A Mathematical Theory of Communcation

By Claude Shannon. In these notes, we will go over this paper. I hope to provide summaries to new points presented, as well as provide some of my own thoughts as they happen. My own thoughts could be wrong, as I am just brain storming so please let me know if you disagree with one of my points. Please keep in mind that this is an old paper, and some of the examples that it presents might be a little outdated (the theories should still stand.)

Introduction

Discrete Noiseless Systems

The Discrete Noiseless Channel

we will be looking at how to measure the capacity of a discrete system to be able to transmit information.

What defines a discrete system? - a message is a sequence that comes from a finite set of symbols. - eg. a bit, or dot/dash in telegraphy - sequence across time. - each element in the sequence may take varying amounts of time. - eg. dot takes 1 timestep, dash takes 3. - not all combinations of sequences need to be accounted for, - eg. if we want representations for 3 things, we won’t need the complete set of possible sequences.

How do we define the capacity of a discrete channel? We need a general way to quantify sequences of different length and time. We can measure the capacity as the amount of information sent per unit of time.

Remember that the amount of information is quantified by the amount of uncertainty solved. The total number of signals/sequences that we could be sending in a given amount of time is what we are uncertain about. Normalizing this with time would give us a general measure of the information in the given time.

TBD