To save content items to your account,
please confirm that you agree to abide by our usage policies.
If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your account.
Find out more about saving content to .
To save content items to your Kindle, first ensure no-reply@cambridge.org
is added to your Approved Personal Document E-mail List under your Personal Document Settings
on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part
of your Kindle email address below.
Find out more about saving to your Kindle.
Note you can select to save to either the @free.kindle.com or @kindle.com variations.
‘@free.kindle.com’ emails are free but can only be saved to your device when it is connected to wi-fi.
‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.
This chapter covers regression and classification, where the goal is to estimate a quantity of interest (the response) from observed features. In regression, the response is a numerical variable. In classification, it belongs to a finite set of predetermined classes. We begin with a comprehensive description of linear regression and discuss how to leverage it to perform causal inference. Then, we explain under what conditions linear models tend to overfit or to generalize robustly to held-out data. Motivated by the threat of overfitting, we introduce regularization and ridge regression, and discuss sparse regression, where the goal is to fit a linear model that only depends on a small subset of the available features. Then, we introduce two popular linear models for binary and multiclass classification: Logistic and softmax regression. At this point, we turn our attention to nonlinear models. First, we present regression and classification trees and explain how to combine them via bagging, random forests, and boosting. Second, we explain how to train neural networks to perform regression and classification. Finally, we discuss how to evaluate classification models.
We analyze generating functions for trees and for connected subgraphs on the complete graph, and identify a single scaling profile which applies for both generating functions in a critical window. Our motivation comes from the analysis of the finite-size scaling of lattice trees and lattice animals on a high-dimensional discrete torus, for which we conjecture that the identical profile applies in dimensions $d \ge 8$.
By reference to nominated attributes, a genus, being a population of objects of one specified kind, may be partitioned into species, being subpopulations of different kinds. A prototype is an object representative of its species within the genus. Using this framework, the paper describes how objects can be relatively differentiated with respect to attributes, and how attributes can be relatively differentiating with respect to objects. Methods and rationale for such differential ordering of objects and attributes are presented by example, formal development, and application.
For a genus Ω comprising n species of object there is a subset P ofn distinct prototypes. With respect to m nominated attributes, each object in Ω has an m-element characterization. Together these determine an n × m objects × attributes matrix, the rows of which are the characterizations of the prototypical objects. Over then species in Ω, an associated relative frequency vector gives the distribution of objects (and of their characterizations). The matrix and vector associate the objects in Ω with points in a metric space (P, δ); and it is with respect to various sums of distances in this attribute space that one can differentially order objects and attributes.
The definition of the distance function δ is generalized across kinds of difference, types of characterization, scale-types of measurement, Minkowski index ≧ 1, and any form of distribution of objects over species. Explanatory and taxonomic applications in psychology and other fields are discussed, with focus on classification, identification, recognition, and search. The Braille code and the identification of its characters provide illustration.
In this paper we consider positional games where the winning sets are edge sets of tree-universal graphs. Specifically, we show that in the unbiased Maker-Breaker game on the edges of the complete graph $K_n$, Maker has a strategy to claim a graph which contains copies of all spanning trees with maximum degree at most $cn/\log (n)$, for a suitable constant $c$ and $n$ being large enough. We also prove an analogous result for Waiter-Client games. Both of our results show that the building player can play at least as good as suggested by the random graph intuition. Moreover, they improve on a special case of earlier results by Johannsen, Krivelevich, and Samotij as well as Han and Yang for Maker-Breaker games.
We show that for every $\eta \gt 0$ every sufficiently large $n$-vertex oriented graph $D$ of minimum semidegree exceeding $(1+\eta )\frac k2$ contains every balanced antidirected tree with $k$ edges and bounded maximum degree, if $k\ge \eta n$. In particular, this asymptotically confirms a conjecture of the first author for long antidirected paths and dense digraphs.
Further, we show that in the same setting, $D$ contains every $k$-edge antidirected subdivision of a sufficiently small complete graph, if the paths of the subdivision that have length $1$ or $2$ span a forest. As a special case, we can find all antidirected cycles of length at most $k$.
Finally, we address a conjecture of Addario-Berry, Havet, Linhares Sales, Reed, and Thomassé for antidirected trees in digraphs. We show that this conjecture is asymptotically true in $n$-vertex oriented graphs for all balanced antidirected trees of bounded maximum degree and of size linear in $n$.
In this article, we introduce a hierarchy on the class of non-archimedean Polish groups that admit a compatible complete left-invariant metric. We denote this hierarchy by $\alpha $-CLI and L-$\alpha $-CLI where $\alpha $ is a countable ordinal. We establish three results:
(1)G is $0$-CLI iff $G=\{1_G\}$;
(2)G is $1$-CLI iff G admits a compatible complete two-sided invariant metric; and
(3)G is L-$\alpha $-CLI iff G is locally $\alpha $-CLI, i.e., G contains an open subgroup that is $\alpha $-CLI.
Subsequently, we show this hierarchy is proper by constructing non-archimedean CLI Polish groups $G_\alpha $ and $H_\alpha $ for $\alpha <\omega _1$, such that:
(1)$H_\alpha $ is $\alpha $-CLI but not L-$\beta $-CLI for $\beta <\alpha $; and
(2)$G_\alpha $ is $(\alpha +1)$-CLI but not L-$\alpha $-CLI.
We study the possible structures which can be carried by sets which have no countable subset, but which fail to be ‘surjectively Dedekind finite’, in two possible senses, that there is surjection to $\omega $, or alternatively, that there is a surjection to a proper superset.
As green spaces, lawns are often thought to capture carbon from the atmosphere. However, once mowing, fertlising and irrigation are taken into account, we show that they become carbon sources, at least in the long run. Converting unused urban and rural lawn and grassland to treescapes can make a substantial contribution to reducing greenhouse gas emissions and increasing carbon absorption from the atmosphere. However, it is imperative for governing bodies to put in place appropriate policies and incentives in order to achieve this.
Technical summary
Mown grass or lawn is a ubiquitous form of vegetation in human-dominated landscapes and it is often claimed to perform an ecosystem service by sequestering soil carbon. If lawn maintenance is included, however, we show that lawns become net carbon emitters. We estimate that globally, if one-third of mown grass in cities was returned to treescapes, 310–1630 million tonnes of carbon could be absorbed from the atmosphere, and up to 43 tonnes of carbon equivalent per hectare of emissions could be avoided over a two-decade time span. We therefore propose that local and central governments introduce policies to incentivise and/or regulate the conversion of underutilised grass into treescapes.
Social media summary
If unused lawns were planted with trees, a gigaton of carbon could be removed from the atmosphere over two decades.
Let T be the regular tree in which every vertex has exactly $d\ge 3$ neighbours. Run a branching random walk on T, in which at each time step every particle gives birth to a random number of children with mean d and finite variance, and each of these children moves independently to a uniformly chosen neighbour of its parent. We show that, starting with one particle at some vertex 0 and conditionally on survival of the process, the time it takes for every vertex within distance r of 0 to be hit by a particle of the branching random walk is $r + ({2}/{\log(3/2)})\log\log r + {\mathrm{o}}(\log\log r)$.
We study the first-order theories of some natural and important classes of coloured trees, including the four classes of trees whose paths have the order type respectively of the natural numbers, the integers, the rationals, and the reals. We develop a technique for approximating a tree as a suitably coloured linear order. We then present the first-order theories of certain classes of coloured linear orders and use them, along with the approximating technique, to establish complete axiomatisations of the four classes of trees mentioned above.
Vatica cauliflora P.S. Ashton (Dipterocarpaceae) is a threatened tree species endemic to Kapuas Hulu District, West Kalimantan Province, Indonesia. The species is only known from the type specimen collected in 1953. After this first collection, there was no record confirming the presence of this tree in its natural habitat. Our recent surveys in 2019 and 2020 located 179 individuals of the species in six unprotected locations. The population's size structure is dominated (62.6%) by young individuals within the 0–5 cm diameter class. Our surveys also showed that the habitat of V. cauliflora is degraded as a result of the negative effects of agriculture and logging. Assessment with the IUCN Red List criteria indicates that V. cauliflora should be categorized as Critically Endangered.
Here, I deal with the issue of how organisms support themselves against the downward pull of gravity. First, I look at the microscopic skeletons of cells, which help both to maintain the shape of a static cell and to alter that of a mobile one. Then I consider the higher-level skeletons that are needed by large multicellular creatures, whether animals or plants. In the animal kingdom, I focus on the solutions to this issue that we find in the three largest phyla. Vertebrates and arthropods have solved the problem of support in what might be called opposite ways – skeletons on the inside and outside respectively (endo- and exo-skeletons). Molluscs are harder to generalize about. An external shell is the commonest hard structure, but in some groups the shell is reduced and internal. Shells are different from endo- and exo-skeletons in that they have to be carried, so are a mixed blessing from a gravitational perspective. Outside the three main animal groups, other skeletal solutions are found. These range from the hydrostatic skeleton of worms to the woody skeleton of trees. Wood has allowed the evolution of the tallest organisms on the planet – coastal redwoods.
In this paper, we mainly study a class of small deviation theorems for Markov chains indexed by an infinite tree with uniformly bounded degree in Markovian environment. Firstly, we give the definition of Markov chains indexed by a tree with uniformly bounded degree in random environment. Then, we introduce the some lemmas which are the basis of the results. Finally, a class of small deviation theorems for functionals of random fields on a tree with uniformly bounded degree in Markovian environment is established.
We consider a countable tree $T$, possibly having vertices with infinite degree, and an arbitrary stochastic nearest neighbour transition operator $P$. We provide a boundary integral representation for general eigenfunctions of $P$ with eigenvalue $\lambda \in \C$, under the condition that the oriented edges can be equipped with complex-valued weights satisfying three natural axioms. These axioms guarantee that one can construct a $\lambda$-Poisson kernel. The boundary integral is with respect to distributions, that is, elements in the dual of the space of locally constant functions. Distributions are interpreted as finitely additive complex measures. In general, they do not extend to $\sigma$-additive measures: for this extension, a summability condition over disjoint boundary arcs is required. Whenever $\lambda$ is in the resolvent of $P$ as a self-adjoint operator on a naturally associated $\ell^2$-space and the diagonal elements of the resolvent (“Green function”) do not vanish at $\lambda$, one can use the ordinary edge weights corresponding to the Green function and obtain the ordinary $\la$-Martin kernel.
We then consider the case when $P$ is invariant under a transitive group action. In this situation, we study the phenomenon that in addition to the $\lambda$-Martin kernel, there may be further choices for the edge weights which give rise to another $\lambda$-Poisson kernel with associated integral representations. In particular, we compare the resulting distributions on the boundary.
The material presented here is closely related to the contents of our “companion” paper\cite{PiWo}.
Acer yangbiense Y.S. Chen & Q.E. Yang (Aceraceae) is a threatened tree species endemic to China, formerly presumed to have declined to only five extant individuals, restricted to Yangbi County, Yunnan Province. Our surveys in 2016, however, located 577 individuals in 12 localities, but only three localities (with a total of 62 individuals) are protected. Nine localities are on private forest land. The population's size structure is an inverse J-curve, but there is a scarcity of trees of the smallest size class and of seedlings. Our surveys also showed that the habitat of A. yangbiense is degraded as a result of the negative effects of agriculture, logging and wood harvesting. Assessment with the IUCN Red List categories and criteria indicates that A. yangbiense should be recategorized from Critically Endangered to Endangered.
The chapter presents and discusses the graph conception of set. According to the graph conception, sets are things depicted by graphs of a certain sort. The chapter begins by presenting four set theories, due to Aczel, which are formulated by using the notion of a graph. The graph conception is then introduced, and a historical excursion into forerunners of the conception is also given. The chapter continues by clarifying the relationship between the conception and the four theories described by Aczel. It concludes by discussing four objections to the graph conception: the objection that set theories based on graphs do not introduce new isomorphism types; the objection that the graph conception does not provide us with an intuitive model for the set theory it sanctions; the objection that the graph conception cannot naturally allow for Urelemente; and the objection that a set theory based on the graph conception cannot provide an autonomous foundation for mathematics. It is argued that whilst the first two objections fail, the remaining two retain their force.
Let $\unicode[STIX]{x1D6E4}\leqslant \text{Aut}(T_{d_{1}})\times \text{Aut}(T_{d_{2}})$ be a group acting freely and transitively on the product of two regular trees of degree $d_{1}$ and $d_{2}$. We develop an algorithm that computes the closure of the projection of $\unicode[STIX]{x1D6E4}$ on $\text{Aut}(T_{d_{t}})$ under the hypothesis that $d_{t}\geqslant 6$ is even and that the local action of $\unicode[STIX]{x1D6E4}$ on $T_{d_{t}}$ contains $\text{Alt}(d_{t})$. We show that if $\unicode[STIX]{x1D6E4}$ is torsion-free and $d_{1}=d_{2}=6$, exactly seven closed subgroups of $\text{Aut}(T_{6})$ arise in this way. We also construct two new infinite families of virtually simple lattices in $\text{Aut}(T_{6})\times \text{Aut}(T_{4n})$ and in $\text{Aut}(T_{2n})\times \text{Aut}(T_{2n+1})$, respectively, for all $n\geqslant 2$. In particular, we provide an explicit presentation of a torsion-free infinite simple group on 5 generators and 10 relations, that splits as an amalgamated free product of two copies of $F_{3}$ over $F_{11}$. We include information arising from computer-assisted exhaustive searches of lattices in products of trees of small degrees. In an appendix by Pierre-Emmanuel Caprace, some of our results are used to show that abstract and relative commensurator groups of free groups are almost simple, providing partial answers to questions of Lubotzky and Lubotzky–Mozes–Zimmer.
In this paper, we extend the strong laws of large numbers and entropy ergodic theorem for partial sums for tree-indexed nonhomogeneous Markov chains fields to delayed versions of nonhomogeneous Markov chains fields indexed by a homogeneous tree. At first we study a generalized strong limit theorem for nonhomogeneous Markov chains indexed by a homogeneous tree. Then we prove the generalized strong laws of large numbers and the generalized asymptotic equipartition property for delayed sums of finite nonhomogeneous Markov chains indexed by a homogeneous tree. As corollaries, we can get the similar results of some current literatures. In this paper, the problem settings may not allow to use Doob's martingale convergence theorem, and we overcome this difficulty by using Borel–Cantelli Lemma so that our proof technique also has some new elements compared with the reference Yang and Ye (2007).