you choose, it just needs to be something basic that you understand. The more important thing is figuring out how to explain it. More details are in assignment as well as an example.
Pick something technical (e.g. an algorithm, a process, a phenomenon, a technique, etc) related to the background of your research, and explain how it works to an audience of undergraduates (who are perhaps taking an undergraduate course in which you are teaching this material). Especially when describing something complicated, it is often useful to first give the big picture or a general overview before jumping into specific details. It is this high-level introduction that you should turn in. Your submission should be in the form of a video that is 3 minutes or less. The camera can be trained on you only, or on you at a whiteboard, or on a piece of paper that you write on with your voice in the background, etc.
Complete The Following Steps
As a side note, giving an overview is useful not just for teaching, but also in your writing. Here’s the corresponding version excerpted from a thesis:
“Simulated annealing is a stochastic means for finding the minimum scoring state of a system through gradual descent[KGJV83, BT92, PTVF95]. It finds its origin in the thermodynamic process of cooling. When a liquid freezes, it crystallizes; if the temperature is reduced gradually, the system settles into its minimum energy state and the crystal formed is pure. With simulated annealing, the algorithm starts at some initial temperature, and with each iteration, this temperature is gradually lowered. Some starting sequence is randomly selected, and the search space is explored by making random modifications, called moves, first to the initial sequence and then to all subsequent sequences. Any move resulting in a better scoring sequence(a lower energy state) is immediately taken. Any move resulting in a higher energy state is allowed with some probability that depends on the current temperature. These uphill events allow for the search to escape local minima, but they are less likely to occur as the algorithm progresses. Thus, the start of the algorithm is essentially a random exploration of the space as long range moves have a higher probability of being taken, while more fine tuning is done at lower temperatures. It has been shown that if the cooling schedule is slow enough, then the algorithm will converge on the global optimum with high certainty[Haj88].”
Peer Feedback Instructions
Find 2-3 friends and perform the lecture in front of them. Ask them the following questions:
Regarding the explanation
- Was the explanation understandable? Were there any areas that were confusing or unclear?
- What can the audience infer of the material you presented as a result of the explanation?
- Ex: “If I understand you correctly, this means that I could in fact move to the optimal answer early on and move away from it with some probability?”
- Was the pacing appropriate?
- Was the presenter “repetitive without being repetitive”?
- Did the presenter give the audience time to absorb the material and mentally catch up?
- If the presenter used any props (such as a whiteboard, models, etc), were they used in a way that was helpful in understanding the presenter’s explanation?
- Did the presenter utilize nonverbal cues appropriately?
- The presenter appeared comfortable and confident
- The presenter made eye contact with the audience
- The presenter spoke from stillness, with open posture
- The Presenter conveyed you are passionate about the material
- Did the presenter use any filler words?
- Meaningless words like “um”, “you know”, “like”, “uh”, etc.
- Was the presenter’s voice clearly understandable?
- (i) good volume
- (ii) good enunciation
- (iii) good pacing (not too fast and not too slow)
- Did the presenter maintain an appropriate connection with the audience?
- The presenter included a narrative to make the content relatable to the audience and/or maintained a conversational tone in addressing the audience.
- Does the presenter have any constructive and specific feedback for improvement?
Finished Model 1
Here is a sample transcript of an example for “simulated annealing” for an undergraduates in Computer Science:
“Given a problem to solve and a space of possible answers, one could use an exhaustive search strategy — just try every single possible answer and find the best one. But for many problems, it would be computationally infeasible to do so because there are too many possible answers and it would simply take too long to try them all. So computer scientists have come up with various ways to cope. One way is to perform a limited guided search of the solution space. Instead of considering all solutions, only consider (or “move to”) a subset of them that are chosen probabilistically according to a set of guiding rules. Here’s the idea: you pick a random solution in the space, then you use your guiding rules to generate a second solution. If the latter is a better solution, you move to it; if not, you move to it with some probability that decreases over time, else you stay where you are. This allows you to avoid getting stuck in local minima. Then you generate the next possible solution, and continue in this manner until you either reach a solution that is “good enough” or until you have explored a pre-set number of solutions (in which case you can return the best answer you encountered). So in the beginning, you jump around from solution to solution, always moving to a better answer if you happen upon on, but as time goes on, fewer and fewer jumps occur because the probability you move to a worse solution decreases. This technique is called simulated annealing and researchers have shown that if the probability is reduced slowly enough, this guided search converges on the optimal solution.”
Notice how we don’t delve immediately into a formal definition or equations, but take the time to explain what is going on a high level, and we try to minimize any intellectual gaps that are too big. In terms of jargon, while there is some (e.g. “local minima”, “exhaustive search”), these terms should already be familiar to a computer science audience, and we didn’t even use any terms specific to simulated annealing to minimize cognitive load. Lastly, the sentence “So in the beginning, … decreases” gives the overall intuition behind how a simulated annealing approach works.
Finished Model 2
Finished Model 3
Finished Model 4
Case Study: Katrina LaCurts On Explaining The Technical When Teaching
WARNING: This video serves as a demonstration of the underlying ideas from this exercise applied to a (then) current MIT graduate student. The video is a little long. While it isn’t necessary to watch this video in order to progress through the next level, it is encouraged.