Meta-Algorithmics: Patterns for Robust, Low Cost, High Quality Systems
Steven J. Simske
Format: PDF / Kindle (mobi) / ePub
The confluence of cloud computing, parallelism and advanced machine intelligence approaches has created a world in which the optimum knowledge system will usually be architected from the combination of two or more knowledge-generating systems. There is a need, then, to provide a reusable, broadly-applicable set of design patterns to empower the intelligent system architect to take advantage of this opportunity.
This book explains how to design and build intelligent systems that are optimized for changing system requirements (adaptability), optimized for changing system input (robustness), and optimized for one or more other important system parameters (e.g., accuracy, efficiency, cost). It provides an overview of traditional parallel processing which is shown to consist primarily of task and component parallelism; before introducing meta-algorithmic parallelism which is based on combining two or more algorithms, classification engines or other systems.
- Explains the entire roadmap for the design, testing, development, refinement, deployment and statistics-driven optimization of building systems for intelligence
- Offers an accessible yet thorough overview of machine intelligence, in addition to having a strong image processing focus
- Contains design patterns for parallelism, especially meta-algorithmic parallelism – simply conveyed, reusable and proven effective that can be readily included in the toolbox of experts in analytics, system architecture, big data, security and many other science and engineering disciplines
- Connects algorithms and analytics to parallelism, thereby illustrating a new way of designing intelligent systems compatible with the tremendous changes in the computing world over the past decade
- Discusses application of the approaches to a wide number of fields; primarily, document understanding, image understanding, biometrics and security printing
- Companion website contains sample code and data sets
fixes), rolling out new features, and supporting new services, plug-ins, and add-ons. The costs of (1) and (2) are therefore relatively fixed, as opposed to the costs of bandwidth—which, if reasonable, are usually passed on to the mobile device owner. For the software producer, then, if all other factors are equal, the code development preference is to have a single codebase. This means either a single (or multiple, compatible, e.g., all Linux-based) mobile platform is supported (e.g., the most
partitioning of data to closely map to the processing capabilities throughout the distributed system. Finally, parallelism by metaalgorithmics was introduced. Meta-algorithmics are the means by which parallel processing strategies are brought to multiple intelligence generators, each of which acts on the same data and provides output of the same type. First-, second-, and third-order meta-algorithmic patterns are introduced, corresponding to increasing complexity (more variable factors in the
Parallelism by Task 119 reducing a 600 × 600 pixels/in.—or ppi—image to 300 × 300 ppi), is given here: ⎡ ⎤ 1 3 3 1 ⎢3 9 9 3⎥ ⎢ ⎥ D=⎢ ⎥. ⎣3 9 9 3⎦ 1 3 3 1 To ensure that uniform areas do not have a change in image intensity or saturation, the D kernel is normalized so that the coefficients sum to 1.0, as shown here by D : ⎡ 1/64 ⎢ 3/64 ⎢ D =⎢ ⎣ 3/64 3/64 9/64 9/64 3/64 9/64 9/64 ⎤ 1/64 3/64 ⎥ ⎥ ⎥. 3/64 ⎦ 1/64 3/64 3/64 1/64 The down-sampling kernel is centered over pixel P(x,y) and
across the deterrent (see below). The former is aided by using predetermined tile dimensions, while the latter is aided by uniform lighting and relatively compact deterrent size (or high-quality capture, such as with a scanner or vision system). However, these assumptions often fail in the mobile world. For the purposes of describing the use of parallelism by task, we need the following definitions: 1. A tile is a uniformly colored glyph, nominally a square, from which the overall deterrent is
Donoho (2005), and elsewhere. As noted in Bishop (2006), direct solution of the optimization problem created by the search for an optimum margin is highly complex, and some classification engineers may prefer genetic, near-exhaustive, and/or artificial neural network (ANN) approaches to the 10 Meta-algorithmics: Patterns for Robust, Low-Cost, High-Quality Systems Dimension 1 Dimension 2 Figure 1.3 Example decision boundary (solid line) with margin to either side (dotted lines) as defined by