DOCSLIB.ORG

1. Abelian Categories Most of Homological Algebra Can Be Carried

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
1. Abelian Categories Most of Homological Algebra Can Be Carried 1. Abelian categories Most of homological algebra can be carried out in the setting of abelian categories, a class of categories which includes on the one hand all categories of modules and on the other hand categories of (abelian) sheaves. In this section we give a quick introduction to abelian categories. There are three layers of structure, and we discuss each in succession. 1.1. Pre-additive categories. In a general category C the homsets C(A; B) are simply sets; however, in many specific cases we find that these sets carry some additional structure. For example, consider the category Pos of posets and order-preserving functions. For two posets A; B, the homset Pos(A; B) is again an ordered set if we let f ≤ g , f(x) ≤ g(x) for all x 2 A. Sim- ilarly, in the category Rel of sets and relations we may define an ordering on Rel(A; B) by saying that R ≤ S if and only if R (as a subset of A × B) is contained in S, i.e. if aRb implies aSb. Also, in the category of abelian groups, the homset AbGrp(A; B) is more than a mere set: it is itself an abelian group under the operation of pointwise addition. More generally, in any category of modules we have that R − Mod(A; B) is an abelian group. The phenomenon that categories may have homsets with additional struc- ture goes by the name of enrichment: we say that the category Rel is en- riched in posets, or that the category of R-modules is enriched in abelian groups. We will not go into the theory of enriched categories in general (a standard reference is Max Kelly's book \Enriched Category Theory") but only concentrate on the abelian group case. Definition 1.1. A category C is called pre-additive (equivalently: enriched in abelian groups) if for each pair of objects A; B of C the homset C(A; B) is an abelian group; moreover, it is required that the composition functions C(B; C) × C(A; B) −! C(A; C) are additive in each variable. Explicitly, the condition on composition means that given a diagram g f / k A / B / C / D h we have (g + h)f = gf + hf; k(g + h) = kg + kh: Note that pre-additivity is really an extra piece of structure on the cat- egory: in particular, a category may be pre-additive in different ways. To see this, take a category with two objects A; B and let Hom(A; A) = 0 = Hom(B; B), and Hom(A; B) a 4-element set. Then since we have ∼ Z=4Z =6 Z=2Z ⊕ Z=2Z there are two different abelian group structures on Hom(A; B). Examples 1.2. (1) As mentioned earlier, any category of R-modules is pre-additive when we define a pointwise group structure on the homsets. In particular the category of abelian groups is pre-additive. 1 2 (2) Recall that a one-object category is the same thing as a monoid. Now a one-object pre-additive category is the same thing as a monoid which also has the structure of an abelian group, in such a way that the multiplication (which is the composition in the one-object category) preserves addition. Such a structure is commonly called a ring. So, if you like, you can regard a pre-additive category as a \multi-object" ring. (3) If C is pre-additive then so is the opposite category Cop. (4) If C is pre-additive then so is any functor category CD. Exercise 1. Work out the last example. Exercise 2. Show that pre-additive structure on a category C corresponds to a lift of the Hom-functor of C, as in AbGrp s9 s s U s s Cop × C / Set C(−;−) where U is the forgetful functor. Definition 1.3 (Additive functor). Let F : C −! D be a functor between pre-additive categories. We say F is additive whenever for each pair of objects A; B of C the induced function C(A; B) −! D(F A; F B) is an additive map. The apparent terminological mismatch (pre-additive vs. additive) will be cleared up in the next section. Examples 1.4. (1) The forgetful functor R − Mod −! AbGrp is additive. (2) When C and D are both pre-additive categories, then we may con- sider the (full) subcategory of CD on those objects which are additive functors C −! D. This is again a pre-additive category. (3) Let R be a ring, considered as one-object pre-additive category. By the previous example the category of additive functors from R to the category of abelian groups is again a pre-additive category. (4) For any pre-additive category with terminal object, the global sec- tions functor factors additively through the category of abelian groups. Exercise 3. Show that global sections is an additive functor. Exercise 4. Prove that the category of additive functors from R to AbGrp is equivalent (via an additive functor!) to the category of R-modules. Exercise 5. Let A be an object of a pre-additive category C. Show that End(A) = C(A; A) is a ring. 3 1.2. Additive categories. We now add some extra structure to our pre- additive categories: Definition 1.5 (Additive category). A pre-additive category C is additive if it has finite products and a zero object. Most examples of pre-additive categories given before are in fact additive: for any ring R, the category of R-modules is additive; if C is additive then so is CD for any D, and so is the subcategory on the additive functors D −! C if D is also (pre-)additive. The correct notion of functor between additive categories is the same as for pre-additive categories. The reason is that, as we shall see below, any additive functor automatically preserves finite products and zero objects. We now show that the notion of additive category is, in spite of the apparent non-duality of the definition, is self-dual. For that, we show that any additive category not only has products but also has finite coproducts. In fact: Lemma 1.6. If C is additive, then C has biproducts. Proof. Given two objects A; B write A ⊕ B for the product of A and B. We shall show that A ⊕ B is (the vertex of) a coproduct of A and B. Define coproduct inclusions by h1A; 0i : A −! A ⊕ B; h0; 1Bi : B −! A ⊕ B: Given morphisms f : A −! C and g : B −! C define [f; g]: A ⊕ B −! C as fπA + gπB. Remaining details are left as exercise. Exercise 6. Complete the proof. We now obtain: Corollary 1.7. If C is additive then so is Cop. Note that we used that additive structure on homsets (and the zero maps) to define the coproduct structure. Thus, the additive structure on C and the finite products are related! To make this clearer, consider two parallel maps f; g : A −! B. On the one hand, since our category is pre-additive, we may use the abelian group structure on C(A; B) to form f + g. But on the other hand we can consider [f; g] A −−!δ A ⊕ A −−−−! B: Here, δ is the diagonal map. From the construction of [f; g] it is clear that f + g = [f; g]δ. Thus, the addition on C(A; B) can be recovered from the biproduct structure. Moreover, the map A −! 0 −! B is indeed the unit element of C(A; B) (check!). This means that in an additive category the pre-additive structure (i.e. the abelian group structures on C(A; B) is uniquely determined. Conse- quently, unlike for pre-additive categories, it is impossible for a category to be additive in more than one way. Exercise 7. Let F : C −! D be an additive functor between additive cate- gories. Show that F preserves finite products and the zero object. 4 1.3. Abelian categories. The final step to make is to add to the notion of additive category some exactness properties: Definition 1.8 (Abelian Category). An additive category C is called abelian if: (i) every map in C has a kernel; (ii) every map in C has a cokernel; (iii) every monomorphism in C is the kernel of its cokernel; (iv) every epimorphism in C is the cokernel of its kernel. Note that this definition is again self-dual so that C is abelian if and only if Cop is. The conditions on monomorphisms and epimorphisms imply that these are regular; in particular, an abelian category is balanced, in the sense that any morphism which is at the same time mono and epi is an isomorphism. Examples 1.9. (1) The leading example of an abelian category is R − Mod for any ring R. (2) If C is abelian then so is CD for any D. (3) If C is abelian and D is (pre-)additive then the category of additive functors D −! C is again abelian. (Thus the fact that R − Mod is abelian is a consequence of the fact that this category is of the form AbGrpR and that AbGrp is abelian.) (4) The category of finite abelian groups is abelian. (5) The category of free abelian groups is additive but not abelian (check!). We begin with a few easy consequences of the definition. Lemma 1.10. In an abelian category, every morphism factors as an epi- morphism followed by a monomorphism.
Load more

Recommended publications
  • BIPRODUCTS WITHOUT POINTEDNESS 1. Introduction
    BIPRODUCTS WITHOUT POINTEDNESS MARTTI KARVONEN Abstract. We show how to define biproducts up to isomorphism in an ar- bitrary category without assuming any enrichment. The resulting notion co- incides with the usual definitions whenever all binary biproducts exist or the category is suitably enriched, resulting in a modest yet strict generalization otherwise. We also characterize when a category has all binary biproducts in terms of an ambidextrous adjunction. Finally, we give some new examples of biproducts that our definition recognizes. 1. Introduction Given two objects A and B living in some category C, their biproduct { according to a standard definition [4] { consists of an object A ⊕ B together with maps p i A A A ⊕ B B B iA pB such that pAiA = idA pBiB = idB pBiA = 0A;B pAiB = 0B;A and idA⊕B = iApA + iBpB. For us to be able to make sense of the equations, we must assume that C is enriched in commutative monoids. One can get a slightly more general definition that only requires zero morphisms but no addition { that is, enrichment in pointed sets { by replacing the last equation with the condition that (A ⊕ B; pA; pB) is a product of A and B and that (A ⊕ B; iA; iB) is their coproduct. We will call biproducts in the first sense additive biproducts and in the second sense pointed biproducts in order to contrast these definitions with our central object of study { a pointless generalization of biproducts that can be applied in any category C, with no assumptions concerning enrichment. This is achieved by replacing the equations referring to zero with the single equation (1.1) iApAiBpB = iBpBiApA, which states that the two canonical idempotents on A ⊕ B commute with one another.
    [Show full text]
  • The Factorization Problem and the Smash Biproduct of Algebras and Coalgebras
    Algebras and Representation Theory 3: 19–42, 2000. 19 © 2000 Kluwer Academic Publishers. Printed in the Netherlands. The Factorization Problem and the Smash Biproduct of Algebras and Coalgebras S. CAENEPEEL1, BOGDAN ION2, G. MILITARU3;? and SHENGLIN ZHU4;?? 1University of Brussels, VUB, Faculty of Applied Sciences, Pleinlaan 2, B-1050 Brussels, Belgium 2Department of Mathematics, Princeton University, Fine Hall, Washington Road, Princeton, NJ 08544-1000, U.S.A. 3University of Bucharest, Faculty of Mathematics, Str. Academiei 14, RO-70109 Bucharest 1, Romania 4Institute of Mathematics, Fudan University, Shanghai 200433, China (Received: July 1998) Presented by A. Verschoren Abstract. We consider the factorization problem for bialgebras. Let L and H be algebras and coalgebras (but not necessarily bialgebras) and consider two maps R: H ⊗ L ! L ⊗ H and W: L ⊗ H ! H ⊗ L. We introduce a product K D L W FG R H and we give necessary and sufficient conditions for K to be a bialgebra. Our construction generalizes products introduced by Majid and Radford. Also, some of the pointed Hopf algebras that were recently constructed by Beattie, Dascˇ alescuˇ and Grünenfelder appear as special cases. Mathematics Subject Classification (2000): 16W30. Key words: Hopf algebra, smash product, factorization structure. Introduction The factorization problem for a ‘structure’ (group, algebra, coalgebra, bialgebra) can be roughly stated as follows: under which conditions can an object X be written as a product of two subobjects A and B which have minimal intersection (for example A \ B Df1Xg in the group case). A related problem is that of the construction of a new object (let us denote it by AB) out of the objects A and B.In the constructions of this type existing in the literature ([13, 20, 27]), the object AB factorizes into A and B.
    [Show full text]
  • Arxiv:2001.09075V1 [Math.AG] 24 Jan 2020
    A topos-theoretic view of difference algebra Ivan Tomašić Ivan Tomašić, School of Mathematical Sciences, Queen Mary Uni- versity of London, London, E1 4NS, United Kingdom E-mail address: [email protected] arXiv:2001.09075v1 [math.AG] 24 Jan 2020 2000 Mathematics Subject Classification. Primary . Secondary . Key words and phrases. difference algebra, topos theory, cohomology, enriched category Contents Introduction iv Part I. E GA 1 1. Category theory essentials 2 2. Topoi 7 3. Enriched category theory 13 4. Internal category theory 25 5. Algebraic structures in enriched categories and topoi 41 6. Topos cohomology 51 7. Enriched homological algebra 56 8. Algebraicgeometryoverabasetopos 64 9. Relative Galois theory 70 10. Cohomologyinrelativealgebraicgeometry 74 11. Group cohomology 76 Part II. σGA 87 12. Difference categories 88 13. The topos of difference sets 96 14. Generalised difference categories 111 15. Enriched difference presheaves 121 16. Difference algebra 126 17. Difference homological algebra 136 18. Difference algebraic geometry 142 19. Difference Galois theory 148 20. Cohomologyofdifferenceschemes 151 21. Cohomologyofdifferencealgebraicgroups 157 22. Comparison to literature 168 Bibliography 171 iii Introduction 0.1. The origins of difference algebra. Difference algebra can be traced back to considerations involving recurrence relations, recursively defined sequences, rudi- mentary dynamical systems, functional equations and the study of associated dif- ference equations. Let k be a commutative ring with identity, and let us write R = kN for the ring (k-algebra) of k-valued sequences, and let σ : R R be the shift endomorphism given by → σ(x0, x1,...) = (x1, x2,...). The first difference operator ∆ : R R is defined as → ∆= σ id, − and, for r N, the r-th difference operator ∆r : R R is the r-th compositional power/iterate∈ of ∆, i.e., → r r ∆r = (σ id)r = ( 1)r−iσi.
    [Show full text]
  • 1.Introduction Given a Class Σ of Morphisms of a Category X, We Can Construct a Cat- Egory of Fractions X[Σ−1] Where All Morphisms of Σ Are Invertible
    Pre-Publicac¸´ oes˜ do Departamento de Matematica´ Universidade de Coimbra Preprint Number 15–41 A CALCULUS OF LAX FRACTIONS LURDES SOUSA Abstract: We present a notion of category of lax fractions, where lax fraction stands for a formal composition s∗f with s∗s = id and ss∗ ≤ id, and a corresponding calculus of lax fractions which generalizes the Gabriel-Zisman calculus of frac- tions. 1.Introduction Given a class Σ of morphisms of a category X, we can construct a cat- egory of fractions X[Σ−1] where all morphisms of Σ are invertible. More X X Σ−1 precisely, we can define a functor PΣ : → [ ] which takes the mor- Σ phisms of to isomorphisms, and, moreover, PΣ is universal with respect to this property. As shown in [13], if Σ admits a calculus of fractions, then the morphisms of X[Σ−1] can be expressed by equivalence classes of cospans (f,g) of morphisms of X with g ∈ Σ, which correspond to the for- mal compositions g−1f . We recall that categories of fractions are closely related to reflective sub- categories and orthogonality. In particular, if A is a full reflective subcat- egory of X, the class Σ of all morphisms inverted by the corresponding reflector functor – equivalently, the class of all morphisms with respect to which A is orthogonal – admits a left calculus of fractions; and A is, up to equivalence of categories, a category of fractions of X for Σ. In [3] we presented a Finitary Orthogonality Deduction System inspired by the left calculus of fractions, which can be looked as a generalization of the Implicational Logic of [20], see [4].
    [Show full text]
  • SHORT INTRODUCTION to ENRICHED CATEGORIES FRANCIS BORCEUX and ISAR STUBBE Département De Mathématique, Université Catholique
    SHORT INTRODUCTION TO ENRICHED CATEGORIES FRANCIS BORCEUX and ISAR STUBBE Departement de Mathematique, Universite Catholique de Louvain, 2 Ch. du Cyclotron, B-1348 Louvain-la-Neuve, Belgium. e-mail: [email protected] — [email protected] This text aims to be a short introduction to some of the basic notions in ordinary and enriched category theory. With reasonable detail but always in a compact fashion, we have brought together in the rst part of this paper the denitions and basic properties of such notions as limit and colimit constructions in a category, adjoint functors between categories, equivalences and monads. In the second part we pass on to enriched category theory: it is explained how one can “replace” the category of sets and mappings, which plays a crucial role in ordinary category theory, by a more general symmetric monoidal closed category, and how most results of ordinary category theory can be translated to this more general setting. For a lack of space we had to omit detailed proofs, but instead we have included lots of examples which we hope will be helpful. In any case, the interested reader will nd his way to the references, given at the end of the paper. 1. Ordinary categories When working with vector spaces over a eld K, one proves such theorems as: for all vector spaces there exists a base; every vector space V is canonically included in its bidual V ; every linear map between nite dimensional based vector spaces can be represented as a matrix; and so on. But where do the universal quantiers take their value? What precisely does “canonical” mean? How can we formally “compare” vector spaces with matrices? What is so special about vector spaces that they can be based? An answer to these questions, and many more, can be formulated in a very precise way using the language of category theory.
    [Show full text]
  • Abelian Categories
    Abelian Categories Lemma. In an Ab-enriched category with zero object every finite product is coproduct and conversely. π1 Proof. Suppose A × B //A; B is a product. Define ι1 : A ! A × B and π2 ι2 : B ! A × B by π1ι1 = id; π2ι1 = 0; π1ι2 = 0; π2ι2 = id: It follows that ι1π1+ι2π2 = id (both sides are equal upon applying π1 and π2). To show that ι1; ι2 are a coproduct suppose given ' : A ! C; : B ! C. It φ : A × B ! C has the properties φι1 = ' and φι2 = then we must have φ = φid = φ(ι1π1 + ι2π2) = ϕπ1 + π2: Conversely, the formula ϕπ1 + π2 yields the desired map on A × B. An additive category is an Ab-enriched category with a zero object and finite products (or coproducts). In such a category, a kernel of a morphism f : A ! B is an equalizer k in the diagram k f ker(f) / A / B: 0 Dually, a cokernel of f is a coequalizer c in the diagram f c A / B / coker(f): 0 An Abelian category is an additive category such that 1. every map has a kernel and a cokernel, 2. every mono is a kernel, and every epi is a cokernel. In fact, it then follows immediatly that a mono is the kernel of its cokernel, while an epi is the cokernel of its kernel. 1 Proof of last statement. Suppose f : B ! C is epi and the cokernel of some g : A ! B. Write k : ker(f) ! B for the kernel of f. Since f ◦ g = 0 the map g¯ indicated in the diagram exists.
    [Show full text]
  • Arxiv:1908.01212V3 [Math.CT] 3 Nov 2020 Step, One Needs to Apply a For-Loop to Divide a Matrix Into Blocks
    Typing Tensor Calculus in 2-Categories Fatimah Ahmadi Department of Computer Science University of Oxford November 4, 2020 Abstract We introduce semiadditive 2-categories, 2-categories with binary 2- biproducts and a zero object, as a suitable framework for typing tensor calculus. Tensors are the generalization of matrices, whose components have more than two indices. 1 Introduction Linear algebra is the primary mathematical toolbox for quantum physicists. Categorical quantum mechanics re-evaluates this toolbox by expressing each tool in the categorical language and leveraging the power of graphical calculus accompanying monoidal categories to gain an alternative insight into quantum features. In the categorical description of quantum mechanics, everything is typed in FHilb; the category whose objects are finite dimensional Hilbert spaces and morphisms are linear maps/matrices. In this category, Hilbert spaces associated with physical systems are typed as objects, and processes between systems as morphisms. MacLane [7] proposed the idea of typing matrices as morphisms while intro- ducing semiadditive categories(categories with well-behaved additions between objects). This line of research, further, has been explored by Macedo and Oliveira in the pursuit of avoiding the cumbersome indexed-based operations of matrices [6]. They address the quest for shifting the traditional perspective of formal methods in software development[10]. To observe why index-level operations do not offer an optimal approach, take the divide-and-conquer algorithm to multiplicate two matrices. In each arXiv:1908.01212v3 [math.CT] 3 Nov 2020 step, one needs to apply a for-loop to divide a matrix into blocks. While the division happens automatically if one takes matrices whose sources or targets are biproducts of objects.
    [Show full text]
  • UNIVERSITY of CALIFORNIA RIVERSIDE the Grothendieck
    UNIVERSITY OF CALIFORNIA RIVERSIDE The Grothendieck Construction in Categorical Network Theory A Dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Mathematics by Joseph Patrick Moeller December 2020 Dissertation Committee: Dr. John C. Baez, Chairperson Dr. Wee Liang Gan Dr. Carl Mautner Copyright by Joseph Patrick Moeller 2020 The Dissertation of Joseph Patrick Moeller is approved: Committee Chairperson University of California, Riverside Acknowledgments First of all, I owe all of my achievements to my wife, Paola. I couldn't have gotten here without my parents: Daniel, Andrea, Tonie, Maria, and Luis, or my siblings: Danielle, Anthony, Samantha, David, and Luis. I would like to thank my advisor, John Baez, for his support, dedication, and his unique and brilliant style of advising. I could not have become the researcher I am under another's instruction. I would also like to thank Christina Vasilakopoulou, whose kindness, energy, and expertise cultivated a deeper appreciation of category theory in me. My expe- rience was also greatly enriched by my academic siblings: Daniel Cicala, Kenny Courser, Brandon Coya, Jason Erbele, Jade Master, Franciscus Rebro, and Christian Williams, and by my cohort: Justin Davis, Ethan Kowalenko, Derek Lowenberg, Michel Manrique, and Michael Pierce. I would like to thank the UCR math department. Professors from whom I learned a ton of algebra, topology, and category theory include Julie Bergner, Vyjayanthi Chari, Wee-Liang Gan, Jos´eGonzalez, Jacob Greenstein, Carl Mautner, Reinhard Schultz, and Steffano Vidussi. Special thanks goes to the department chair Yat-Sun Poon, as well as Margarita Roman, Randy Morgan, and James Marberry, and many others who keep the whole thing together.
    [Show full text]
  • All (,1)-Toposes Have Strict Univalent Universes
    All (1; 1)-toposes have strict univalent universes Mike Shulman University of San Diego HoTT 2019 Carnegie Mellon University August 13, 2019 One model is not enough A (Grothendieck{Rezk{Lurie)( 1; 1)-topos is: • The category of objects obtained by \homotopically gluing together" copies of some collection of \model objects" in specified ways. • The free cocompletion of a small (1; 1)-category preserving certain well-behaved colimits. • An accessible left exact localization of an (1; 1)-category of presheaves. They are a powerful tool for studying all kinds of \geometry" (topological, algebraic, differential, cohesive, etc.). It has long been expected that (1; 1)-toposes are models of HoTT, but coherence problems have proven difficult to overcome. Main Theorem Theorem (S.) Every (1; 1)-topos can be given the structure of a model of \Book" HoTT with strict univalent universes, closed under Σs, Πs, coproducts, and identity types. Caveats for experts: 1 Classical metatheory: ZFC with inaccessible cardinals. 2 We assume the initiality principle. 3 Only an interpretation, not an equivalence. 4 HITs also exist, but remains to show universes are closed under them. Towards killer apps Example 1 Hou{Finster{Licata{Lumsdaine formalized a proof of the Blakers{Massey theorem in HoTT. 2 Later, Rezk and Anel{Biedermann{Finster{Joyal unwound this manually into a new( 1; 1)-topos-theoretic proof, with a generalization applicable to Goodwillie calculus. 3 We can now say that the HFLL proof already implies the (1; 1)-topos-theoretic result, without manual translation. (Modulo closure under HITs.) Outline 1 Type-theoretic model toposes 2 Left exact localizations 3 Injective model structures 4 Remarks Review of model-categorical semantics We can interpret type theory in a well-behaved model category E : Type theory Model category Type Γ ` A Fibration ΓA Γ Term Γ ` a : A Section Γ ! ΓA over Γ Id-type Path object .
    [Show full text]
  • Enriched Categories
    Enriched Categories Ross Tate May 16, 2018 Definition (Enriched Category over a Multicategory). For a multicategory M, an M-enriched category is made of: Objects A set Ob of objects Morphisms For each A and B in Ob, an object Hom(A; B) of M Identities For each A in Ob, a nullary multimorphism idA :[] !M Hom(A; A) Compositions For each A, B, and C in Ob, a multimorphism compA;B;C : [Hom(A; B); Hom(B; C)] !M Hom(A; C) Identity With the property that comp and comp equal id aidA;idHom(A;B) A;B;C aid Hom(A;B);idB A;B;C Hom(A;B) Associativity And the property that comp equals comp acompA;B;C ;idHom(C;D) A;C;D aidHom(A;B);compB;C;D A;B;D Example. A Set-enriched category is simply a category. Ob is the set of objects of the category, and Hom(A; B) is the set of morphisms from A to B. idA is the identity morphism id A, and compA;B;C is the composition function mapping f : A ! B and g : B ! C to f ; g : A ! C. The identity and associativity requirements of Set-enriched categories are exactly the identity and associativity requirements of categories. Example. Many examples of enriched categories are enriched over a multicategory M with an obvious faithful functor to Set. In such cases, M-enriched categories can be viewed as categories whose sets of morphisms are enriched with additional M-structure, and whose identities and compositions respect that structure.
    [Show full text]
  • An Aop Approach to Typed Linear Algebra
    An AoP approach to typed linear algebra J.N. Oliveira (joint work with Hugo Macedo) Dept. Inform´atica, Universidade do Minho Braga, Portugal IFIP WG2.1 meeting #65 27th January 2010 Braga, Portugal Motivation Matrices = arrows Abelian category Abide laws Divide & conquer Vectorization References Context and Motivation • The advent of on-chip parallelism poses many challenges to current programming languages. • Traditional approaches (compiler + hand-coded optimization are giving place to trendy DSL-based generative techniques. • In areas such as scientific computing, image/video processing, the bulk of the work performed by so-called kernel functions. • Examples of kernels are matrix-matrix multiplication (MMM), the discrete Fourier transform (DFT), etc. • Kernel optimization is becoming very difficult due to the complexity of current computing platforms. Motivation Matrices = arrows Abelian category Abide laws Divide & conquer Vectorization References Teaching computers to write fast numerical code In the Spiral Group (CMU), a DSL has been defined (OL) (Franchetti et al., 2009) to specify kernels in a data-independent way. • Divide-and-conquer algorithms are described as OL breakdown rules. • By recursively applying these rules a space of algorithms for a desired kernel can be generated. Rationale behind Spiral: • Target imperative code is too late for numeric processing kernel optimization. • Such optimization can be elegantly and efficiently performed well above in the design chain once the maths themselves are expressed in an index-free style. Motivation Matrices = arrows Abelian category Abide laws Divide & conquer Vectorization References Starting point Synergy: • Parallel between the pointfree notation of OL and relational algebra (relations are Boolean matrices) • Rich calculus of algebraic rules.
    [Show full text]
  • Introduction to Categories
    6 Introduction to categories 6.1 The definition of a category We have now seen many examples of representation theories and of operations with representations (direct sum, tensor product, induction, restriction, reflection functors, etc.) A context in which one can systematically talk about this is provided by Category Theory. Category theory was founded by Saunders MacLane and Samuel Eilenberg around 1940. It is a fairly abstract theory which seemingly has no content, for which reason it was christened “abstract nonsense”. Nevertheless, it is a very flexible and powerful language, which has become totally indispensable in many areas of mathematics, such as algebraic geometry, topology, representation theory, and many others. We will now give a very short introduction to Category theory, highlighting its relevance to the topics in representation theory we have discussed. For a serious acquaintance with category theory, the reader should use the classical book [McL]. Definition 6.1. A category is the following data: C (i) a class of objects Ob( ); C (ii) for every objects X; Y Ob( ), the class Hom (X; Y ) = Hom(X; Y ) of morphisms (or 2 C C arrows) from X; Y (for f Hom(X; Y ), one may write f : X Y ); 2 ! (iii) For any objects X; Y; Z Ob( ), a composition map Hom(Y; Z) Hom(X; Y ) Hom(X; Z), 2 C × ! (f; g) f g, 7! ∞ which satisfy the following axioms: 1. The composition is associative, i.e., (f g) h = f (g h); ∞ ∞ ∞ ∞ 2. For each X Ob( ), there is a morphism 1 Hom(X; X), called the unit morphism, such 2 C X 2 that 1 f = f and g 1 = g for any f; g for which compositions make sense.
    [Show full text]