Basic Information

• Project Title: Revenue Optimization in Network-Based Auctions
• Name: Parvik Dave
• Project: BTP
• Semester(s): 7,8
• Guide: Prof. Swaprava Nath

Abstract

The auction of a single indivisible object is one of the most studied problems in mechanism design. Myerson gave a fundamental result that defined incentive-compatible auctions and revenue optimality. However, when information is spread via agents in a network, the result does not hold. This is because the agents need to be incentivized to spread the information. Recently, a few mechanisms have been proposed that incentivize buyers to forward auction information in their networks. This work involves investigating the properties of a new class of mechanisms based on a Myerson-like characterization of incentive-compatible auctions on a network. In particular, a special subclass called referral auctions is considered, inspired by the multi-level marketing mechanisms, and obtain the structure of the revenue optimal referral auction for i.i.d. bidders.

Any courses you completed relevant to the project

CS6001, CS6002

Describe your experience on the project

I’d explored game theory as part of SoS during freshie year, and then didn’t really do much about it. But I wanted to do something related to this, and so this came naturally when I was looking for a project for my BTP. I approached Prof. Swaprava Nath to work on something in this field. Since I’d already covered significant ground beforehand through my SoS project, we got started almost immediately as part of BTP-I. We were working on finding a characterization for all truthful auction mechanisms in the case of diffusion auctions. The work during BTP-I was mostly theoretical, except for some simulations that were performed to test out our designed mechanisms. We went on to submit a paper at a conference for this, the results of which are still awaited. As part of BTP-II, I continued the work on finding the optimal mechanism and extending our work. This involved LP to solve optimal mechanisms for smaller networks, and then building the intuition to create a rigorous proof for the structure of optimal mechanisms. On a tangent, I also started working on multi-object network based auctions, which is a direct extension of single-object network based auctions, during summers. The work was enjoyable, and didn’t get too intensive, about 8-10 hours of work apart from weekly meetings.

Describe your experience with the guide

Prof. Swaprava Nath likes to be involved in all his projects, which is why he personally worked along with me on what I was working. We divided the work and we would have weekly meetings where we’d discuss my progress and for some questions that needed discussion. Sir is very approachable. He replies fairly quickly in case of any doubts, but I believe some digging by oneself is recommended. He is also very perceptive to what you have to say and will factor in your feedback/opinion on any matter. The evaluation was through the presentation at the end of the semester. I was required to prepare a detailed report summarizing my work, my findings, literature review, etc. and also prepare a presentation that explains my work to the external examiner.

Any advice for anyone considering a project under the same guide? Any other professors working in similar fields?

One piece of advice would be to be pro-active. Push for meetings when you want to discuss something, and proactively update the prof on your work.