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
Confusion Matrix and Weighted Confusion Matrix 6.3.2 Confusion Matrix with Output Space Transformation (Probability Space Transformation) 6.3.3 Tessellation and Recombination with Expert Decisioner 6.3.4 Predictive Selection with Secondary Engines 6.3.5 Single Engine with Required Precision 6.3.6 Majority Voting or Weighted Confusion Matrix 6.3.7 Majority Voting or Best Engine 6.3.8 Best Engine with Differential Confidence or Second Best Engine 6.3.9 Best Engine with Absolute Confidence or
domain space that contain no data are much more likely to be actual nonrelevant domain sections. Such gaps in the domain of one or more intelligence-generating systems significantly aid in the selection of individual systems for the application of a meta-algorithmic pattern. 3. Improved data layering: With more data comes better association among the data. That is, more nuanced data sets afford greater possibilities for multiple meta-algorithmic patterns to be evaluated on the same data set in
considered in the context of document understanding systems. These operations cannot be defined by a single pattern per se. Instead, they depend on domain knowledge to prepare for parallelism. A LI-PSO is parallelism by task wherein the original operation is deconstructed for the purposes of preparing it for subsequent parallel processing. Each of the four large systems described in Table 4.2 are likely highly amenable to LI-PSO if, in fact, they are not already explicitly designed internally for
from the original image. In (c), the sharpened image of Figure 4.4d is down-sampled using the D kernel. Although convolution with D should introduce some blur to the image, the effect of the dramatic sharpening is only exacerbated. The relatively poor results indicate that sharpening should not occur before down-sampling for the kernels defined and the particular image. In (d), the P kernel, which combines the down-sampling and sharpening into a single operation, is shown. This high-quality image
a different halftone implementation of, the following: (1) differential down-sampling of image regions based on their frequency, entropy, and so on, measurements; (2) palettization of different image regions based on their histograms, chroma variance, entropy, and so on; and (3) filtering (e.g., sharpening, blurring, unsharp masking, etc.) of different image regions based on their regional characteristics. It is clear that such approaches naturally lead to meta-algorithmic patterns such as