By Zainab Khan and Jaspreet Kaur

In our project for the quarter, we studied MIMO channels with polar-encoded inputs, and the process of decoding them using conventional and neural network-based techniques (abstract at the end). While as a whole this was much too involved of a project to explain to kids, we hoped to explain the process of encoding a message, transmitting it across a noisy channel, and then decoding the message. We translated the concepts involved in the communication pipeline into a kid-friendly medium: invisible ink and birthdays!

“Do you want me to guess your birthday month?” We open with this to invite the children to something they are very excited about: when they next get a year older.

We provide a constellation chart with drawings of the 12 Western horoscopes in constellation format: Aries, Taurus, etc. Children get to use the concept of a lookup table by finding their birthday in the provided birthday ranges, and translating that into a constellation.

Now they have encoded their message, the birthday range, into a constellation. (If we consider the birthday as the message, this is a lossy encoding, but we would never use this encoding in practice to encode a birth-date.)

This encoding is easily accessible to children: the 12 different constellations look different enough that even when they struggled drawing (it’s hard to draw with invisible ink!), the message was recognizable. In this way, we can see that the horoscope constellations are a type of channel encoding, since they allow for message recovery even in the presence of noise.

Finally, we invited kids to trade with others visiting the table, or trade with us, so that someone could decode their message. We demonstrated that even though the picture was not always close to the original picture, the original message was retrievable.

The invisible ink serves as the encoder (and fun!), the white paper serves as the noisy channel, and the black light as a decoder.

By using the Invisible Birthday Channel activity, we translated the conventional communication diagram into something that kids could intuitively get.

As we previously mentioned, the Invisible Birthday Channel is a simplification of the general communication problem. Our main project abstract is below:

There are a lot of interesting problems at the intersection of machine learning and information theory. The field of communications uses important Information Theory concepts to maximize the amount of information communicated without losing information. There has been an increased interest in using Neural Networks for solving channel based communication problems. This quarter, we looked at the problem of decoding messages which were polar-encoded and transmitted in a Multiple Input Multiple Output (MIMO) system. In this paper, we explore the benefits of polar coding, MIMO systems, and joint channel decoding. Additionally, we explore the use of a six layer Deep Neural Network (DNN) for MIMO systems with equal number of transmit and receive antennas. The DNN performs joint channel detection and decoding and we compare our results against a linear MIMO receiver.

The link to our full report is here.