{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true,
"tags": [
"remove_cell"
]
},
"outputs": [],
"source": [
"# HIDDEN\n",
"from datascience import *\n",
"from prob140 import *\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"plt.style.use('fivethirtyeight')\n",
"%matplotlib inline\n",
"import math\n",
"from scipy import stats\n",
"from scipy import misc"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Reversibility ##"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The reflecting random walk of the previous section has states 0, 1, 2, 3, 4 arranged in sequence clockwise on a circle. At each step the chain stays in place with probability $s$, moves to its clockwise neighbor with probability $r$ and to its counterclockwise neighbor with probability $p$. The stationary distribution of the chain assigns chance 0.2 to each state.\n",
"\n",
"If $r > p$, then the chain is more likely to be moving clockwise than counterclockwise. For example, in steady state, the probability of the path $0, 1, 2, 3$ is\n",
"\n",
"$$\n",
"P(X_0 = 0)P(0, 1)P(1, 2)P(2, 3) = 0.2r^3\n",
"$$\n",
"\n",
"The probability of the *reversed* path $3, 2, 1, 0$ is\n",
"\n",
"$$\n",
"P(X_0 = 3)P(3, 2)P(2, 1)P(1, 0) = 0.2p^3\n",
"$$\n",
"\n",
"If $r > p$, then the original path has higher chance.\n",
"\n",
"But if $r = p$, the chance of the original path is the same as that of the reversed path; in steady state, the chain is just as likely to be running in either direction. If someone simulates the chain in steady state and shows you the original path as well as the reversed path, you will not be able to tell which is which.\n",
"\n",
"In this section we define what it means for a Markov Chain to be *reversible* in this way."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Reversed Process ###\n",
"Let $X_0, X_1, \\ldots $ be an irreducible Markov Chain with a finite state space and stationary distribution $\\pi$. Start the chain off with this stationary distribution; that is, let $X_0$ have distribution $\\pi$. Then for all $n \\ge 1$, the distribution of $X_n$ is also $\\pi$.\n",
"\n",
"Fix $n > 0$ and consider the *reversed* sequence $Y_0, Y_1, \\ldots, Y_n$ defined by $Y_k = X_{n-k}$ for $k = 0, 1, \\ldots, n$. Call $X_0, X_1, \\ldots, X_n$ the *forwards* sequence.\n",
"\n",
"It is a wonderful fact that the reversed sequence is a time homogenous Markov Chain. To see why, we will check that the Markov property holds.\n",
"\n",
"Before we prove the general fact, let's make some exploratory calculations. Start with $n = 1$, so that $Y_0 = X_1$ and $Y_1 = X_0$. For states $i$ and $j$.\n",
"$$\n",
"\\begin{align*}\n",
"P(Y_1 = j \\mid Y_0 = i) ~ &= ~ \\frac{P(Y_1 = j, Y_0 = i)}{P(Y_0 = i)} \\\\\n",
"&= ~ \\frac{P(X_0 = j, X_1 = i)}{P(X_1 = i)} \\\\\n",
"&= ~ \\frac{\\pi(j)P(j, i)}{\\pi(i)}\n",
"\\end{align*}\n",
"$$\n",
"because the forwards sequence is in steady state. We have found a transition probability for the reversed sequence using the transition matrix and stationary distribution of the forwards sequence.\n",
"\n",
"For $n = 2$, we have $Y_0 = X_2$, $Y_1 = X_1$, and $Y_2 = X_0$. For states $k$, $i$, and $j$,\n",
"$$\n",
"\\begin{align*}\n",
"P(Y_2 = j \\mid Y_0 = k, Y_1 = i) ~ &= ~ \\frac{P(Y_2 = j, Y_1 = i, Y_0 = k)}{P(Y_1 = i, Y_0 = k)} \\\\\n",
"&= ~ \\frac{P(X_0 = j, X_1 = i, X_2 = k)}{P(X_1 = i, X_2 = k)} \\\\\n",
"&= ~ \\frac{\\pi(j)P(j, i)P(i, k)}{\\pi(i)P(i, k)} \\\\\n",
"&= ~ \\frac{\\pi(j)P(j, i)}{\\pi(i)}\n",
"\\end{align*}\n",
"$$\n",
"The answer doesn't depend on $k$. That's consistent with the Markov property. Also, put together the two facts we have just proved to notice that the transition probabilities are time homogenous.\n",
"\n",
"For general $n$, fix states $i$ and $j$ and an integer $m$ in the range 0 through $n-1$.\n",
"$$\n",
"\\begin{align*}\n",
"& P(Y_{m+1} = j \\mid Y_0 = i_0, Y_1 = i_1, \\ldots, Y_{m-1} = i_{m-1}, Y_m = i) \\\\ \\\\ \n",
"&=\n",
"\\frac{P(Y_0 = i_0, Y_1 = i_1 \\ldots, Y_{m-1} = i_{m-1}, Y_m = i, Y_{m+1} = j)}\n",
"{P(Y_0 = i_0, Y_1 = i_1 \\ldots, Y_{m-1} = i_{m-1}, Y_m = i)} \\\\ \\\\\n",
"&= \\frac{P(X_n = i_0, X_{n-1} = i_1, \\ldots, X_{n-m+1} = i_{m-1}, X_{n-m} = i, X_{n-m-1} = j)}\n",
"{P(X_n = i_0, X_{n-1} = i_1, \\ldots, X_{n-m+1)} = i_{m-1}, X_{n-m} = i)} \\\\ \\\\\n",
"&= \\frac{\\pi(j)P(j, i)P(i, i_{m-1}) \\cdots P(i_1, i_0)}\n",
"{\\pi(i)P(i, i_{m-1}) \\cdots P(i_1, i_0)} \\\\ \\\\\n",
"&= \\frac{\\pi(j)P(j, i)}{\\pi(i)}\n",
"\\end{align*}\n",
"$$\n",
"This involves only $i$ and $j$, and not on $i_0, i_1, \\ldots, i_{m-1}$ nor on $m$. So the Markov property is satisfied and the transition probabilities are time homogenous. The one-step \"$i$ to $j$\" transition probability for the reversed sequence is\n",
"\n",
"$$\n",
"P(Y_1 = j \\mid Y_0 = i) = \\frac{\\pi(j)P(j, i)}{\\pi(i)}\n",
"$$"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Reversible Chains ###\n",
"The original \"forwards\" Markov Chain $X_0, X_1, \\ldots $ is called *reversible* if for every $n$, the reversed sequence $Y_0, Y_1, \\ldots Y_n$\n",
"has *the same one-step transition probabilities as the original*; that is, if\n",
"\n",
"$$\n",
"\\frac{\\pi(j)P(j, i)}{\\pi(i)} = P(i, j) ~~~ \\text{for all } i, j\n",
"$$\n",
"\n",
"That is, the chain is reversible if\n",
"\n",
"$$\n",
"\\pi(i)P(i, j) = \\pi(j)P(j, i) ~~~ \\text{for all } i, j\n",
"$$\n",
"\n",
"In other words:\n",
"\n",
"**The chain is reversible if the detailed balance equations have a positive solution.** This is consistent with our image of particles moving according to this chain in steady state: at each instant, the proportion of particles moving from $i$ to $j$ is exactly the same as the proportion moving from $j$ to $i$, for every pair of states $i$ and $j$. "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"At the start of this section we looked at a random walk on a circle. Let's see what the definition of reversibility implies for this chain.\n",
"\n",
"- In the previous section we showed that when $p \\ne r$, the detailed balance equations have no positive solution. Therefore, when $p \\ne r$, the chain is not reversible. This is consistent with our earlier analysis.\n",
"\n",
"- When $p = r$, we found a solution to the detailed balance equations, and therefore the chain is reversible. This formalizes our idea that if $p = r$ then in steady state the chain \"looks the same run forwards or backwards.\""
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Reversibility of Birth and Death Chains ###\n",
"Recall that a *birth and death chain* is a Markov Chain on the integers, with one-step transitions restricted to going up by 1, going down by 1, or staying in place. It is not hard to check that every irreducible birth and death chain with a finite state space is reversible. You can simply solve the detailed balance equations just as we did for the Ehrenfest chain in the previous section.\n",
"\n",
"Go back and look through the examples in the text and exercises. The switching chains, the reflecting random walks (both lazy and not), both of the Ehrenfest chains, and the Bernoulli-Laplace chain are all irreducible birth and death chains, and hence are reversible."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's confirm this in the case of a birth and death chain which at first glance seems not to be reversible. Here is the transition diagram of a Markov Chain $X_0, X_1, \\ldots $.\n",
"\n",
"![B&D](trans_b_and_d.png)\n",
"\n",
"This chain moves right (that is, has births) with high probability, so it seems as though we should be able to tell whether it's moving forwards or backwards. But remember that **time reversal happens in the steady state**. In the steady state, the chain is overwhelmingly likely to be shuttling between states 3 and 4. You can see this by solving the detailed balance equations.\n",
"\n",
"$$\n",
"\\begin{align*}\n",
"\\pi(1)\\cdot 1 &= \\pi(2) \\cdot 0.1 ~~~~ \\implies \\pi(2) = 10\\pi(1) \\\\\n",
"\\pi(2) \\cdot 0.9 &= \\pi(3) \\cdot 0.1 ~~~~ \\implies \\pi(3) = 90\\pi(1) \\\\\n",
"\\pi(3) \\cdot 0.9 &= \\pi(4) \\cdot 1 ~~~~~~~ \\implies \\pi(4) = 81\\pi(1)\n",
"\\end{align*}\n",
"$$\n",
"\n",
"It will visit states 2 and 1 as well, but rarely, state 1 being particularly rare. These vists will intersperse the sojourns in 3 and 4, and the paths will be indistinguishable forwards and backwards.\n",
"\n",
"Let's simulate paths of this process. First, we construct the transition matrix and confirm our calculations of $\\pi$."
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"s = np.arange(1, 5)\n",
"\n",
"def trans(i, j):\n",
" if i == 1:\n",
" if j == 2:\n",
" return 1\n",
" else:\n",
" return 0\n",
" elif i == 4:\n",
" if j == 3:\n",
" return 1\n",
" else:\n",
" return 0\n",
" elif j == i+1:\n",
" return 0.9\n",
" elif j == i-1:\n",
" return 0.1\n",
" else:\n",
" return 0\n",
"\n",
"bnd = MarkovChain.from_transition_function(s, trans)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
"

Value

Probability

1

0.00549451

2

0.0549451

3

0.494505

4

0.445055

"
],
"text/plain": [
"Value | Probability\n",
"1 | 0.00549451\n",
"2 | 0.0549451\n",
"3 | 0.494505\n",
"4 | 0.445055"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"pi = bnd.steady_state()\n",
"pi"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can use `simulate_path` to plot a path of the chain. Notice that unlike our previous uses of this method, we are now passing an initial distribution as the first argument, not a particular state. The second argument is the number of steps, as before.\n",
"\n",
"The graph below shows one path of length 200. Run the cell a few times and look at each path forwards as well as backwards. You won't find a systematic difference between the two. "
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"