Mathematical Strategies P.S.Subramanian CSRD group 21 Jan 2001, IIT/ Mumbai 2000 SASKEN All Rights Reserved Mathematical StrategiesStrategy vs Tactics - in Chess Tactics is situation specific and concrete Strategy is generic and abstract Pros and Cons of Strategy and Tactics 2000 SASKEN All Rights Reserved Mathematical Strategies Why study the Strategies of Mathematics? Helps us to `see the forest for the trees’. Makes the learning of `new’ topics easier. Makes the study of `History of Mathematics’ more meaningful. 2000 SASKEN All Rights Reserved Some Common Strategies Encapsulation for representation independence Step-wise refinement Coordinatisation (Cartesian, Positional and Mixed) Reuse Linearisation Localisation Crowding Dualisation 2000 SASKEN All Rights Reserved Encapsulation Need to study properties independent of the `representation’. In Computer Science the essence of OOP Representation = Implementation 2000 SASKEN All Rights Reserved Encapsulation - Example Injectivity of function f : A —› B, where A, B are Sets un-encapsulated definition is a, b in A, f(a) = f(b) => a = b Can we give a definition without in? 2000 SASKEN All Rights Reserved Encapsulation - example Encapsulated Definition let C be another set and g , h : C —› A, be two maps f is injective iff, f ° g= f ° h => g=h Elements have vanished. 2000 SASKEN All Rights Reserved Encapsulation This line of thinking leads to `Category Theory’ For a gentle introduction see `Conceptual Mathematics’ by William Lawvere - Prentice Hall. Strongly Recommended for CS Students 2000 SASKEN All Rights Reserved Step-wise Refinement Given a collection of problems P which we know how to solve, and a new problem Q Find a sequence of subproblems with the property that we have a method of transforming the solution of problems occurring later in the sequence to those of the earlier. 2000 SASKEN All Rights Reserved Stepwise Refinement In particular if the tail of the sequence has problems only from the set P then we can solve Q. 2000 SASKEN All Rights Reserved Stepwise Refinement Gaussian Elimination - What is P and Q? Galois Theory - What is P and Q? Let P be a set of Software specifications for which we have already written programs and Q is new specification for which we want to develop a program. 2000 SASKEN All Rights Reserved Stepwise Refinement Component based Software (and Hardware) Engineering is an important and evolving area. Sample reference see http://www.kestrel.edu 2000 SASKEN All Rights Reserved Co-ordinatisation Cartesian Positional Mixed 2000 SASKEN All Rights Reserved Cartesian Synthetic Projective Geometry Underlying `Mathematics’ is Wedderburn’s Representation Theorem of Semi-simple rings in terms of Matrix rings over division algebras. 2000 SASKEN All Rights Reserved Cartesian The idea of coordinatising the Space of Functions enables us to transport many ideas from the usual coordinate geometry to these spaces. 2000 SASKEN All Rights Reserved Positional Decimal Number System Wavelets Underlying Mathematics is that of Wreath Products Krasner-Kaloujnine Theorem of Embedding a group in the wreath product of the factors of it’s composition series. 2000 SASKEN All Rights Reserved Mixed Krohn- Rhodes Theorem in Automata Theory and it’s generalisations Underlying Mathematics is the theory of Semigroup Decompositions 2000 SASKEN All Rights Reserved Reuse If we have already solved a problem in some domain and if can establish a suitable connection between domains then we can `reuse’ the solutions of problems of the former domain. 2000 SASKEN All Rights Reserved Reuse Example (NOT historically accurate!) Galois Theory (again) Original Domain - Groups Problem- Stepwise Refinement New Domain - Fields Suitable Connection - Galois Connection 2000 SASKEN All Rights Reserved Reuse The Specware software from the Kestrel Institute provides mechanisms for reuse of ideas in the domain of Algorithm Design. But, contrary to Galois theory which is fully automatic one has to provide the connection manually. 2000 SASKEN All Rights Reserved Linearisation Newton-Raphson Temporarily pretend that the situation is linear Generalisation - Kantorovich to Fn Spaces Structural Linearisation - Algebraic Topology Linear to Module to Abelian Categories 2000 SASKEN All Rights Reserved Mathematical Strategies Localisation - Sheaf Theory Representation Theorem of Rings Minkowski-Hasse on Quadratic Forms Many Computer Science uses of Sheaf Theory 2000 SASKEN All Rights Reserved Mathematical Strategies Crowding - Contraction Maps, Ramsey Theory Fixed point Theorems and their uses. Duality- Fourier Transforms, Spectral Methods, Chu Spaces, Ramsey = Discontinuous Duality, 2000 SASKEN All Rights Reserved Mathematical Strategies Conclusion One gets more insight into Mathematics and it’s applications by reflecting on the strategies. 2000 SASKEN All Rights Reserved Some Mathematical Topics relevant to Sasken Separating the strands in Signal Processing. Generalising Shannon’s Information Theory New Coding Techniques Mathematics of Image processing Mathematical aspects of Componentisation 2000 SASKEN All Rights Reserved