DOCSLIB.ORG

Learning the Irreducible Representations of Commutative Lie Groups Groups Only

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Learning the Irreducible Representations of Commutative Lie Groups Groups Only Learning the Irreducible Representations of Commutative Lie Groups Taco Cohen [email protected] Max Welling [email protected] Machine Learning Group, University of Amsterdam Abstract for meaning for future work. We present a new probabilistic model of compact What do we mean, intuitively, when we speak of invari- commutative Lie groups that produces invariant- ance and disentangling? A disentangled representation is equivariant and disentangled representations of one that explicitly represents the distinct factors of varia- data. To define the notion of disentangling, we tion in the data. For example, visual data (i.e. pixels) can borrow a fundamental principle from physics that be thought of as a composition of object identity, position is used to derive the elementary particles of a sys- and pose, lighting conditions, etc. Once disentangling is tem from its symmetries. Our model employs a achieved, invariance follows easily: to build a representa- newfound Bayesian conjugacy relation that en- tion that is invariant to the transformation of a factor of vari- ables fully tractable probabilistic inference over ation (e.g. object position) that is considered a nuisance for compact commutative Lie groups – a class that a particular task (e.g. object classification), one can sim- includes the groups that describe the rotation and ply ignore the units in the representation that encode the cyclic translation of images. We train the model nuisance factor. on pairs of transformed image patches, and show To get a mathematical handle on the concept of disentan- that the learned invariant representation is highly gling, we borrow a fundamental principle from physics, effective for classification. which we refer to as Weyl’s principle, following Kanatani (1990). In physics, this idea is used to tease apart (i.e. disentangle) the elementary particles of a physical system 1. Introduction from mere measurement values that have no inherent phys- Recently, the field of deep learning has produced some re- ical significance. We apply this principle to the area of vi- markable breakthroughs. The hallmark of the deep learn- sion, for after all, pixels are nothing but physical measure- ing approach is to learn multiple layers of representa- ments. tion of data, and much work has gone into the develop- Weyl’s principle presupposes a symmetry group that acts ment of representation learning modules such as RBMs on the data. By this we mean a set of transformations and their generalizations (Welling et al., 2005), and autoen- that does not change the “essence” of the measured phe- coders (Vincent et al., 2008). However, at this point it is not nomenon, although it may change the “superficial appear- arXiv:1402.4437v2 [cs.LG] 25 May 2014 quite clear what characterizes a good representation. In this ance”, i.e. the measurement values. As a concrete example paper, we take a fresh look at the basic principles behind that we will use throughout this paper, consider the group unsupervised representation learning from the perspective known as SO(2), acting on images by 2D rotation about the of Lie group theory1. origin. A transformation from this group (a rotation) may Various desiderata for learned representations have change the value of every pixel in the image, but leaves in- been expressed: representations should be meaning- variant the identity of the imaged object. Weyl’s principle ful (Bengio & Lecun, 2014), invariant (Goodfellow et al., states that the elementary components of this system are 2009), abstract and disentangled (Bengio et al., 2013), but given by the irreducible representations of the symmetry so far most of these notions have not been defined in a group – a concept that will be explained in this paper. mathematically precise way. Here we focus on the no- Although this theoretical principle is widely applicable, tions of invariance and disentangling, leaving the search we demonstrate it for real-valued compact commutative Proceedings of the 31 st International Conference on Machine 1. We will at times assume a passing familiarity with Lie Learning, Beijing, China, 2014. JMLR: W&CP volume 32. Copy- groups, but the main ideas of this paper should be accessible to right 2014 by the author(s). a broad audience. Learning the Irreducible Representations of Commutative Lie Groups groups only. We introduce a probabilistic model that de- 2. Preliminaries scribes a representation of such a group, and show how it can be learned from pairs of images related by arbitrary and 2.1. Equivalence, Invariance and Reducibility unobserved transformations in the group. Compact com- In this section, we discuss three fundamental concepts on mutative groups are also known as toroidal groups, so we which the analysis in the rest of this paper is based: equiv- refer to this model as Toroidal Subgroup Analysis (TSA). alence, invariance and reducibility. Using a novel conjugate prior, the model integrates proba- RD bility theory and Lie group theory in a very elegant way. Consider a function Φ: X that assigns to each pos- RD → All the relevant probabilistic quantities such as normal- sible data point x a class-label (X = 1,...,L ) ∈ R{L } ization constants, moments, KL-divergences, the posterior or some distributed representation (e.g. X = ). Such a density over the transformation group, the marginal density function induces an equivalence relation on the input space RD: we say that two vectors x, y RD are Φ-equivalent if in data space, and their gradients can be obtained in closed ∈ form. they are mapped onto the same representation by Φ. Sym- bolically, x y Φ(x)=Φ(y). ≡Φ ⇔ 1.1. Related work Every equivalence relation on the input space fully de- termines a symmetry group acting on the space. This The first to propose a model and algorithm for learning group, call it G, contains all invertible transformations Lie group representations from data were Rao & Ruder- ρ : RD RD that leave Φ invariant: G = ρ x man (1999). This model deals only with one-parameter RD : Φ(ρ→(x)) = Φ(x) . G describes the symmetries{ | ∀ of∈ groups, a limitation that was later lifted by Miao and Rao Φ, or, stated differently, the} label function/representation Φ (2007). Both works rely on MAP-inference proceduresthat is invariant to transformations in G. Hence, we can speak can only deal with infinitesimally small transformations. of G-equivalence: x y ρ G : ρ(x) = y. For This problem was solved by Sohl-Dickstein et al. (2010) G example, if some elements≡ of⇔G ∃act∈ by rotating the image, using an elegant adaptive smoothing technique, making it two images are G-equivalent if they are rotations of each possible to learn from large transformations. This model other. uses a general linear transformation to diagonalize a one- parameter group, and combines multiple one-parameter Before we can introduce Weyl’s principle, we need one groups multiplicatively. more concept: the reduction of a group representation (Kanatani, 1990). Let us restrict our attention to linear Other, non-group-theoretical approaches to learning representations of Lie groups: ρ becomes a matrix-valued transformations and invariant representations ex- function ρ of an abstract group element g G, such that ist (Memisevic & Hinton, 2010). These gating models g ∈ g,h G : ρ ◦ = ρ ρ . In general, every coordinate y were found to perform a kind of joint eigenspace analy- g h g h i of∀ y =∈ρ x can depend on every coordinate x of x. Now, sis (Memisevic, 2012), which is somewhat similar to the g j since x is G-equivalent to y, it makes no sense to consider irreducible reduction of a toroidal group. the coordinates xi as separate quantities; we can only con- Motivated by a number of statistical phenomena observed sider the vector x as a single unit because the symmetry in natural images, Cadieu & Olshausen (2012) describe a transformations ρg tangle all coordinates. In other words, model that decomposes a signal into invariant amplitudes we cannot say that coordinate xi is an independent part of ′ and covariant phase variables. the aggregate x, because a mapping x x = ρgx that is supposed to leave the intrinsic properties→ of x unchanged, None of the mentioned methods take into account the full will in fact induce induce a functional dependence between uncertainty over transformation parameters, as does TSA. all supposed parts x′ and x . Due to exact or approximate symmetries in the data, there i j is in general no unique transformation relating two images, However, we are free to change the basis of the measure- so that only a multimodal posterior distribution over the ment space. It may be possible to use a change of basis to group gives a complete description of the geometric situa- expose an invariant subspace, i.e. a subspace V RD that tion. Furthermore, posterior inference in our model is per- is mapped onto itself by every transformation in⊂ the group: formed by a very fast feed-forward procedure, whereas the g G : x V ρgx V . If such a subspace ex- MAP inference algorithm by Sohl-Dicksteint et al. requires ists∀ ∈ and its orthogonal∈ ⇒ complement∈ V ⊥ RD is also an a more expensive iterative optimization. invariant subspace, then it makes sense to⊂ consider the two parts of x that lie in V and V ⊥ to be distinct, because they remain distinct under symmetry transformations. Let W be a change of basis matrix that exposes the invari- Learning the Irreducible Representations of Commutative Lie Groups ant subspaces, that is, the total energy (norm) of the image, they can be repre- 1 sented by orthogonal matrices acting on vectorized images. ρg −1 ρg = W 2 W , (1) ρg As is well known, commuting matrices can be simulta- neously diagonalized, so one could represent a toroidal for all g G.
Load more

Recommended publications
  • Notes 9: Eli Cartan's Theorem on Maximal Tori
    Notes 9: Eli Cartan’s theorem on maximal tori. Version 1.00 — still with misprints, hopefully fewer Tori Let T be a torus of dimension n. This means that there is an isomorphism T S1 S1 ... S1. Recall that the Lie algebra Lie T is trivial and that the exponential × map× is× a group homomorphism, so there is an exact sequence of Lie groups exp 0 N Lie T T T 1 where the kernel NT of expT is a discrete subgroup Lie T called the integral lattice of T . The isomorphism T S1 S1 ... S1 induces an isomorphism Lie T Rn under which the exponential map× takes× the× form 2πit1 2πit2 2πitn exp(t1,...,tn)=(e ,e ,...,e ). Irreducible characters. Any irreducible representation V of T is one dimensio- nal and hence it is given by a multiplicative character; that is a group homomorphism 1 χ: T Aut(V )=C∗. It takes values in the unit circle S — the circle being the → only compact and connected subgroup of C∗ — hence we may regard χ as a Lie group map χ: T S1. → The character χ has a derivative at the unit which is a map θ = d χ :LieT e → Lie S1 of Lie algebras. The two Lie algebras being trivial and Lie S1 being of di- 1 mension one, this is just a linear functional on Lie T . The tangent space TeS TeC∗ = ⊆ C equals the imaginary axis iR, which we often will identify with R. The derivative θ fits into the commutative diagram exp 0 NT Lie T T 1 θ N θ χ | T exp 1 1 0 NS1 Lie S S 1 The linear functional θ is not any linear functional, the values it takes on the discrete 1 subgroup NT are all in N 1 .
    [Show full text]
  • Simultaneous Diagonalization and SVD of Commuting Matrices
    Simultaneous Diagonalization and SVD of Commuting Matrices Ronald P. Nordgren1 Brown School of Engineering, Rice University Abstract. We present a matrix version of a known method of constructing common eigenvectors of two diagonalizable commuting matrices, thus enabling their simultane- ous diagonalization. The matrices may have simple eigenvalues of multiplicity greater than one. The singular value decomposition (SVD) of a class of commuting matrices also is treated. The effect of row/column permutation is examined. Examples are given. 1 Introduction It is well known that if two diagonalizable matrices have the same eigenvectors, then they commute. The converse also is true and a construction for the common eigen- vectors (enabling simultaneous diagonalization) is known. If one of the matrices has distinct eigenvalues (multiplicity one), it is easy to show that its eigenvectors pertain to both commuting matrices. The case of matrices with simple eigenvalues of mul- tiplicity greater than one requires a more complicated construction of their common eigenvectors. Here, we present a matrix version of a construction procedure given by Horn and Johnson [1, Theorem 1.3.12] and in a video by Sadun [4]. The eigenvector construction procedure also is applied to the singular value decomposition of a class of commuting matrices that includes the case where at least one of the matrices is real and symmetric. In addition, we consider row/column permutation of the commuting matrices. Three examples illustrate the eigenvector construction procedure. 2 Eigenvector Construction Let A and B be diagonalizable square matrices that commute, i.e. AB = BA (1) and their Jordan canonical forms read −1 −1 A = SADASA , B = SBDB SB .
    [Show full text]
  • Math 210C. Simple Factors 1. Introduction Let G Be a Nontrivial Connected Compact Lie Group, and Let T ⊂ G Be a Maximal Torus
    Math 210C. Simple factors 1. Introduction Let G be a nontrivial connected compact Lie group, and let T ⊂ G be a maximal torus. 0 Assume ZG is finite (so G = G , and the converse holds by Exercise 4(ii) in HW9). Define Φ = Φ(G; T ) and V = X(T )Q, so (V; Φ) is a nonzero root system. By the handout \Irreducible components of root systems", (V; Φ) is uniquely a direct sum of irreducible components f(Vi; Φi)g in the following sense. There exists a unique collection of nonzero Q-subspaces Vi ⊂ V such that for Φi := Φ \ Vi the following hold: each pair ` (Vi; Φi) is an irreducible root system, ⊕Vi = V , and Φi = Φ. Our aim is to use the unique irreducible decomposition of the root system (which involves the choice of T , unique up to G-conjugation) and some Exercises in HW9 to prove: Theorem 1.1. Let fGjgj2J be the set of minimal non-trivial connected closed normal sub- groups of G. (i) The set J is finite, the Gj's pairwise commute, and the multiplication homomorphism Y Gj ! G is an isogeny. Q (ii) If ZG = 1 or π1(G) = 1 then Gi ! G is an isomorphism (so ZGj = 1 for all j or π1(Gj) = 1 for all j respectively). 0 (iii) Each connected closed normal subgroup N ⊂ G has finite center, and J 7! GJ0 = hGjij2J0 is a bijection between the set of subsets of J and the set of connected closed normal subgroups of G. (iv) The set fGjg is in natural bijection with the set fΦig via j 7! i(j) defined by the condition T \ Gj = Ti(j).
    [Show full text]
  • The Algebra Generated by Three Commuting Matrices
    THE ALGEBRA GENERATED BY THREE COMMUTING MATRICES B.A. SETHURAMAN Abstract. We present a survey of an open problem concerning the dimension of the algebra generated by three commuting matrices. This article concerns a problem in algebra that is completely elementary to state, yet, has proven tantalizingly difficult and is as yet unsolved. Consider C[A; B; C] , the C-subalgebra of the n × n matrices Mn(C) generated by three commuting matrices A, B, and C. Thus, C[A; B; C] consists of all C- linear combinations of \monomials" AiBjCk, where i, j, and k range from 0 to infinity. Note that C[A; B; C] and Mn(C) are naturally vector-spaces over C; moreover, C[A; B; C] is a subspace of Mn(C). The problem, quite simply, is this: Is the dimension of C[A; B; C] as a C vector space bounded above by n? 2 Note that the dimension of C[A; B; C] is at most n , because the dimen- 2 sion of Mn(C) is n . Asking for the dimension of C[A; B; C] to be bounded above by n when A, B, and C commute is to put considerable restrictions on C[A; B; C]: this is to require that C[A; B; C] occupy only a small portion of the ambient Mn(C) space in which it sits. Actually, the dimension of C[A; B; C] is already bounded above by some- thing slightly smaller than n2, thanks to a classical theorem of Schur ([16]), who showed that the maximum possible dimension of a commutative C- 2 subalgebra of Mn(C) is 1 + bn =4c.
    [Show full text]
  • Arxiv:1705.10957V2 [Math.AC] 14 Dec 2017 Ideal Called Is and Commutator X Y Vrafield a Over Let Matrix Eae Leri Es Is E Sdfierlvn Notions
    NEARLY COMMUTING MATRICES ZHIBEK KADYRSIZOVA Abstract. We prove that the algebraic set of pairs of matrices with a diag- onal commutator over a field of positive prime characteristic, its irreducible components, and their intersection are F -pure when the size of matrices is equal to 3. Furthermore, we show that this algebraic set is reduced and the intersection of its irreducible components is irreducible in any characteristic for pairs of matrices of any size. In addition, we discuss various conjectures on the singularities of these algebraic sets and the system of parameters on the corresponding coordinate rings. Keywords: Frobenius, singularities, F -purity, commuting matrices 1. Introduction and preliminaries In this paper we study algebraic sets of pairs of matrices such that their commutator is either nonzero diagonal or zero. We also consider some other related algebraic sets. First let us define relevant notions. Let X =(x ) and Y =(y ) be n×n matrices of indeterminates arXiv:1705.10957v2 [math.AC] 14 Dec 2017 ij 1≤i,j≤n ij 1≤i,j≤n over a field K. Let R = K[X,Y ] be the polynomial ring in {xij,yij}1≤i,j≤n and let I denote the ideal generated by the off-diagonal entries of the commutator matrix XY − YX and J denote the ideal generated by the entries of XY − YX. The ideal I defines the algebraic set of pairs of matrices with a diagonal commutator and is called the algebraic set of nearly commuting matrices. The ideal J defines the algebraic set of pairs of commuting matrices.
    [Show full text]
  • Lecture 2: Spectral Theorems
    Lecture 2: Spectral Theorems This lecture introduces normal matrices. The spectral theorem will inform us that normal matrices are exactly the unitarily diagonalizable matrices. As a consequence, we will deduce the classical spectral theorem for Hermitian matrices. The case of commuting families of matrices will also be studied. All of this corresponds to section 2.5 of the textbook. 1 Normal matrices Definition 1. A matrix A 2 Mn is called a normal matrix if AA∗ = A∗A: Observation: The set of normal matrices includes all the Hermitian matrices (A∗ = A), the skew-Hermitian matrices (A∗ = −A), and the unitary matrices (AA∗ = A∗A = I). It also " # " # 1 −1 1 1 contains other matrices, e.g. , but not all matrices, e.g. 1 1 0 1 Here is an alternate characterization of normal matrices. Theorem 2. A matrix A 2 Mn is normal iff ∗ n kAxk2 = kA xk2 for all x 2 C : n Proof. If A is normal, then for any x 2 C , 2 ∗ ∗ ∗ ∗ ∗ 2 kAxk2 = hAx; Axi = hx; A Axi = hx; AA xi = hA x; A xi = kA xk2: ∗ n n Conversely, suppose that kAxk = kA xk for all x 2 C . For any x; y 2 C and for λ 2 C with jλj = 1 chosen so that <(λhx; (A∗A − AA∗)yi) = jhx; (A∗A − AA∗)yij, we expand both sides of 2 ∗ 2 kA(λx + y)k2 = kA (λx + y)k2 to obtain 2 2 ∗ 2 ∗ 2 ∗ ∗ kAxk2 + kAyk2 + 2<(λhAx; Ayi) = kA xk2 + kA yk2 + 2<(λhA x; A yi): 2 ∗ 2 2 ∗ 2 Using the facts that kAxk2 = kA xk2 and kAyk2 = kA yk2, we derive 0 = <(λhAx; Ayi − λhA∗x; A∗yi) = <(λhx; A∗Ayi − λhx; AA∗yi) = <(λhx; (A∗A − AA∗)yi) = jhx; (A∗A − AA∗)yij: n ∗ ∗ n Since this is true for any x 2 C , we deduce (A A − AA )y = 0, which holds for any y 2 C , meaning that A∗A − AA∗ = 0, as desired.
    [Show full text]
  • Loop Groups and Twisted K-Theory III
    Annals of Mathematics 174 (2011), 947{1007 http://dx.doi.org/10.4007/annals.2011.174.2.5 Loop groups and twisted K-theory III By Daniel S. Freed, Michael J. Hopkins, and Constantin Teleman Abstract In this paper, we identify the Ad-equivariant twisted K-theory of a compact Lie group G with the \Verlinde group" of isomorphism classes of admissible representations of its loop groups. Our identification pre- serves natural module structures over the representation ring R(G) and a natural duality pairing. Two earlier papers in the series covered founda- tions of twisted equivariant K-theory, introduced distinguished families of Dirac operators and discussed the special case of connected groups with free π1. Here, we recall the earlier material as needed to make the paper self-contained. Going further, we discuss the relation to semi-infinite co- homology, the fusion product of conformal field theory, the r^oleof energy and a topological Peter-Weyl theorem. Introduction Let G be a compact Lie group and LG the space of smooth maps S1 !G, the loop group of G. The latter has a distinguished class of irreducible pro- jective unitary representations, the integrable lowest-weight representations. They are rigid; in fact, if we fix the central extension LGτ of LG there is a finite number of isomorphism classes. Our main theorem identifies the free abelian group Rτ (LG) they generate with a twisted version of the equivari- ant topological K-theory group KG(G), where G acts on itself by conjugation. The twisting is constructed from the central extension of the loop group.
    [Show full text]
  • On Low Rank Perturbations of Complex Matrices and Some Discrete Metric Spaces∗
    Electronic Journal of Linear Algebra ISSN 1081-3810 A publication of the International Linear Algebra Society Volume 18, pp. 302-316, June 2009 ELA ON LOW RANK PERTURBATIONS OF COMPLEX MATRICES AND SOME DISCRETE METRIC SPACES∗ LEV GLEBSKY† AND LUIS MANUEL RIVERA‡ Abstract. In this article, several theorems on perturbations ofa complex matrix by a matrix ofa given rank are presented. These theorems may be divided into two groups. The first group is about spectral properties ofa matrix under such perturbations; the second is about almost-near relations with respect to the rank distance. Key words. Complex matrices, Rank, Perturbations. AMS subject classifications. 15A03, 15A18. 1. Introduction. In this article, we present several theorems on perturbations of a complex matrix by a matrix of a given rank. These theorems may be divided into two groups. The first group is about spectral properties of a matrix under such perturbations; the second is about almost-near relations with respect to the rank distance. 1.1. The first group. Theorem 2.1gives necessary and sufficient conditions on the Weyr characteristics, [20], of matrices A and B if rank(A − B) ≤ k.In one direction the theorem is known; see [12, 13, 14]. For k =1thetheoremisa reformulation of a theorem of Thompson [16] (see Theorem 2.3 of the present article). We prove Theorem 2.1by induction with respect to k. To this end, we introduce a discrete metric space (the space of Weyr characteristics) and prove that this metric space is geodesic. In fact, the induction (with respect to k) is hidden in this proof (see Proposition 3.6 and Proposition 3.1).
    [Show full text]
  • Exceptional Groups of Lie Type: Subgroup Structure and Unipotent Elements
    Exceptional groups of Lie type: subgroup structure and unipotent elements Donna Testerman EPF Lausanne 17 December 2012 Donna Testerman (EPF Lausanne) Exceptional groups of Lie type: subgroup structure and unip17 Decemberotent elements 2012 1 / 110 Outline: I. Maximal subgroups of exceptional groups, finite and algebraic II. Lifting results III. Overgroups of unipotent elements Donna Testerman (EPF Lausanne) Exceptional groups of Lie type: subgroup structure and unip17 Decemberotent elements 2012 2 / 110 Maximal subgroups of exceptional algebraic groups Let G be a simple algebraic group of exceptional type, defined over an algebraically closed field k of characteristic p ≥ 0. Let M ⊂ G be a maximal positive-dimensional closed subgroup. By the Borel-Tits theorem, if M◦ is not reductive, then M is a maximal parabolic subgroup of G. Now if M◦ is reductive, it is possible that M◦ contains a maximal torus of G. It is straightforward to describe subgroups containing maximal tori of G via the Borel-de Siebenthal algorithm. Then M is the full normalizer of such a group, and finding such M which are maximal is again straightfoward. Donna Testerman (EPF Lausanne) Exceptional groups of Lie type: subgroup structure and unip17 Decemberotent elements 2012 3 / 110 So finally, one is left to consider the subgroups M such that M◦ is reductive, and M does not contain a maximal torus of G. A classification of the positive-dimensional maximal closed subgroups of G was completed in 2004 by Liebeck and Seitz. (Earlier work of Seitz had reduced this problem to those cases where char(k) is ‘small’.) Donna Testerman (EPF Lausanne) Exceptional groups of Lie type: subgroup structure and unip17 Decemberotent elements 2012 4 / 110 We may assume the exceptional group G to be an adjoint type group, as maximal subgroups of an arbitrary simple algebraic group G˜ must contain Z(G˜ ) and hence will have an image which is a maximal subgroup of the adjoint group.
    [Show full text]
  • Representation Theory
    M392C NOTES: REPRESENTATION THEORY ARUN DEBRAY MAY 14, 2017 These notes were taken in UT Austin's M392C (Representation Theory) class in Spring 2017, taught by Sam Gunningham. I live-TEXed them using vim, so there may be typos; please send questions, comments, complaints, and corrections to [email protected]. Thanks to Kartik Chitturi, Adrian Clough, Tom Gannon, Nathan Guermond, Sam Gunningham, Jay Hathaway, and Surya Raghavendran for correcting a few errors. Contents 1. Lie groups and smooth actions: 1/18/172 2. Representation theory of compact groups: 1/20/174 3. Operations on representations: 1/23/176 4. Complete reducibility: 1/25/178 5. Some examples: 1/27/17 10 6. Matrix coefficients and characters: 1/30/17 12 7. The Peter-Weyl theorem: 2/1/17 13 8. Character tables: 2/3/17 15 9. The character theory of SU(2): 2/6/17 17 10. Representation theory of Lie groups: 2/8/17 19 11. Lie algebras: 2/10/17 20 12. The adjoint representations: 2/13/17 22 13. Representations of Lie algebras: 2/15/17 24 14. The representation theory of sl2(C): 2/17/17 25 15. Solvable and nilpotent Lie algebras: 2/20/17 27 16. Semisimple Lie algebras: 2/22/17 29 17. Invariant bilinear forms on Lie algebras: 2/24/17 31 18. Classical Lie groups and Lie algebras: 2/27/17 32 19. Roots and root spaces: 3/1/17 34 20. Properties of roots: 3/3/17 36 21. Root systems: 3/6/17 37 22. Dynkin diagrams: 3/8/17 39 23.
    [Show full text]
  • Lecture Notes: Basic Group and Representation Theory
    Lecture notes: Basic group and representation theory Thomas Willwacher February 27, 2014 2 Contents 1 Introduction 5 1.1 Definitions . .6 1.2 Actions and the orbit-stabilizer Theorem . .8 1.3 Generators and relations . .9 1.4 Representations . 10 1.5 Basic properties of representations, irreducibility and complete reducibility . 11 1.6 Schur’s Lemmata . 12 2 Finite groups and finite dimensional representations 15 2.1 Character theory . 15 2.2 Algebras . 17 2.3 Existence and classification of irreducible representations . 18 2.4 How to determine the character table – Burnside’s algorithm . 20 2.5 Real and complex representations . 21 2.6 Induction, restriction and characters . 23 2.7 Exercises . 25 3 Representation theory of the symmetric groups 27 3.1 Notations . 27 3.2 Conjugacy classes . 28 3.3 Irreducible representations . 29 3.4 The Frobenius character formula . 31 3.5 The hook lengths formula . 34 3.6 Induction and restriction . 34 3.7 Schur-Weyl duality . 35 4 Lie groups, Lie algebras and their representations 37 4.1 Overview . 37 4.2 General definitions and facts about Lie algebras . 39 4.3 The theorems of Lie and Engel . 40 4.4 The Killing form and Cartan’s criteria . 41 4.5 Classification of complex simple Lie algebras . 43 4.6 Classification of real simple Lie algebras . 44 4.7 Generalities on representations of Lie algebras . 44 4.8 Representation theory of sl(2; C) ................................ 45 4.9 General structure theory of semi-simple Lie algebras . 48 4.10 Representation theory of complex semi-simple Lie algebras .
    [Show full text]
  • On the Borel Transgression in the Fibration $ G\Rightarrow G/T$
    On the Borel transgression in the fibration G → G/T Haibao Duan∗ Institute of Mathematics, Chinese Academy of Sciences; School of Mathematical Sciences, University of the Chinese Academy of Sciences [email protected] September 16, 2018 Abstract Let G be a semisimple Lie group with a maximal torus T . We present 1 2 an explicit formula for the Borel transgression τ : H (T ) → H (G/T ) of the fibration G → G/T . This formula corrects an error in the paper [9], and has been applied to construct the integral cohomology rings of compact Lie groups in the sequel works [4, 7]. 2010 Mathematical Subject Classification: 55T10, 57T10 Key words and phrases: Lie groups; Cohomology, Leray–Serre spec- tral sequence 1 Introduction A Lie group is called semisimple if its center is finite; is called adjoint if its center is trivial. In this paper the Lie groups G under consideration are compact, connected and semisimple. The homology and cohomology are over the ring of integers, unless otherwise stated. For a Lie group G with a maximal torus T let π : G → G/T be the quotient fibration. Consider the diagram with top row the cohomology exact sequence of the pair (G, T ) arXiv:1710.03014v1 [math.AT] 9 Oct 2017 ∗ i∗ δ j 0 → H1(G) → H1(T ) → H2(G, T ) → H2(G) → · · · ∗ ց τ =∼↑ π H2(G/T ) where, since the pair (G, T ) is 1–connected, the induced map π∗ is an isomor- phism. The Borel transgression [11, p.185] in the fibration π is the composition τ = (π∗)−1 ◦ δ : H1(T ) → H2(G/T ).
    [Show full text]