The sunflower lemma via Shannon entropy

A family of sets for some is a sunflower if there is a core set contained in each of the such that the petal sets are disjoint. If , let denote the smallest natural number with the property that any family of distinct sets of cardinality at most contains distinct elements that form a sunflower. The celebrated Erdös-Rado theorem asserts that is finite; in fact Erdös and Rado gave the bounds

The sunflower conjecture asserts in fact that the upper bound can be improved to . This remains open at present despite much effort (including a Polymath project); after a long series of improvements to the upper bound, the best general bound known currently is

for all , established in 2019 by Rao (building upon a recent breakthrough a month previously of Alweiss, Lovett, Wu, and Zhang). Here we remove the easy cases or in order to make the logarithmic factor a little cleaner.

Rao’s argument used the Shannon noiseless coding theorem. It turns out that the argument can be arranged in the very slightly different language of Shannon entropy, and I would like to present it here. The argument proceeds by locating the core and petals of the sunflower separately (this strategy is also followed in Alweiss-Lovett-Wu-Zhang). In both cases the following definition will be key. In this post all random variables, such as random sets, will be understood to be discrete random variables taking values in a finite range. We always use boldface symbols to denote random variables, and non-boldface for deterministic quantities.

Definition 1 (Spread set) Let . A random set is said to be -spread if one has

for all sets . A family of sets is said to be -spread if is non-empty and the random variable is -spread, where is drawn uniformly from .

The core can then be selected greedily in such a way that the remainder of a family becomes spread:

Lemma 2 (Locating the core) Let be a family of subsets of a finite set , each of cardinality at most , and let . Then there exists a “core” set of cardinality at most such that the set

has cardinality at least , and such that the family is -spread. Furthermore, if and the are distinct, then .

Proof: We may assume is non-empty, as the claim is trivial otherwise. For any , define the quantity

and let be a subset of that maximizes . Since and when , we see that . If the are distinct and , then we also have when , thus in this case we have .

Let be the set (3). Since , is non-empty. It remains to check that the family is -spread. But for any and drawn uniformly at random from one has

Observe that , and the probability is only non-empty when are disjoint, so that . The claim follows.

In view of the above lemma, the bound (2) will then follow from

Proposition 3 (Locating the petals) Let be natural numbers, and suppose that for a sufficiently large constant . Let be a finite family of subsets of a finite set , each of cardinality at most which is -spread. Then there exist such that is disjoint.

Indeed, to prove (2), we assume that is a family of sets of cardinality greater than for some ; by discarding redundant elements and sets we may assume that is finite and that all the are contained in a common finite set . Apply Lemma 2 to find a set of cardinality such that the family is -spread. By Proposition 3 we can find such that are disjoint; since these sets have cardinality , this implies that the are distinct. Hence form a sunflower as required.

Remark 4 Proposition 3 is easy to prove if we strengthen the condition on to . In this case, we have for every , hence by the union bound we see that for any with there exists such that is disjoint from the set , which has cardinality at most . Iterating this, we obtain the conclusion of Proposition 3 in this case. This recovers a bound of the form , and by pursuing this idea a little further one can recover the original upper bound (1) of Erdös and Rado.

It remains to prove Proposition 3. In fact we can locate the petals one at a time, placing each petal inside a random set.

Proposition 5 (Locating a single petal) Let the notation and hypotheses be as in Proposition 3. Let be a random subset of , such that each lies in with an independent probability of . Then with probability greater than , contains one of the .

To see that Proposition 5 implies Proposition 3, we randomly partition into by placing each into one of the , chosen uniformly and independently at random. By Proposition 5 and the union bound, we see that with positive probability, it is simultaneously true for all that each contains one of the . Selecting one such for each , we obtain the required disjoint petals.

We will prove Proposition 5 by gradually increasing the density of the random set and arranging the sets to get quickly absorbed by this random set. The key iteration step is

Proposition 6 (Refinement inequality) Let and . Let be a random subset of a finite set which is -spread, and let be a random subset of independent of , such that each lies in with an independent probability of . Then there exists another -spread random subset of whose support is contained in the support of , such that and

Note that a direct application of the first moment method gives only the bound

but the point is that by switching from to an equivalent we can replace the factor by a quantity significantly smaller than .

One can iterate the above proposition, repeatedly replacing with (noting that this preserves the -spread nature of ) to conclude

Corollary 7 (Iterated refinement inequality) Let , , and . Let be a random subset of a finite set which is -spread, and let be a random subset of independent of , such that each lies in with an independent probability of . Then there exists another random subset of with support contained in the support of , such that

Now we can prove Proposition 5. Let be chosen shortly. Applying Corollary 7 with drawn uniformly at random from the , and setting , or equivalently , we have

In particular, if we set , so that , then by choice of we have , hence

In particular with probability at least , there must exist such that , giving the proposition.

It remains to establish Proposition 6. This is the difficult step, and requires a clever way to find the variant of that has better containment properties in than does. The main trick is to make a conditional copy of that is conditionally independent of subject to the constraint . The point here is that this constrant implies the inclusions

and

Because of the -spread hypothesis, it is hard for to contain any fixed large set. If we could apply this observation in the contrapositive to we could hope to get a good upper bound on the size of and hence on thanks to (4). One can also hope to improve such an upper bound by also employing (5), since it is also hard for the random set to contain a fixed large set. There are however difficulties with implementing this approach due to the fact that the random sets are coupled with in a moderately complicated fashion. In Rao’s argument a somewhat complicated encoding scheme was created to give information-theoretic control on these random variables; below the fold we accomplish a similar effect by using Shannon entropy inequalities in place of explicit encoding. A certain amount of information-theoretic sleight of hand is required to decouple certain random variables to the extent that the Shannon inequalities can be effectively applied. The argument bears some resemblance to the “entropy compression method” discussed in this previous blog post; there may be a way to more explicitly express the argument below in terms of that method. (There is also some kinship with the method of dependent random choice, which is used for instance to establish the Balog-Szemerédi-Gowers lemma, and was also translated into information theoretic language in these unpublished notes of Van Vu and myself.)

254A, Notes 9 – second moment and entropy methods

In these notes we presume familiarity with the basic concepts of probability theory, such as random variables (which could take values in the reals, vectors, or other measurable spaces), probability, and expectation. Much of this theory is in turn based on measure theory, which we will also presume familiarity with. See for instance this previous set of lecture notes for a brief review.

The basic objects of study in analytic number theory are deterministic; there is nothing inherently random about the set of prime numbers, for instance. Despite this, one can still interpret many of the averages encountered in analytic number theory in probabilistic terms, by introducing random variables into the subject. Consider for instance the form

of the prime number theorem (where we take the limit ). One can interpret this estimate probabilistically as

where is a random variable drawn uniformly from the natural numbers up to , and denotes the expectation. (In this set of notes we will use boldface symbols to denote random variables, and non-boldface symbols for deterministic objects.) By itself, such an interpretation is little more than a change of notation. However, the power of this interpretation becomes more apparent when one then imports concepts from probability theory (together with all their attendant intuitions and tools), such as independence, conditioning, stationarity, total variation distance, and entropy. For instance, suppose we want to use the prime number theorem (1) to make a prediction for the sum

After dividing by , this is essentially

With probabilistic intuition, one may expect the random variables to be approximately independent (there is no obvious relationship between the number of prime factors of , and of ), and so the above average would be expected to be approximately equal to

which by (2) is equal to . Thus we are led to the prediction

The asymptotic (3) is widely believed (it is a special case of the Chowla conjecture, which we will discuss in later notes; while there has been recent progress towards establishing it rigorously, it remains open for now.
How would one try to make these probabilistic intuitions more rigorous? The first thing one needs to do is find a more quantitative measurement of what it means for two random variables to be “approximately” independent. There are several candidates for such measurements, but we will focus in these notes on two particularly convenient measures of approximate independence: the “” measure of independence known as covariance, and the “” measure of independence known as mutual information (actually we will usually need the more general notion of conditional mutual information that measures conditional independence). The use of type methods in analytic number theory is well established, though it is usually not described in probabilistic terms, being referred to instead by such names as the “second moment method”, the “large sieve” or the “method of bilinear sums”. The use of methods (or “entropy methods”) is much more recent, and has been able to control certain types of averages in analytic number theory that were out of reach of previous methods such as methods. For instance, in later notes we will use entropy methods to establish the logarithmically averaged version

of (3), which is implied by (3) but strictly weaker (much as the prime number theorem (1) implies the bound , but the latter bound is much easier to establish than the former).
As with many other situations in analytic number theory, we can exploit the fact that certain assertions (such as approximate independence) can become significantly easier to prove if one only seeks to establish them on average, rather than uniformly. For instance, given two random variables and of number-theoretic origin (such as the random variables and mentioned previously), it can often be extremely difficult to determine the extent to which behave “independently” (or “conditionally independently”). However, thanks to second moment tools or entropy based tools, it is often possible to assert results of the following flavour: if are a large collection of “independent” random variables, and is a further random variable that is “not too large” in some sense, then must necessarily be nearly independent (or conditionally independent) to many of the , even if one cannot pinpoint precisely which of the the variable is independent with. In the case of the second moment method, this allows us to compute correlations such as for “most” . The entropy method gives bounds that are significantly weaker quantitatively than the second moment method (and in particular, in its current incarnation at least it is only able to say non-trivial assertions involving interactions with residue classes at small primes), but can control significantly more general quantities for “most” thanks to tools such as the Pinsker inequality.

Yves Meyer wins the 2017 Abel Prize

Just a short post to note that Norwegian Academy of Science and Letters has just announced that the 2017 Abel prize has been awarded to Yves Meyer, “for his pivotal role in the development of the mathematical theory of wavelets”.  The actual prize ceremony will be at Oslo in May.

I am actually in Oslo myself currently, having just presented Meyer’s work at the announcement ceremony (and also having written a brief description of some of his work).  The Abel prize has a somewhat unintuitive (and occasionally misunderstood) arrangement in which the presenter of the work of the prize is selected independently of the winner of the prize (I think in part so that the choice of presenter gives no clues as to the identity of the laureate).  In particular, like other presenters before me (which in recent years have included Timothy Gowers, Jordan Ellenberg, and Alex Bellos), I agreed to present the laureate’s work before knowing who the laureate was!  But in this case the task was very easy, because Meyer’s areas of (both pure and applied) harmonic analysis and PDE fell rather squarely within my own area of expertise.  (I had previously written about some other work of Meyer in this blog post.)  Indeed I had learned about Meyer’s wavelet constructions as a graduate student while taking a course from Ingrid Daubechies.   Daubechies also made extremely important contributions to the theory of wavelets, but due to a conflict of interest (as per the guidelines for the prize committee) arising from Daubechies’ presidency of the International Mathematical Union (which nominates the majority of the members of the Abel prize committee, who then serve for two years) from 2011 to 2014 (and her continuing service ex officio on the IMU executive committee from 2015 to 2018), she will not be eligible for the prize until 2021 at the earliest, and so I do not think this prize should be necessarily construed as a judgement on the relative contributions of Meyer and Daubechies to this field.  (In any case I fully agree with the Abel prize committee’s citation of Meyer’s pivotal role in the development of the theory of wavelets.)

[Update, Mar 28: link to prize committee guidelines and clarification of the extent of Daubechies’ conflict of interest added. -T]

Special cases of Shannon entropy

Given a random variable that takes on only finitely many values, we can define its Shannon entropy by the formula

with the convention that . (In some texts, one uses the logarithm to base rather than the natural logarithm, but the choice of base will not be relevant for this discussion.) This is clearly a nonnegative quantity. Given two random variables taking on finitely many values, the joint variable is also a random variable taking on finitely many values, and also has an entropy . It obeys the Shannon inequalities

so we can define some further nonnegative quantities, the mutual information

and the conditional entropies

More generally, given three random variables , one can define the conditional mutual information

and the final of the Shannon entropy inequalities asserts that this quantity is also non-negative.

The mutual information is a measure of the extent to which and fail to be independent; indeed, it is not difficult to show that vanishes if and only if and are independent. Similarly, vanishes if and only if and are conditionally independent relative to . At the other extreme, is a measure of the extent to which fails to depend on ; indeed, it is not difficult to show that if and only if is determined by in the sense that there is a deterministic function such that . In a related vein, if and are equivalent in the sense that there are deterministic functional relationships , between the two variables, then is interchangeable with for the purposes of computing the above quantities, thus for instance , , , , etc..

One can get some initial intuition for these information-theoretic quantities by specialising to a simple situation in which all the random variables being considered come from restricting a single random (and uniformly distributed) boolean function on a given finite domain to some subset of :

In this case, has the law of a random uniformly distributed boolean function from to , and the entropy here can be easily computed to be , where denotes the cardinality of . If is the restriction of to , and is the restriction of to , then the joint variable is equivalent to the restriction of to . If one discards the normalisation factor , one then obtains the following dictionary between entropy and the combinatorics of finite sets:

Random variables	Finite sets
Entropy	Cardinality
Joint variable	Union
Mutual information	Intersection cardinality
Conditional entropy	Set difference cardinality
Conditional mutual information
independent	disjoint
determined by	a subset of
conditionally independent relative to

Every (linear) inequality or identity about entropy (and related quantities, such as mutual information) then specialises to a combinatorial inequality or identity about finite sets that is easily verified. For instance, the Shannon inequality becomes the union bound , and the definition of mutual information becomes the inclusion-exclusion formula

For a more advanced example, consider the data processing inequality that asserts that if are conditionally independent relative to , then . Specialising to sets, this now says that if are disjoint outside of , then ; this can be made apparent by considering the corresponding Venn diagram. This dictionary also suggests how to prove the data processing inequality using the existing Shannon inequalities. Firstly, if and are not necessarily disjoint outside of , then a consideration of Venn diagrams gives the more general inequality

and a further inspection of the diagram then reveals the more precise identity

Using the dictionary in the reverse direction, one is then led to conjecture the identity

which (together with non-negativity of conditional mutual information) implies the data processing inequality, and this identity is in turn easily established from the definition of mutual information.

On the other hand, not every assertion about cardinalities of sets generalises to entropies of random variables that are not arising from restricting random boolean functions to sets. For instance, a basic property of sets is that disjointness from a given set is preserved by unions:

Indeed, one has the union bound

Applying the dictionary in the reverse direction, one might now conjecture that if was independent of and was independent of , then should also be independent of , and furthermore that

but these statements are well known to be false (for reasons related to pairwise independence of random variables being strictly weaker than joint independence). For a concrete counterexample, one can take to be independent, uniformly distributed random elements of the finite field of two elements, and take to be the sum of these two field elements. One can easily check that each of and is separately independent of , but the joint variable determines and thus is not independent of .

From the inclusion-exclusion identities

one can check that (1) is equivalent to the trivial lower bound . The basic issue here is that in the dictionary between entropy and combinatorics, there is no satisfactory entropy analogue of the notion of a triple intersection . (Even the double intersection only exists information theoretically in a “virtual” sense; the mutual information allows one to “compute the entropy” of this “intersection”, but does not actually describe this intersection itself as a random variable.)

However, this issue only arises with three or more variables; it is not too difficult to show that the only linear equalities and inequalities that are necessarily obeyed by the information-theoretic quantities associated to just two variables are those that are also necessarily obeyed by their combinatorial analogues . (See for instance the Venn diagram at the Wikipedia page for mutual information for a pictorial summation of this statement.)

One can work with a larger class of special cases of Shannon entropy by working with random linear functions rather than random boolean functions. Namely, let be some finite-dimensional vector space over a finite field , and let be a random linear functional on , selected uniformly among all such functions. Every subspace of then gives rise to a random variable formed by restricting to . This random variable is also distributed uniformly amongst all linear functions on , and its entropy can be easily computed to be . Given two random variables formed by restricting to respectively, the joint random variable determines the random linear function on the union on the two spaces, and thus by linearity on the Minkowski sum as well; thus is equivalent to the restriction of to . In particular, . This implies that and also , where is the quotient map. After discarding the normalising constant , this leads to the following dictionary between information theoretic quantities and linear algebra quantities, analogous to the previous dictionary:

Random variables	Subspaces
Entropy	Dimension
Joint variable	Sum
Mutual information	Dimension of intersection
Conditional entropy	Dimension of projection
Conditional mutual information
independent	transverse ()
determined by	a subspace of
conditionally independent relative to	, transverse.

The combinatorial dictionary can be regarded as a specialisation of the linear algebra dictionary, by taking to be the vector space over the finite field of two elements, and only considering those subspaces that are coordinate subspaces associated to various subsets of .

As before, every linear inequality or equality that is valid for the information-theoretic quantities discussed above, is automatically valid for the linear algebra counterparts for subspaces of a vector space over a finite field by applying the above specialisation (and dividing out by the normalising factor of ). In fact, the requirement that the field be finite can be removed by applying the compactness theorem from logic (or one of its relatives, such as Los’s theorem on ultraproducts, as done in this previous blog post).

The linear algebra model captures more of the features of Shannon entropy than the combinatorial model. For instance, in contrast to the combinatorial case, it is possible in the linear algebra setting to have subspaces such that and are separately transverse to , but their sum is not; for instance, in a two-dimensional vector space , one can take to be the one-dimensional subspaces spanned by , , and respectively. Note that this is essentially the same counterexample from before (which took to be the field of two elements). Indeed, one can show that any necessarily true linear inequality or equality involving the dimensions of three subspaces (as well as the various other quantities on the above table) will also be necessarily true when applied to the entropies of three discrete random variables (as well as the corresponding quantities on the above table).

However, the linear algebra model does not completely capture the subtleties of Shannon entropy once one works with four or more variables (or subspaces). This was first observed by Ingleton, who established the dimensional inequality

for any subspaces . This is easiest to see when the three terms on the right-hand side vanish; then are transverse, which implies that ; similarly . But and are transverse, and this clearly implies that and are themselves transverse. To prove the general case of Ingleton’s inequality, one can define and use (and similarly for instead of ) to reduce to establishing the inequality

which can be rearranged using (and similarly for instead of ) and as

but this is clear since .

Returning to the entropy setting, the analogue

of (3) is true (exercise!), but the analogue

of Ingleton’s inequality is false in general. Again, this is easiest to see when all the terms on the right-hand side vanish; then are conditionally independent relative to , and relative to , and and are independent, and the claim (4) would then be asserting that and are independent. While there is no linear counterexample to this statement, there are simple non-linear ones: for instance, one can take to be independent uniform variables from , and take and to be (say) and respectively (thus are the indicators of the events and respectively). Once one conditions on either or , one of has positive conditional entropy and the other has zero entropy, and so are conditionally independent relative to either or ; also, or are independent of each other. But and are not independent of each other (they cannot be simultaneously equal to ). Somehow, the feature of the linear algebra model that is not present in general is that in the linear algebra setting, every pair of subspaces has a well-defined intersection that is also a subspace, whereas for arbitrary random variables , there does not necessarily exist the analogue of an intersection, namely a “common information” random variable that has the entropy of and is determined either by or by .

I do not know if there is any simpler model of Shannon entropy that captures all the inequalities available for four variables. One significant complication is that there exist some information inequalities in this setting that are not of Shannon type, such as the Zhang-Yeung inequality

One can however still use these simpler models of Shannon entropy to be able to guess arguments that would work for general random variables. An example of this comes from my paper on the logarithmically averaged Chowla conjecture, in which I showed among other things that

whenever was sufficiently large depending on , where is the Liouville function. The information-theoretic part of the proof was as follows. Given some intermediate scale between and , one can form certain random variables . The random variable is a sign pattern of the form where is a random number chosen from to (with logarithmic weighting). The random variable was tuple of reductions of to primes comparable to . Roughly speaking, what was implicitly shown in the paper (after using the multiplicativity of , the circle method, and the Matomaki-Radziwill theorem on short averages of multiplicative functions) is that if the inequality (5) fails, then there was a lower bound

on the mutual information between and . From translation invariance, this also gives the more general lower bound

for any , where denotes the shifted sign pattern . On the other hand, one had the entropy bounds

and from concatenating sign patterns one could see that is equivalent to the joint random variable for any . Applying these facts and using an “entropy decrement” argument, I was able to obtain a contradiction once was allowed to become sufficiently large compared to , but the bound was quite weak (coming ultimately from the unboundedness of as the interval of values of under consideration becomes large), something of the order of ; the quantity needs at various junctures to be less than a small power of , so the relationship between and becomes essentially quadruple exponential in nature, . The basic strategy was to observe that the lower bound (6) causes some slowdown in the growth rate of the mean entropy, in that this quantity decreased by as increased from to , basically by dividing into components , and observing from (6) each of these shares a bit of common information with the same variable . This is relatively clear when one works in a set model, in which is modeled by a set of size , and is modeled by a set of the form

for various sets of size (also there is some translation symmetry that maps to a shift while preserving all of the ).

However, on considering the set model recently, I realised that one can be a little more efficient by exploiting the fact (basically the Chinese remainder theorem) that the random variables are basically jointly independent as ranges over dyadic values that are much smaller than , which in the set model corresponds to the all being disjoint. One can then establish a variant

of (6), which in the set model roughly speaking asserts that each claims a portion of the of cardinality that is not claimed by previous choices of . This leads to a more efficient contradiction (relying on the unboundedness of rather than ) that looks like it removes one order of exponential growth, thus the relationship between and is now . Returning to the entropy model, one can use (7) and Shannon inequalities to establish an inequality of the form

for a small constant , which on iterating and using the boundedness of gives the claim. (A modification of this analysis, at least on the level of the back of the envelope calculation, suggests that the Matomaki-Radziwill theorem is needed only for ranges greater than or so, although at this range the theorem is not significantly simpler than the general case).

Entropy and rare events

Let and be two random variables taking values in the same (discrete) range , and let be some subset of , which we think of as the set of “bad” outcomes for either or . If and have the same probability distribution, then clearly

In particular, if it is rare for to lie in , then it is also rare for to lie in .

If and do not have exactly the same probability distribution, but their probability distributions are close to each other in some sense, then we can expect to have an approximate version of the above statement. For instance, from the definition of the total variation distance between two random variables (or more precisely, the total variation distance between the probability distributions of two random variables), we see that

for any . In particular, if it is rare for to lie in , and are close in total variation, then it is also rare for to lie in .

A basic inequality in information theory is Pinsker’s inequality

where the Kullback-Leibler divergence is defined by the formula

(See this previous blog post for a proof of this inequality.) A standard application of Jensen’s inequality reveals that is non-negative (Gibbs’ inequality), and vanishes if and only if , have the same distribution; thus one can think of as a measure of how close the distributions of and are to each other, although one should caution that this is not a symmetric notion of distance, as in general. Inserting Pinsker’s inequality into (1), we see for instance that

Thus, if is close to in the Kullback-Leibler sense, and it is rare for to lie in , then it is rare for to lie in as well.

We can specialise this inequality to the case when a uniform random variable on a finite range of some cardinality , in which case the Kullback-Leibler divergence simplifies to

where

is the Shannon entropy of . Again, a routine application of Jensen’s inequality shows that , with equality if and only if is uniformly distributed on . The above inequality then becomes

Thus, if is a small fraction of (so that it is rare for to lie in ), and the entropy of is very close to the maximum possible value of , then it is rare for to lie in also.

The inequality (2) is only useful when the entropy is close to in the sense that , otherwise the bound is worse than the trivial bound of . In my recent paper on the Chowla and Elliott conjectures, I ended up using a variant of (2) which was still non-trivial when the entropy was allowed to be smaller than . More precisely, I used the following simple inequality, which is implicit in the arguments of that paper but which I would like to make more explicit in this post:

Lemma 1 (Pinsker-type inequality) Let be a random variable taking values in a finite range of cardinality , let be a uniformly distributed random variable in , and let be a subset of . Then

Proof: Consider the conditional entropy . On the one hand, we have

by Jensen’s inequality. On the other hand, one has

where we have again used Jensen’s inequality. Putting the two inequalities together, we obtain the claim.

Remark 2 As noted in comments, this inequality can be viewed as a special case of the more general inequality

for arbitrary random variables taking values in the same discrete range , which follows from the data processing inequality

for arbitrary functions , applied to the indicator function . Indeed one has

where is the entropy function.

Thus, for instance, if one has

and

for some much larger than (so that ), then

More informally: if the entropy of is somewhat close to the maximum possible value of , and it is exponentially rare for a uniform variable to lie in , then it is still somewhat rare for to lie in . The estimate given is close to sharp in this regime, as can be seen by calculating the entropy of a random variable which is uniformly distributed inside a small set with some probability and uniformly distributed outside of with probability , for some parameter .

It turns out that the above lemma combines well with concentration of measure estimates; in my paper, I used one of the simplest such estimates, namely Hoeffding’s inequality, but there are of course many other estimates of this type (see e.g. this previous blog post for some others). Roughly speaking, concentration of measure inequalities allow one to make approximations such as

with exponentially high probability, where is a uniform distribution and is some reasonable function of . Combining this with the above lemma, we can then obtain approximations of the form

with somewhat high probability, if the entropy of is somewhat close to maximum. This observation, combined with an “entropy decrement argument” that allowed one to arrive at a situation in which the relevant random variable did have a near-maximum entropy, is the key new idea in my recent paper; for instance, one can use the approximation (3) to obtain an approximation of the form

for “most” choices of and a suitable choice of (with the latter being provided by the entropy decrement argument). The left-hand side is tied to Chowla-type sums such as through the multiplicativity of , while the right-hand side, being a linear correlation involving two parameters rather than just one, has “finite complexity” and can be treated by existing techniques such as the Hardy-Littlewood circle method. One could hope that one could similarly use approximations such as (3) in other problems in analytic number theory or combinatorics.

An entropy Plünnecke-Ruzsa inequality

A handy inequality in additive combinatorics is the Plünnecke-Ruzsa inequality:

Theorem 1 (Plünnecke-Ruzsa inequality) Let be finite non-empty subsets of an additive group , such that for all and some scalars . Then there exists a subset of such that .

The proof uses graph-theoretic techniques. Setting , we obtain a useful corollary: if has small doubling in the sense that , then we have for all , where is the sum of copies of .

In a recent paper, I adapted a number of sum set estimates to the entropy setting, in which finite sets such as in are replaced with discrete random variables taking values in , and (the logarithm of) cardinality of a set is replaced by Shannon entropy of a random variable . (Throughout this note I assume all entropies to be finite.) However, at the time, I was unable to find an entropy analogue of the Plünnecke-Ruzsa inequality, because I did not know how to adapt the graph theory argument to the entropy setting.

I recently discovered, however, that buried in a classic paper of Kaimonovich and Vershik (implicitly in Proposition 1.3, to be precise) there was the following analogue of Theorem 1:

Theorem 2 (Entropy Plünnecke-Ruzsa inequality) Let be independent random variables of finite entropy taking values in an additive group , such that for all and some scalars . Then .

In fact Theorem 2 is a bit “better” than Theorem 1 in the sense that Theorem 1 needed to refine the original set to a subset , but no such refinement is needed in Theorem 2. One corollary of Theorem 2 is that if , then for all , where are independent copies of ; this improves slightly over the analogous combinatorial inequality. Indeed, the function is concave (this can be seen by using the version of Theorem 2 (or (2) below) to show that the quantity is decreasing in ).

Theorem 2 is actually a quick consequence of the submodularity inequality

in information theory, which is valid whenever are discrete random variables such that and each determine (i.e. is a function of , and also a function of ), and and jointly determine (i.e is a function of and ). To apply this, let be independent discrete random variables taking values in . Observe that the pairs and each determine , and jointly determine . Applying (1) we conclude that

which after using the independence of simplifies to the sumset submodularity inequality

(this inequality was also recently observed by Madiman; it is the case of Theorem 2). As a corollary of this inequality, we see that if , then

and Theorem 2 follows by telescoping series.

The proof of Theorem 2 seems to be genuinely different from the graph-theoretic proof of Theorem 1. It would be interesting to see if the above argument can be somehow adapted to give a stronger version of Theorem 1. Note also that both Theorem 1 and Theorem 2 have extensions to more general combinations of than ; see this paper and this paper respectively.

More Mahler lectures

I am posting here four more of my Mahler lectures, each of which is based on earlier talks of mine:

• Compressed sensing.  This is an updated and reformatted version of my ANZIAM talk on this topic.
• Discrete random matrices.  This talk is a survey on recent developments on the universality phenomenon in random matrices, including work of myself and Van Vu.  It covers similar material to my Netanyahu lecture, which has not previously appeared in electronic form.
• Recent progress in additive prime number theory.  This is an updated and reformatted version of my AMS lecture on this topic.
• Recent progress on the Kakeya conjecture.  This is an updated and reformatted version of my Fefferman conference lecture.

As always, comments, corrections, and other feedback are welcome.

Moser’s entropy compression argument

There are many situations in combinatorics in which one is running some sort of iteration algorithm to continually “improve” some object ; each loop of the algorithm replaces with some better version of itself, until some desired property of is attained and the algorithm halts. In order for such arguments to yield a useful conclusion, it is often necessary that the algorithm halts in a finite amount of time, or (even better), in a bounded amount of time. (In general, one cannot use infinitary iteration tools, such as transfinite induction or Zorn’s lemma, in combinatorial settings, because the iteration processes used to improve some target object often degrade some other finitary quantity in the process, and an infinite iteration would then have the undesirable effect of making infinite.)

A basic strategy to ensure termination of an algorithm is to exploit a monotonicity property, or more precisely to show that some key quantity keeps increasing (or keeps decreasing) with each loop of the algorithm, while simultaneously staying bounded. (Or, as the economist Herbert Stein was fond of saying, “If something cannot go on forever, it must stop.”)

Here are four common flavours of this monotonicity strategy:

• The mass increment argument. This is perhaps the most familiar way to ensure termination: make each improved object “heavier” than the previous one by some non-trivial amount (e.g. by ensuring that the cardinality of is strictly greater than that of , thus ). Dually, one can try to force the amount of “mass” remaining “outside” of in some sense to decrease at every stage of the iteration. If there is a good upper bound on the “mass” of that stays essentially fixed throughout the iteration process, and a lower bound on the mass increment at each stage, then the argument terminates. Many “greedy algorithm” arguments are of this type. The proof of the Hahn decomposition theorem in measure theory also falls into this category. The general strategy here is to keep looking for useful pieces of mass outside of , and add them to to form , thus exploiting the additivity properties of mass. Eventually no further usable mass remains to be added (i.e. is maximal in some sense), and this should force some desirable property on .
• The density increment argument. This is a variant of the mass increment argument, in which one increments the “density” of rather than the “mass”. For instance, might be contained in some ambient space , and one seeks to improve to (and to ) in such a way that the density of the new object in the new ambient space is better than that of the previous object (e.g. for some ). On the other hand, the density of is clearly bounded above by . As long as one has a sufficiently good lower bound on the density increment at each stage, one can conclude an upper bound on the number of iterations in the algorithm. The prototypical example of this is Roth’s proof of his theorem that every set of integers of positive upper density contains an arithmetic progression of length three. The general strategy here is to keep looking for useful density fluctuations inside , and then “zoom in” to a region of increased density by reducing and appropriately. Eventually no further usable density fluctuation remains (i.e. is uniformly distributed), and this should force some desirable property on .
• The energy increment argument. This is an “” analogue of the ““-based mass increment argument (or the ““-based density increment argument), in which one seeks to increments the amount of “energy” that captures from some reference object , or (equivalently) to decrement the amount of energy of which is still “orthogonal” to . Here and are related somehow to a Hilbert space, and the energy involves the norm on that space. A classic example of this type of argument is the existence of orthogonal projections onto closed subspaces of a Hilbert space; this leads among other things to the construction of conditional expectation in measure theory, which then underlies a number of arguments in ergodic theory, as discussed for instance in this earlier blog post. Another basic example is the standard proof of the Szemerédi regularity lemma (where the “energy” is often referred to as the “index”). These examples are related; see this blog post for further discussion. The general strategy here is to keep looking for useful pieces of energy orthogonal to , and add them to to form , thus exploiting square-additivity properties of energy, such as Pythagoras’ theorem. Eventually, no further usable energy outside of remains to be added (i.e. is maximal in some sense), and this should force some desirable property on .
• The rank reduction argument. Here, one seeks to make each new object to have a lower “rank”, “dimension”, or “order” than the previous one. A classic example here is the proof of the linear algebra fact that given any finite set of vectors, there exists a linearly independent subset which spans the same subspace; the proof of the more general Steinitz exchange lemma is in the same spirit. The general strategy here is to keep looking for “collisions” or “dependencies” within , and use them to collapse to an object of lower rank. Eventually, no further usable collisions within remain, and this should force some desirable property on .

Much of my own work in additive combinatorics relies heavily on at least one of these types of arguments (and, in some cases, on a nested combination of two or more of them). Many arguments in nonlinear partial differential equations also have a similar flavour, relying on various monotonicity formulae for solutions to such equations, though the objective in PDE is usually slightly different, in that one wants to keep control of a solution as one approaches a singularity (or as some time or space coordinate goes off to infinity), rather than to ensure termination of an algorithm. (On the other hand, many arguments in the theory of concentration compactness, which is used heavily in PDE, does have the same algorithm-terminating flavour as the combinatorial arguments; see this earlier blog post for more discussion.)

Recently, a new species of monotonicity argument was introduced by Moser, as the primary tool in his elegant new proof of the Lovász local lemma. This argument could be dubbed an entropy compression argument, and only applies to probabilistic algorithms which require a certain collection of random “bits” or other random choices as part of the input, thus each loop of the algorithm takes an object (which may also have been generated randomly) and some portion of the random string to (deterministically) create a better object (and a shorter random string , formed by throwing away those bits of that were used in the loop). The key point is to design the algorithm to be partially reversible, in the sense that given and and some additional data that logs the cumulative history of the algorithm up to this point, one can reconstruct together with the remaining portion not already contained in . Thus, each stage of the argument compresses the information-theoretic content of the string into the string in a lossless fashion. However, a random variable such as cannot be compressed losslessly into a string of expected size smaller than the Shannon entropy of that variable. Thus, if one has a good lower bound on the entropy of , and if the length of is significantly less than that of (i.e. we need the marginal growth in the length of the history file per iteration to be less than the marginal amount of randomness used per iteration), then there is a limit as to how many times the algorithm can be run, much as there is a limit as to how many times a random data file can be compressed before no further length reduction occurs.

It is interesting to compare this method with the ones discussed earlier. In the previous methods, the failure of the algorithm to halt led to a new iteration of the object which was “heavier”, “denser”, captured more “energy”, or “lower rank” than the previous instance of . Here, the failure of the algorithm to halt leads to new information that can be used to “compress” (or more precisely, the full state ) into a smaller amount of space. I don’t know yet of any application of this new type of termination strategy to the fields I work in, but one could imagine that it could eventually be of use (perhaps to show that solutions to PDE with sufficiently “random” initial data can avoid singularity formation?), so I thought I would discuss it here.

Below the fold I give a special case of Moser’s argument, based on a blog post of Lance Fortnow on this topic.

P=NP, relativisation, and multiple choice exams

The most fundamental unsolved problem in complexity theory is undoubtedly the P=NP problem, which asks (roughly speaking) whether a problem which can be solved by a non-deterministic polynomial-time (NP) algorithm, can also be solved by a deterministic polynomial-time (P) algorithm. The general belief is that , i.e. there exist problems which can be solved by non-deterministic polynomial-time algorithms but not by deterministic polynomial-time algorithms.

One reason why the question is so difficult to resolve is that a certain generalisation of this question has an affirmative answer in some cases, and a negative answer in other cases. More precisely, if we give all the algorithms access to an oracle, then for one choice of this oracle, all the problems that are solvable by non-deterministic polynomial-time algorithms that calls (), can also be solved by a deterministic polynomial-time algorithm algorithm that calls (), thus ; but for another choice of this oracle, there exist problems solvable by non-deterministic polynomial-time algorithms that call , which cannot be solved by a deterministic polynomial-time algorithm that calls , thus . One particular consequence of this result (which is due to Baker, Gill, and Solovay) is that there cannot be any relativisable proof of either or , where “relativisable” means that the proof would also work without any changes in the presence of an oracle.

The Baker-Gill-Solovay result was quite surprising, but the idea of the proof turns out to be rather simple. To get an oracle such that , one basically sets to be a powerful simulator that can simulate non-deterministic machines (and, furthermore, can also simulate itself); it turns out that any PSPACE-complete oracle would suffice for this task. To get an oracle for which , one has to be a bit sneakier, setting to be a query device for a sparse set of random (or high-complexity) strings, which are too complex to be guessed at by any deterministic polynomial-time algorithm.

Unfortunately, the simple idea of the proof can be obscured by various technical details (e.g. using Turing machines to define and precisely), which require a certain amount of time to properly absorb. To help myself try to understand this result better, I have decided to give a sort of “allegory” of the proof, based around a (rather contrived) story about various students trying to pass a multiple choice test, which avoids all the technical details but still conveys the basic ideas of the argument. This allegory was primarily for my own benefit, but I thought it might also be of interest to some readers here (and also has some tangential relation to the proto-polymath project of determinstically finding primes), so I reproduce it below.

Reflections on compressed sensing

[This post should have appeared several months ago, but I didn’t have a link to the newsletter at the time, and I subsequently forgot about it until now.  -T.]

Last year, Emmanuel Candés and I were two of the recipients of the 2008 IEEE Information Theory Society Paper Award, for our paper “Near-optimal signal recovery from random projections: universal encoding strategies?” published in IEEE Inf. Thy..  (The other recipient is David Donoho, for the closely related paper “Compressed sensing” in the same journal.)  These papers helped initiate the modern subject of compressed sensing, which I have talked about earlier on this blog, although of course they also built upon a number of important precursor results in signal recovery, high-dimensional geometry, Fourier analysis, linear programming, and probability.  As part of our response to this award, Emmanuel and I wrote a short piece commenting on these developments, entitled “Reflections on compressed sensing“, which appears in the Dec 2008 issue of the IEEE Information Theory newsletter.  In it we place our results in the context of these precursor results, and also mention some of the many active directions (theoretical, numerical, and applied) that compressed sensing is now developing in.