# When is a Markov chain a Markov chain?

I’ve been taking tutorials on the third quarter of the second-year probability course, in which the student have met discrete-time Markov chains for the first time. The hardest aspect of this introduction (apart from the rapid pace – they cover only slightly less material than I did in Cambridge, but in half the time) is, in my opinion, choosing which definition of the Markov property is most appropriate to use in a given setting.

We have the wordy “conditional on the present, the future is independent of the past”, which is probably too vague for any precise application. Then you can ask more formally that the transition probabilities are the same under two types of conditioning, that is conditioning on the whole history, and conditioning on just the current value

$\mathbb{P}(X_{n+1}=i_{n+1} \,\big|\, X_n=i_n,\ldots,X_0=i_0) = \mathbb{P}(X_{n+1}=i_{n+1} \,\big |\, X_n=i_n),$ (*)

and furthermore this must hold for all sets of values $(i_{n+1},\ldots,i_0)$ and if we want time-homogeneity (as is usually assumed at least implicitly when we use the word ‘chain’), then these expressions should be functions of $(i_n,i_{n+1})$ but not n.

Alternatively, one can define everything in terms of the probability of seeing a given path:

$\mathbb{P}(X_0=i_0,\ldots,X_n=i_n)= \lambda_{i_0}p_{i_0,i_1}\ldots p_{i_{n-1}i_n},$

where $\lambda$ is the initial distribution, and the $p_{i,j}$s are the entries of the transition matrix P.

Fortunately, these latter two definitions are equivalent, but it can be hard to know how to proceed when you’re asked to show that a given process is a Markov chain. I think this is partly because this is one of the rare examples of a concept that students meet, then immediately find it hard to think of any examples of similar processes which are not Markov chains. The only similar concept I can think of are vector spaces, which share this property mainly because almost everything in first-year mathematics is linear in some regard.

Non-examples of Markov chains

Anyway, during the tutorials I was asking for some suggestions of discrete-time processes on a countable or finite state space which are not Markov chains. Here are some things we came up with:

• Consider a bag with a finite collection of marbles of various colours. Record the colours of marbles sampled repeatedly without replacement. Then the colour of the next marble depends on the set you’ve already seen, not on the current colour. And of course, the process terminates.
• Non-backtracking random walk. Suppose you are on a graph where every vertex has degree at least 2, and in a step you move to an adjacent vertex, chosen uniformly among the neighbours, apart from the one from which you arrived.
• In a more applied setting, it’s reasonable to assume that if we wanted to know the chance it will rain tomorrow, this will be informed by the weather over the past week (say) rather than just today.

Showing a process is a Markov chain

We often find Markov chains embedded in other processes, for example a sequence of IID random variables $X_1,X_2,\ldots$. Let’s consider the random walk $S_n=\sum_{i=1}^n X_i$, where each $X_i =\pm 1$ with probability p and (1-p). Define the running maximum $M_n=\max_{m\le n}S_m$, and then we are interested in $Y_n:=M_n-S_n$, which we claim is a Markov chain, and we will use this as an example for our recipe to show this in general.

We want to show (*) for the process $Y_n$. We start with the LHS of (*)

$\mathbb{P}(Y_{n+1}=i_{n+1} \,\big|\, Y_n=i_n,\ldots,Y_0=i_0),$

and then we rewrite $Y_{n+1}$ as much as possible in terms of previous and current values of Y, and quantities which might be independent of previous values of Y. At this point it’s helpful to split into the cases $i_n=0$ and $i_n\ne 0$. We’ll treat the latter for now. Then

$Y_{n+1}=Y_n+X_{n+1},$

so we rewrite as

$=\mathbb{P}(X_{n+1}=i_{n+1}-i_n \, \big |\, Y_n=i_n,\ldots, Y_0=i_0),$

noting that we substitute $i_n$ for $Y_n$ since that’s in the conditioning. But this is now ideal, since $X_{n+1}$ is actually independent of everything in the conditioning. So we could get rid of all the conditioning. But we don’t really want to do that, because we want to have conditioning on $Y_n$ left. So let’s get rid of everything except that:

$=\mathbb{P}(X_{n+1}=i_{n+1}-i_n\, \big |\, Y_n=i_n).$

Now we can exactly reverse all of the other steps to get back to

$= \mathbb{P}(Y_{n+1}=i_{n+1} \,\big|\, Y_n=i_n),$

which is exactly what we required.

The key idea is that we stuck to the definition in terms of Y, and held all the conditioning in terms of Y, since that what actually determines the Markov property for Y, rearranging the event until it’s in terms of one of the underlying Xs, at which point it’s easy to use independence.

Showing a process is not a Markov chain

Let’s show that $M_n$ is not a Markov chain. The classic mistake to make here is to talk about possible paths the random walk S could take, which is obviously relevant, but won’t give us a clear reason why M is not Markov. What we should instead do is suggest two paths taken by M, which have the same ‘current’ value, but induce transition probabilities, because they place different restrictions on the possible paths taken by S.

In both diagrams, the red line indicates a possible path taken by $(M_0,M_1,\ldots,M_4)$, and the blue lines show possible paths of S which could induce these.

In the left diagram, clearly there’s only one such path that S could take, and so we know immediately what happens next. Either $X_5=+1$ (with probability p) in which case $M_5=S_5=3$, otherwise it’s -1, in which case $M_5=2$.

In the right diagram, there are two possibilities. In the case that $S_4=0$, clearly there’s no chance of the maximum increasing. So in the absence of other information, for $M_5=3$, we must have $X_4=X_5=+1$, and so the chance of this is $p^2$.

So although the same transitions are possible, they have different probabilities with different information about the history, and so the Markov property does not hold here.

# Gromov-Hausdorff Distance on Trees

This post continues the exposition of Gromov-Hausdorff distance, as introduced in Chapter Four of Steve Evans’ Probability and Real Trees, which we are reading as a group in the Stats Dept in Oxford at the moment. In this post, we consider applications of the Gromov-Hausdorff distance we have just introduced in the context of trees, viewed as metric spaces.

First we consider a direct application of the previous result, which related the Gromov-Hausdorff distance to the infimum of the distortion across the family of correspondences between the two relevant metric spaces. I found this as Corollary 3.7 in notes by Le Gall and Miermont on scaling limits of random trees and maps, which can be found here. I’m not clear whether there is an original source, but the result is simple enough that probably it does not matter hugely.

Proposition – Given f,g excursions above [0,1], and $T_f, T_g$ the real trees associated with these excursions in the standard way, then

$d_{GH}(T_f,T_g)\le 2 ||f-g||.$

Proof: We construct an appropriate correspondence

$\mathcal {R}=\left\{(a,b) \,:\, \exists t\in[0,1]\text{ s.t. } a=p_f(t), b=p_g(t)\right\}.$

In other words, the trees are defined as projections from [0.1] (with some equivalence structure), so taking pairs of projections from [0,1] gives a natural correspondence. Now, suppose $(p_f(s),p_g(s)), (p_f(t),p_g(t))\in \mathcal {R}$. Then

$d_{T_f}(p_f(s),p_f(t)) = f(s) + f(t) - 2\hat f(s,t),$

where $\hat f(s,t):= \min{r\in[s,t]}f(r)$. Obviously, we can replace f by g to obtain

$d_{T_g}(p_g(s),p_g(t)) = g(s) + g(t) - 2\hat g(s,t).$

By thinking slightly carefully about where the functions achieve their minima compared with where the minimum in the sup norm is achieved, we can conclude that

$|d_{T_f}(p_f(s),p_f(t)) - d_{T_g}(p_g(s),p_g(t)) | \le 4||f-g||,$

and so the result follows from the relation between Gromov-Hausdorff distance and the infimum of distortions over the set of correspondences proved at the end of the previous post.

Gromov-Hausdorff Limits of Real Trees

If we are going to consider the Gromov-Hausdorff topology for limits of tree, we want to be sure that the limit of a sequence of real trees is another real tree. In particular, we want to show that this convergence preserves the property of being a geodesic space.

Theorem 4.19 – Given $X_n$ geodesic spaces, and $X$ complete, such that $X_n \stackrel{d_{GH}}\rightarrow X$, then X is geodesic.

Proof: Here, I’ll go in the opposite order to Evans’ book, as I think it’s easier understand why a special case of this implies the whole result once you’ve actually shown that special case.

Anyway, we want to show that given $x,y\in X$, there is a geodesic segment $x\leftrightarrow y$ in X. We will start by showing that there is a well-defined midpoint of the geodesic, that is a point $z \in X$ such that $d(x,z)=d(y,z)=\frac12 d(x,y)$.

Given arbitrary $\epsilon>0$, we can take n such that $d_{GH}(X,X_n) < \frac{\epsilon}{3}$ and then a correspondence $\mathcal{R}$ between $X,X_n$ with $\mathrm{dis}\mathcal {R}<\frac{2\epsilon}{3}$. Now, by definition of correspondence, we have $x',y'\in X_n$ such that $(x,x'),(y,y')\in\mathcal{R}$. But we do have geodesics in $X_n$, so we can take $z'\in X_n$ the midpoint of geodesic $x'\leftrightarrow y'$. Predictably, we now take $z\in X$ such that $(z,z')\in\mathcal {R}$.

We can show that z is ‘almost’ the midpoint of [x,y] in the sense that

$|d(x,z) - \frac12 d(x,y)| \le |d(x',z')-\frac12 d(x',y')+\frac32 \mathrm{dis}\mathcal{R} \le \epsilon.$

Similarly, we have $|d(y,z)-\frac12 d(x,y)|\le \epsilon$.

Perhaps it’s helpful to think of this point z that we’ve constructed as being like a taut, but not quite rigid string. The midpoint of the string has to be fairly near the midpoint of the endpoints, and in particular, as we let $\epsilon\rightarrow 0$, the z’s we deal with form a Cauchy sequence, and thus converge to some point, which (in a case of poor notation planning) we also call $z \in X$, which is the midpoint of $[x,y]$ as described before.

We can now iterate this iteration, to demonstrate that whenever q is a dyadic rational in [0,1], there exists $z_q \in X$ such that

$d(x,z_q)=q d(x,y),\quad d(y,z_q) = (1-q)d(x,y).$

Again then, if we want the above to hold for some general real r in [0,1], we can approximate r arbitrarily well by dyadic rationals q, and the associated points $z_q$ are Cauchy and thus have a well-defined limit with the required properties. We thus have our geodesic segment $x\leftrightarrow y$.

Rooted Gromov-Hausdorff Distance

In the end, the trees we want to compare might be rooted. For example, we talk about finite trees being invariant under random re-rooting, and we might be interested in similar results for real trees, in particular the BCRT. So we need to compare metric spaces as viewed outwards from particular identified points of each space.

An isometry of rooted spaces must map the root to the root, and so we adjust to obtain rooted Gromov-Hausdorff distance. We might try to consider embeddings into a common space so that the roots are shared, but it will be more convenient to maintain the infimum over metrics on the disjoint union as before. But to ensure the roots are in roughly the ‘same’ place in both set embeddings, we minimise the maximum of the Hausdorff distance between the sets, and the distance between the images of the roots in the common covering space.

Similar results apply as in the unrooted case, and normally the proofs are very similar. As we might expect, when defining a correspondence between rooted spaces, we demand that the pair of roots is one of the roots in the correspondence, and then the same equivalence between the minimal distortion and the rooted G-H distance applies.

Evans shows that $\mathbb{T}^{\mathrm{root}}$, the set of compact rooted trees is complete and separable under the rooted G-H topology. Separability is relatively easy to see. Compact trees have finite $\epsilon$-nets, and there is a canonical way to view the net as the vertices of a finite tree with edge lengths. Approximating these edge lengths by rationals and consider the countable family of isomorphism classes of rooted finite trees gives separability.

If we want, we can also define k-pointed Gromov-Hausdorff distance, where we demand k points in each space are held fixed.

Tree $\eta$-erasure

To show that this is a natural topology to consider for the family of trees, Evans devotes a short section to the operation of $\eta$-erasure, where all points within $\eta$ of a leaf are removed from a given tree.

Formally, $R_\eta$ is a map $\mathbb{T}^{\mathrm{root}}\rightarrow \mathbb{T}^{\mathrm{root}}$ (recalling that these are compact real trees), so that $R_\eta(T)$ consists of the tree

$\{\rho\}\cup\{a \in T \, : \, \exists x\in T,\, a\in[\rho, x],d_T(a,x)\ge \eta\},$

rooted again at $\rho$. We claim that the range of $R_\eta$ is the set of compact rooted trees with a finite number of leaves. In this setting, we want a geodesic definition of leaf, for example a leaf is a point that doesn’t lie in the interior of the (unique) geodesic segment between any other point and the root.

Given a tree T with a finite number of leaves, we can glue disjoint segments of length $\eta$ onto every leaf. Taking $R_\eta$ of this deeper tree will give T. Similarly, suppose $R_\eta(T)$ has infinitely many leaves, which we can label $a_1,a_2,\ldots$. Thus we also have $x_1,x_2,\ldots \in T$ such that $a_i\in[\rho,x_i]$, and the segments $\{[a_i,x_i]\}$ are disjoint, and all have length at least $\eta$, hence T cannot be compact, as it has no finite $\frac{\eta}{2}$ net.

It is clear that for fixed T, the family of these maps applied to T is continuous with respect to $\eta$ in the G-H topology. When $\eta$ is changed a small amount, the amount of extra tree removed is locally small, and so approximating correspondences by points in what is left is absolutely fine. Indeed, the operations $T_{\eta_1}, T_{\eta_2}$ commute to give $T_{\eta_1+\eta_2}$.

We want to show that for fixed $\eta$ this map $R_\eta$ is continuous with respect to G-H topology. Suppose two compact rooted trees S,T are covered by a rooted correspondence $\mathcal R$ with distortion $\epsilon \ll \eta$. We can’t immediately restrict $\mathcal R$ to $R_\eta(S)\times R_\eta(T)$, as it won’t necessarily be a surjection under the projection maps any more.

But we can get around this. Note that if $(x,y)\in\mathcal{R}$, then $|d_S(\rho_S,x) - d_T(\rho_T,y)| \le \epsilon$ by assumption. We will construct a correspondence $\mathcal R'$ on the erased trees as follows. Given $(x,y)\in\mathcal{R}$, if $(x,y)\in R_\eta(S)\times R_\eta(T)$, we keep it, and if $x\not\in R_\eta(S), y\not \in R_\eta(y)$, we throw it away. Suppose we have $(s,t) in \mathcal{R}$ with $s\in R_\eta(S), t\not \in R_\eta(T)$. Let $l_s$ be a leaf of S such that $s \in [\rho_S,l_s]$, and let $\bar t$ be the farthest point from the root of the geodesic $[\rho_T,t]$ restricted to $R_\eta(T)$. So the tree above $\bar t$ includes t and has height at most $\eta$.

In words, if a point appears in a pair in the correspondence but is removed by the erasure, we replace it in the pair with the point closest to the original point that was not removed by the erasure.

It remains to prove that this works. My original proof was short but false, and its replacement is long (and I hope true now), but will postpone writing this down either until another post, or indefinitely. The original proof by Evans and co-authors can be found as the main content of Lemma 2.6 in the original paper [1].

REFERENCES

[1] Evans, Pitman, Winter – Rayleigh processes, real trees, and root growth with regrafting.

# Critical Components in Erdos-Renyi

In various previous posts, I’ve talked about the phase transition in the Erdos-Renyi random graph process. Recall the definition of the process. Here we will use the Gilbert model G(n,p), where we have n vertices, and between any pair of vertices we add an edge, independently of other pairs with probability p. We are interested in the sparse scaling, where the typical vertex has degree O(1) in n, and so p=c/n for constant c>0, and we assume throughout that n is large. We could alternatively have considered the alternative Erdos-Renyi model where we choose uniformly at random from the set of graphs with n vertices and some fixed number of edges. Almost all the results present work equally well in this setting.

As proved by Erdos and Renyi, the typical component structure of such a graph changes noticeably around the threshold c=1. Below this, in the subcritical regime, all the components are small, meaning of size at most order O(log n). Above this, in the supercritical regime, there is a single giant component on some non-zero proportion of the vertices. The rest of the graph looks subcritical. The case c=1 exhibits a phase transition between these qualitatively different behaviours. They proved that here, the largest component is with high probability O(n^2/3). It seems that they thought this result held whenever c=1-o(1), but it turns out that this is not the case. In this post, I will discuss some aspects of behaviour around criticality, and the tools needed to treat them.

The first question to address is this: how many components of size n^{2/3} are there? It might be plausible that there is a single such component, like for the subsequent giant component. It might also be plausible that there are n^1/3 such components, so O(n) vertices are on such critical components. As then it is clear how we transition out of criticality into supercriticality – all the vertices on critical components coalesce to form the new giant component.

In fact neither of these are correct. The answer is that for all integers k>0, with high probability the k-th largest component is on a size scale of n^2/3. This is potentially a confusing statement. It looks like there are infinitely many such components, but of course for any particular value of n, this cannot be the case. We should think of there being w(1) components, but o(n^b) for any b>0.

The easiest way to see this is by a duality argument, as we have discussed previously for the supercritical phase. If we remove a component of size O(n^2/3), then what remains is a random graph with n-O(n^2/3) vertices, and edge probability the same as originally. It might make sense to rewrite this probability 1/n as

$\frac{1}{n-O(n^{2/3})}\cdot \frac{n-O(n^{2/3})}{n}=\frac{1-O(n^{-1/3})}{n-O(n^{2/3})}.$

The approximation in the final numerator is basically the same as

$1-o\left(n-O(n^{2/3})\right).$

Although we have no concrete reasoning, it seems at least plausible that this should look similar in structure to G(n,1/n). In particular, there should be another component of size

$O\left([n-O(n^{2/3})]^{2/3}\right)=O(n^{2/3})$.

In fact, the formal proof of this proceeds by an identical argument, only using the exploration process. Because I’ve described this several times before, I’ll be brief. We track how far we have gone through each component in a depth-first walk. In both the supercritical and subcritical cases, when we scale correctly we get a random path which is basically deterministic in the limit (in n). For exactly the same reasons as visible CLT fluctuations for partial sums of RVs with expectation zero, we start seeing interesting effects at criticality.

The important question is the order of rescaling to choose. At each stage of the exploration process, the number of vertices added to the stack is binomial. We want to distinguish between components of size $O(n^{2/3})$ so we should look at the exploration process at time $sn^{2/3}$. The drift of the exploration process is given by the expectation of a binomial random variable minus one (since we remove the current vertex from the stack as we finish exploring it). This is given by

$\mathbb{E}=\left[n-sn^{2/3}\right]\cdot \frac{1}{n}-1=-sn^{-1/3}.$

Note that this is the drift in one time-step. The drift in $n^{2/3}$ time-steps will accordingly by $sn^{1/3}$. So, if we rescale time by $n^{2/3}$ and space by $n^{1/3}$, we should get a nice stochastic process. Specifically, if Z is the exploration process, then we obtain:

$\frac{1}{n^{1/3}}Z^{(n)}_{sn^{2/3}} \rightarrow_d W_s,$

where W is a Brownian motion with inhomogeneous drift -s at time s. The net effect of such a drift at a fixed positive time is given by integrating up to that time, and hence we might say the process has quadratic drift, or is parabolic.

We should remark that our binomial expectation is not entirely correct. We have discounted those $sn^{2/3}$ vertices that have already been explored, but we have not accounted for the vertices currently in the stack. We should also be avoiding considering these. However, we now have a heuristic for the approximate number of these. The number of vertices in the stack should be $O(n^{1/3})$ at all times, and so in particular will always be an order of magnitude smaller than the number of vertices already considered. Therefore, they won’t affect this drift term, though this must be accounted for in any formal proof of convergence. On the subject of which, the mode of convergence is, unsurprisingly, weak convergence uniformly on compact sets. That is, for any fixed S, the convergence holds weakly on the random functions up to time $sn^{2/3}$.

Note that this process will tend to minus infinity almost surely. Component sizes are given by excursions above the running minimum. The process given by the height of the original process above the running minimum is called reflected. Essentially, we construct the reflected process by having the same generator when the current value is positive, and forcing the process up when it is at zero. There are various ways to construct this more formally, including as the scaling limit of some simple random walks conditioned never to stay non-negative.

The cute part of the result is that it holds equally well in a so-called critical window either side of the critical probability 1/n. When the probability is $\frac{1+tn^{-1/3}}{n}$, for any $t\in \mathbb{R}$, the same argument holds. Now the drift at time s is t-s, though everything else still holds.

This result was established by Aldous in [1], and gives a mechanism for calculating distributions of component sizes and so on through this critical window.

In particular, we are now in a position to answer the original question regarding how many such components there were. The key idea is that because whenever we exhaust a component in the exploration process, we choose a new vertex uniformly at random, we are effectively choosing a component according to the size-biased distribution. Roughly speaking, the largest components will show up near the beginning. Note that a critical $O(n^{2/3})$ component will not necessarily be exactly the first component in the exploration process, but the components that are explored before this will take up sufficiently few vertices that they won’t show up in the scaling of the limit.

In any case, the reflected Brownian motion ‘goes on forever’, and the drift is eventually very negative, so there cannot be infinitely wide excursions, hence there are infinitely many such critical components.

If we care about the number of cycles, we can treat this also via the exploration process. Note that in any depth-first search we are necessarily only interested in a spanning tree of the host graph. Anyway, when we are exploring a vertex, there could be extra edges to other vertices in the stack, but not to vertices we’ve already finished exploring (otherwise the edge would have been exposed then). So the expected number of excess edges into a vertex is proportional to the height of the exploration process at that vertex. So the overall expected number of excess edges, conditional on the exploration process is the area under the curve. This carries over perfectly well into the stochastic process limit. It is then a calculation to verify that the area under the curve is almost surely infinite, and thus that we expect there to be infinitely many cycles in a critical random graph.

REFERENCES

[1] Aldous D. – Brownian excursions, critical random graphs and the multiplicative coalescent

# Uniform Spanning Trees

For applications to random graphs, the local binomial structure and independence means that the Galton-Watson branching process is a useful structure to consider embedding in the graph. In several previous posts, I have shown how we can set up the so-called exploration process which visits the sites in a component as if the component were actually a tree. The typical degree is O(1), and so in particular small components will be trees with high probability in the limit. In the giant component for a supercritical graph, this is not the case, but it doesn’t matter, as we ignore vertices we have already explored in our exploration process. We can consider the excess edges separately by ‘sprinkling’ them back in once we have the tree-like backbone of all the components. Again, independence is crucial here.

I am now thinking about a new model. We take an Erdos-Renyi process as before, with edges arriving at some fixed rate, but whenever a cycle appears, we immediately delete all the edges that make up the cycle. Thus at all times the system consists of a collection (or forest) of trees on the n vertices. So initially this process will look exactly like the normal E-R process, but as soon as the components start getting large, we start getting excess edges which destroy the cycles and make everything small again. The question to ask is: if we run the process for long enough, roughly how large are all the components? It seems unlikely that the splitting mechanism is so weak that we will get true giant components forming, ie O(n) sizes, so we might guess that, in common with some other split-merge models of this type, we end up with components of size $n^{2/3}$, as in the critical window for the E-R process.

In any case, the scaling limit process is likely to have components whose sizes grow with n, so we will have a class of trees larger than those we have considered previously, which have typically been O(1). So it’s worth thinking about some ways to generate random trees on a fixed number of vertices.

Conditioned Galton-Watson

Our favourite method of creating trees is inductive. We take a root and connect the root to a number of offspring given by a fixed distribution, and each of these some offspring given by an independent sample from the same distribution and so on. The natural formulation gives no control over the size of the tree. This is a random variable whose distribution depends on the offspring distribution, and which in some circumstances be computed explicitly, for example when the offspring distribution is geometric. In other cases, it is easier to make recourse to generating functions or to a random walk analogue as described in the exploration process discussion.

Of course, there is nothing to stop us conditioning on the total size of the population. This is equivalent to conditioning on the hitting time of -1 for the corresponding random walk, and Donsker’s theorem gives several consequences of a convergence relation towards a rescaled Brownian excursion. Note that there is no a priori labelling for the resulting tree. This will have to be supplied later, with breadth-first and depth-first the most natural choices, which might cause annoyance if you actually want to use it. In particular, it is not obvious, and probably not true unless you are careful, that the distribution is invariant under permuting the labels (having initially assumed 1 is the root etc) which is not ideal if you are embedding into the complete graph.

However, we would like to have some more direct constructions of random trees on n vertices. We now consider perhaps the two best known such methods. These are of particular interest as they are applicable to finding random spanning trees embedded in any graph, rather than just the complete graph.

Uniform Spanning Tree

Given a connected graph, consider the set of all subgraphs which are trees and span the vertex set of the original graph. An element of this set is called a spanning tree. A uniform spanning tree is chosen uniformly at random from the set of spanning trees on the complex graph on n vertices. A famous result of Arthur Cayley says that the number of such spanning trees is $n^{n-2}$. There are various neat proofs, many of which consider a mild generalisation which gives us a more natural framework for using induction. This might be a suitable subject for a subsequent post.

While there is no objective answer to the question of what is the right model for random trees on n vertices, this is what you get from the Erdos-Renyi process. Formally, conditional on the sizes of the (tree) components, the structures of the tree components are given by UST.

To see why this is the case, observe that when we condition that a component has m vertices and is a tree, we are demanding that it be connected and have m-1 edges. Since the probability of a particular configuration appearing in G(n,p) is a function only of the number of edges in the configuration, it follows that the probability of each spanning tree on the m vertices in question is equal.

Interesting things happen when you do this dynamically. That is, if we have two USTs of sizes m and n at some time t, and condition that the next edge to be added in the process joins them, then the resulting component is not a UST on m+n vertices. To see why, consider the probability of a ‘star’, that is a tree with a single distinguished vertex to which every other vertex is joined. Then the probability that the UST on m vertices is a star is $\frac{m}{m^{m-2}}=m^{-(m-3)}$. By contrast, it is not possible to obtain a star on m+n vertices by joining a tree on m vertices and a tree on n vertices with an additional edge.

However, I think the UST property is preserved by the cycle deletion mechanism mentioned at the very start of this post. My working has been very much of the back of the envelope variety, but I am fairly convinced that once you have taken a UST and conditioned on the sizes of the smaller trees which result from cycle deletion. My argument is that you might as well fix the cycle to be deleted, then condition on how many vertices are in each of the trees coming off this cycle. Now the choice of each of these trees is clearly uniform among spanning trees on the correct number of vertices.

However, it is my current belief that the combination of these two mechanisms does not give UST-like trees even after conditioning on the sizes at fixed time.

# Exploring the Supercritical Random Graph

I’ve spent a bit of time this week reading and doing all the exercises from some excellent notes by van der Hofstad about random graphs. I think they are absolutely excellent and would not be surprised if they become the standard text for an introduction to probabilistic combinatorics. You can find them hosted on the author’s website. I’ve been reading chapters 4 and 5, which approaches the properties of phase transitions in G(n,p) by formalising the analogy between component sizes and population sizes in a binomial branching process. When I met this sort of material for the first time during Part III, the proofs generally relied on careful first and second moment bounds, which is fine in many ways, but I enjoyed vdH’s (perhaps more modern?) approach, as it seems to give a more accurate picture of what is actually going on. In this post, I am going to talk about using the branching process picture to explain why the giant component emerges when it does, and how to get a grip on how large it is at any time after it has emerged.

Background

A quick tour through the background, and in particular the notation will be required. At some point I will write a post about this topic in a more digestible format, but for now I want to move on as quickly as possible.

We are looking at the sparse random graph $G(n,\frac{\lambda}{n})$, in the super-critical phase $\lambda>1$. With high probability (that is, with probability tending to 1 as n grows), we have a so-called giant component, with O(n) vertices.

Because all the edges in the configuration are independent, we can view the component containing a fixed vertex as a branching process. Given vertex v(1), the number of neighbours is distributed like $\text{Bi}(n-1,\frac{\lambda}{n})$. The number of neighbours of each of these which we haven’t already considered is then $\text{Bi}(n-k,\frac{\lambda}{n})$, conditional on k, the number of vertices we have already discounted. After any finite number of steps, k=o(n), and so it is fairly reasonable to approximate this just by $\text{Bi}(n,\frac{\lambda}{n})$. Furthermore, as n grows, this distribution converges to $\text{Po}(\lambda)$, and so it is natural to expect that the probability that the fixed vertex lies in a giant component is equal to the survival probability $\zeta_\lambda$ (that is, the probability that it is infinite) of a branching process with $\text{Po}(\lambda)$ offspring distribution. Note that given a graph, the probability of a fixed vertex lying in a giant component is equal to the fraction of the vertex in the giant component. At this point it is clear why the emergence of the giant component must happen at $\lambda=1$, because we require $\mathbb{E}\text{Po}(\lambda)>1$ for the survival probability to be non-zero. Obviously, all of this needs to be made precise and rigorous, and this is treated in sections 4.3 and 4.4 of the notes.

Exploration Process

A common functional of a rooted branching process to consider is the following. This is called in various places an exploration process, a depth-first process or a Lukasiewicz path. We take a depth-first labelling of the tree v(0), v(1), v(2),… , and define c(k) to be the number of children of vertex v(k). We then define the exploration process by:

$S(0)=0,\quad S(k+1)=S(k)+c(k)-1.$

By far the best way to think of this is to imagine we are making the depth-first walk on the tree. S(k) records how many vertices we have seen (because they are connected by an edge to a vertex we have visited) but have not yet visited. To clarify understanding of the definition, note that when you arrive at a vertex with no children, this should decrease by one, as you can see no new vertices, but have visited an extra one.

This exploration process is useful to consider for a couple of reasons. Firstly, you can reconstruct the branching process directly from it. Secondly, while other functionals (eg the height, or contour process) look like random walks, the exploration process genuinely is a random walk. The distribution of the number of children of the next vertex we arrive at is independent of everything we have previously seen in the tree, and is the same for every vertex. If we were looking at branching processes in a different context, we might observe that this gives some information in a suitably-rescaled limit, as rescaled random walks converge to Brownian motion if the variance of the (offspring) distribution is finite. (This is Donsker’s result, which I should write something about soon…)

The most important property is that the exploration process returns to 0 precisely when we have exhausted all the vertices in a component. At that point, we have seen exactly the vertices which we have explored. There is no reason not to extend the definition to forests, that is a union of trees. The depth-first exploration is the same – but when we have exhausted one component, we move onto another component, chosen according to some labelling property. Then, running minima of the exploration process (ie times when it is smaller than it has been before) correspond to jumping between components, and thus excursions above the minimum to components themselves. The running minimum will be non-positive, with absolute value equal to the number of components already exhausted.

Although the exploration process was defined with reference to and in the language of trees, the result of a branching process, this is not necessary. With some vertex denoted as the root, we can construct a depth-first labelling of a general graph, and the exploration process follows exactly as before. Note that we end up ignoring all edges except a set that forms a forest. This is what we will apply to G(n,p).

Exploring G(n,p)

When we jump between components in the exploration process on a supercritical (that is $\lambda>1$) random graph, we move to a component chosen randomly with size-biased distribution. If there is a giant component, as we know there is in the supercritical case, then this will dominate the size-biased distribution. Precisely, if the giant component takes up a fraction H of the vertices, then the number of components to be explored before we get to the giant component is geometrically distributed with parameter H. All other components have size O(log n), so the expected number of vertices explored before we get to the giant component is O(log n)/H = o(n), and so in the limit, we explore the giant component immediately.

The exploration process therefore gives good control on the giant component in the limit, as roughly speaking the first time it returns to 0 is the size of the giant component. Fortunately, we can also control the distribution of S_t, the exploration process at time t. We have that:

$S_t+(t-1)\sim \text{Bi}(n-1,1-(1-p)^t).$

This is not too hard to see. $S_t+(t-1)$ is number of vertices we have explored or seen, ie are connected to a vertex we have explored. Suppose the remaining vertices are called unseen, and we began the exploration at vertex 1. Then any vertex with label in {2,…,n} is unseen if it successively avoids being in the neighbourhood of v(1), v(2), … v(t). This happens with probability $(1-p)^t$, and so the probability of being an explored or seen vertex is the complement of this.

In the supercritical case, we are taking $p=\frac{\lambda}{n}$ with $\lambda>1$, and we also want to speed up S, so that all the exploration processes are defined on [0,1], and rescale the sizes by n, so that it records the fraction of the graph rather than the number of vertices. So we set consider the rescaling $\frac{1}{n}S_{nt}$.

It is straightforward to use the distribution of S_t we deduce that the asymptotic mean $\mathbb{E}\frac{1}{n}S_{nt}=\mu_t = 1-t-e^{-\lambda t}$.

Now we are in a position to provide more concrete motivation for the claim that the proportion of vertices in the giant component is $\zeta_\lambda$, the survival probability of a branching process with $\text{Po}(\lambda)$ offspring distribution. It helps to consider instead the extinction probability $1-\zeta_\lambda$. We have:

$1-\zeta_\lambda=\sum_{k\geq 0}\mathbb{P}(\text{Po}(\lambda)=k)(1-\zeta_\lambda)^k=e^{-\lambda\zeta_\lambda},$

where the second equality is a consequence of the simple form for the moment generating function of the Poisson distribution.

As a result, we have that $\mu_{\zeta_\lambda}=0$. In fact we also have a central limit theorem for S_t, which enables us to deduce that $\frac{1}{n}S_{n\zeta_\lambda}=0$ with high probability, as well as in expectation, which is precisely what is required to prove that the giant component of $G(n,\frac{\lambda}{n})$ has size $n(\zeta_\lambda+o(1))$.

# Markovian Excursions

In the previous post, I talked about the excursions of a Brownian motion. Today I’m thinking about how to extend these ideas to more general Markov chains. First we want to rule out some situations. In particular, we aren’t hugely interested in discrete time Markov chains. The machinery is fairly well established for excursions, whether or not the chain is transient. Furthermore, if the state space is discrete, as for a Poisson process for example, the discussion is not hugely interesting either. Remember that the technical challenges in the constructions of local time arise because of the Blumenthal 0-1 law property that Brownian motion visits 0 infinitely often in the small window after the start time. We therefore want the process to be regular at the point of the state space under discussion. This is precisely the condition described above for BM about 0.

Why is it harder in general?

The informal notion of a local time should transfer to a more general Markov chain, but there are some problems. Firstly, to define something in terms of an integral is not general enough. The state space E needs some topological structure, but any meaningful definition must be in terms of functions from E into the reals. There were also all sorts of special properties of Brownian motion, including the canonical time-space rescaling that came in handy in that particular case. It turns out to be easiest to consider the excursion measure on a general Markov chain through its Laplace transform.

Definition and Probabilistic interpretations

The resolvent is the Laplace transform of the transition probability $P_t(x,A)$, viewed as an operator on functions $f:E\rightarrow \mathbb{R}$.

$R_\lambda f(x):=\mathbb{E}_x\left[\int_0^\infty e^{-\lambda t}f(X_t)dt\right]=\int_0^\infty e^{-\lambda t}P_tf(x)dt$.

We can interpret this in terms of the original process in a couple of ways which may be useful later. The motivation is that we would like to specify a Poisson process of excursions, for which we need to know the rate. We hope that the rate will in fact be constant, so it will in fact to suffice to work out the properties of the expected number of excursions (or whatever) up to various random times, in particular those given by exponential RVs.

So, we take $\zeta\sim\text{Exp}(\lambda)$ independent of everything else, and assume that we ‘kill the chain’ at time $\zeta$. Then, by shuffling expectations in and out of the integral and separating independent bits, we get:

$R_\lambda f(x)=\mathbb{E}_x\int_0^\zeta f(X_s)ds = \frac{1}{\lambda}\mathbb{E}_xf(X_\zeta)$.

As in the Brownian local time description

$R_\lambda 1_A(x)=\mathbb{E}(\text{time spent in }A\text{ before death at time }\zeta_\lambda)$.

Markovian property

We want to show that excursions are Markov, once we’ve sorted out what an ‘excursion’ actually means in this context. We do know how to deal with the Markovian property once we are already on an excursion. It is relatively straightforward to define an extension of the standard transition probability operator, to include a condition that the chain should not hit point a during the transition. That is

$_aP_t(x,A):= \mathbb{P}_x(X_t\in A\cap H_a>t)$.

This will suffice to define the behaviour once an excursion has started. The more complicated bit will be the entrance law $n_t(A)$, being the probability of arriving at A after time t of an excursion. To summarise, as with BM, all the technical difficulties with an excursion happen at the beginning, ie bouncing around the start point, but once it is ‘up-and-running’, its path is Markovian and controlled by $_aP_t$.

Marking

The link between the resolvent and the excursions, is provided as in the Brownian case, by supplying a PPP of marks at uniform rate $\lambda$ to real time. This induces a mark process on excursions, weighted by an (exponential) function of excursion length. We make no distinction between an excursion including one mark or many marks. Then the measure on marked excursion is, in a mild abuse of notation:

$n_\lambda=(1-e^{-\lambda\zeta(f)})\cdot n.$

We compare with the Laplace transform $n_\lambda(dx)=\int_0^\infty e^{-\lambda t}dtn_t(dx)$ using a probabilistic argument.

We can apply the measure to a function in the usual way: $\lambda n_\lambda(1_A)$ is the measure of those excursions for which the first mark occurs in $A$. So by taking $A=E$, we get

$\lambda n_\lambda(1)=\text{ Excursion measure }=\int_U n(df)(1-e^{-\lambda\zeta(f)}).$

We have therefore linked the exponential mark process on excursion measure with the Laplace transform of the entrance law. So in particular:

$\frac{\lambda n_\lambda(A)}{\lambda n_\lambda(1)}=\mathbb{P}(\text{first mark when in }A)=\int_0^\infty \lambda e^{-\lambda t}P_t(0,A)dt=\lambda R_\lambda 1_A(0)$.

The resolvent is relatively easy to calculate explicitly, and so we can find the Laplace transform $n_\lambda(A)$. From this it is generally possible to invert the transform to find the entrance law $n_t$.

References

A Guided Tour Through Excursions – L. C. G. Rogers.

This pair of posts is very much a paraphrase of chapters 3 and 4 of this excellent text. The original can be found here (possibly not open access?)

# Brownian Excursions and Local Time

I’ve been spending a fair bit of time this week reading and thinking about the limits of various combinatorial objects, in particular letting the number of vertices tend to $\infty$ in models of random graphs with various constraints. Perhaps predictably, like so many continuous stochastic objects, yet again the limiting ‘things’ turn out to be closely linked to Brownian Motion. As a result, I’ve ended up reading a bit about the notion of local time, and thought it was sufficiently elegant even by itself to justify a quick post.

Local Time

In general, we might be interested in calculating a stochastic integral like

$\int_0^t f(B_s)ds$.

Note that, except in some highly non-interesting cases, this is a random variable. Our high school understanding of Riemannian integration encourages thinking of this as a ‘pathwise’ integral along the path evolving in time. But of course, that’s orthogonal to the approach we start thinking about when we are introduced to the Lebesgue integral. There we think about potential values of the integrand, and weight their contribution by the (Lebesgue) measure of the subset of the domain in which they appear.

Can we do the same for the stochastic integral? That is, can we find a measure which records how long the Brownian Motion spends at a point x? This measure will not be deterministic – effectively the stochastic behaviour of BM will be encoded through the measure rather than the argument of the function.

The answer is yes, and the measure in question is referred to as local time. More formally, we want

$\int_0^t f(B_s)ds=\int_\mathbb{R}f(x)L(t,x)dx.$ (*)

where the local time L(t,x) is a random process, increasing for fixed x. Informally, one could take

$\partial_t L(t,x) \propto 1(B_t=x)$

but clearly in practice that won’t do at all for a definition, and so instead we use (*). In the usual way, if we want (*) to hold for all reasonably nice functions f, it suffices to check it for the indicator functions of Borel sets. L(t,.) is therefore often referred to as occupation density, while L(.,A) is local time.

Local Time as natural index for Excursions

An excursion, for example of Brownian Motion, is a segment of the path that has zero value only at its endpoints. Alternatively, it is a maximal open interval of time such that the path is away from 0. We want to specify the measure on these excursions. Here are some obvious difficulties.

By Blumenthal’s 0-1 law, BM started from zero hits zero infinitely often in any time interval [0,e], so in the same way that there is no first positive rational, there is no first excursion. We could pick the excursion occurring in progress at a fixed time t, but this is little better. Firstly, the resultant measure is size-biased by the length of the excursion, and more importantly, the proximity of t to the origin may be significant unless we know of some memorylessness type of property to excursions.

Local time allows us to solve these problems. We restrict attention to $L_t:=L(t,0)$, the occupation density of 0. Let’s think about some advantages of indexing excursions by local time rather than by the start time:

• The key observation is that local time remains constant on excursions. That is, if we are avoiding 0, the local time at 0 cannot grow because the BM spends no time there!
• If we use start time, then we have a countably infinite number of small excursions accumulating close to 0, ie with very small start time. However, local time increases rapidly when there are lots of small excursions. Remember, lots of small excursions means that the BM hits 0 lots of times. So local time grows quickly through the annoying bits, and effectively provides a size-biasing for excursions that allows us to ignore the effects of the ‘Blumenthal excursions’ near time 0.
• When indexed by time, excursions might be Markovian, in the sense that subsequent excursions (and in particular their lengths) are independent of past excursions.This is certainly not the case if you index by start time! If an excursion starts at time t and has length u, then the ‘next’ excursions, in as much as that makes sense, must surely start at time t+u.

We know there are only countably many excursions, hence there are only countably many local times which pertain to an excursion. This motivates considering the set of excursions as a Poisson Point Process on local time. Once you’ve had this idea, everything follows quite nicely. Working out the distribution of the constant rate (which is a measure on the set of excursions) remains, but essentially we now have a sensible framework for tracking the process of excursions, and from this we can reconstruct the original Brownian Motion.