Basis Lab

Markov Chains

Chains That Look the Same Run Backwards

Last timeHow Long Until It Gets There

Some chains balance pair by pair rather than only situation by situation. When a chain does, its resting distribution can be read straight off the rule with no equations to solve at all.

Finding a resting distribution means solving a linear system. For a chain with

a million situations that is a serious computation, and for a chain whose rule

depends on a parameter you want to vary it has to be redone every time.

A large class of chains escapes this entirely. Their resting distribution can

be written down by looking at the rule. The property that makes it possible is

stronger than resting, and it has a physical meaning: the chain looks the same

run backwards.

A stronger kind of balance

The previous lesson asked that the total flow into each situation match the

total flow out. Nothing in that required any individual pair to match. The

flow could arrive from one neighbour and leave towards another, going round in

a circle, and the totals would still balance.

Detailed balance rules that out. It asks every pair to match on its own.

FIG 1The flow between a pair, matching in both directions
Compare this with the balance condition of the previous lesson, which was a sum over all of the partners of one situation. This has no sum in it. The practical difference is that the earlier condition is a linear system to be solved and this one is a collection of two term relations that can be chased through the structure of the chain, which is why it so often gives an answer by eye. Not every chain satisfies it, and the ones that do not are exactly the ones with a circulation.

Adding these up shows that detailed balance is enough. Fix a situation and add

its pair conditions over every partner. The left side becomes its share

multiplied by the total chance of hopping anywhere, which is just its share.

The right side becomes the total arriving from everywhere. So the distribution

is resting.

The reverse is false. A chain on three situations that hops clockwise nine

times in ten has the even split as its resting distribution, and sitting at that

distribution it still carries a steady clockwise circulation. The flow from one

situation to the next is nine times the flow back. Resting, and not reversible.

The lesson stops here

4 more paragraphs to go

You have read the opening. The rest of the argument, the problems that check whether it landed, and the lines worth keeping at the end all come with a plan.

The first lesson of every course in the library reads the whole way through, free, so you can see exactly what the rest of them are.

See the planThe contents

This is the reading half

Starting the course gives you your own copy of it. Every idea on every page has problems standing under it, marked with a reason rather than a tick, and any sentence you do not believe can be opened and argued with. None of that can happen on a page nobody owns.

The contents

The rest of this course

  1. 01Where You Are, and One Hop
  2. 02Where You Are After Many Hopsopening only
  3. 03What Is Reachable, and What Traps Youopening only
  4. 04The Distribution the Rule Leaves Aloneopening only
  5. 05Why the Chain Forgets Where It Startedopening only
  6. 06How Long Until It Gets Thereopening only
  7. 07Chains That Look the Same Run Backwardsyou are here
  8. 08Building a Chain to Reach a Chosen Distributionopening only

Read alongside