Gaurav's Blog

cat-ting Binary Files

14 Sep 2012 - less than 1 minute read

I wanted something that would dump a binary file on the terminal in hex. Turns out there is a tool called hexdump!

hexdump filename

Voila!

Read more

Giving Awesome Talks

21 Aug 2012 - 2 minute read

Face it, giving talks seems like a performing art to some extent. Making sure you retain everybody’s attention throughout the talk, is easier said than done. Last semester Prof. Bender gave us some tips on how to give an awesome talk without having the charisma of <Insert-your-favorite-heartthrob’s-name-here/>.

Deepak has very carefully jotted down some of the tips in his blog post.

Edit (August 2025): The link above doesn’t work anymore for some reason. Regardless, these are some of Prof. Bender’s tips contributed by Deepak, Dhruv, and yours truly, in our own words, and based on what we recollect (so there might be errors). We hope to collect more in the future.

Content

  1. Most important - most interesting content first (e.g. the experiment results or the findings) - attention span is highest in the beginning. So keep the content prefix-competitive.
  2. Tell people upfront what they will learn by the end of the presentation (the one or two or three things), and then start delivering that content immediately. Too many people start with an agenda slide, or a long introduction before getting to the most interesting content at the end.
  3. Explain why the problem is important. Even if the audience doesn’t understand the solution, they should understand why do we need a solution in the first place. I (Dhruv) find that Dr. Dan Gusfield does a brilliant job of motivating problems before presenting the solutions.
  4. Know your audience. For example, presenting the workings of a toaster to a homemaker is different from presenting it to an electrical engineer. You’ll need to motivate the problem differently in both cases. Same goes with presenting it to someone who has had a toast before v/s someone who hasn’t.
  5. Make use of plots effectively. Explain the axes, and know what to plot
  6. Give credit wherever possible, and do this in the beginning.

While Presenting

  1. Start with humor / joke, but for a technical talk, don’t over do the humor. The purpose of humor is to engage the audience at the beginning, so you can deliver the most interesting content at the beginning.
  2. Use the stage effectively (don’t just stand behind the podium), walk over to the big screen and point to specific things from the slides. It creates a psychological impact that the speaker knows what they’re talking about, and the audience tends to absorb / remember specific slides better that way.
  3. Refer to everything in the slides, don’t skip over content or slides or diagrams - conversely, if your slides are too busy, it’s probably not easy for you to refer to everything, so make slides less dense, or with little text. If you skip content, it subconsciously tells the audience that the content is not important and dilutes the overall impact.
  4. Don’t just read text from the slides, paraphrase if possible.
  5. Maintain a steady pace (1.2x speed), suggest that you will entertain questions at the end.
  6. Gaze / look at all sides of the room while delivering the talk.
  7. End on time, leave room for questions.

Hope this is helpful in your future talks.

Read more

Graphs in Talks: How to present curves?

13 May 2012 - less than 1 minute read

Prof. Michael Bender showed us a neat trick when presenting graphs in talks.

Often, we do experiments with Data-Structures and Algorithms, and we want to show that the cost of doing blah with the foo bar data structure, is $O(n)$ (which means that if we plot the actual cost against the value $n$, we should get a straight line).

Okay. What intuition do you have for something which costs $O(n\log{n})$. I would say its slope is somewhere between a line and a parabola. But how does the audience visually verify this instantly when seeing the graph?

On the y-axis, you do not plot the actual cost, but plot the ratio of the actual cost to the value that you expected, and show that the ratio is constant.

This is how we plotted something that we expected to be $O(n\log^2{n})$:

moves

It was perfectly clear that the ratio is almost constant, and hence our hypothesis was correct.

Read more