Content-Driven Neural Network Design of a PSP
Abstract:The PSP design technique that we present is, at its core, the generalization of structured job data via neural network learning techniques. While current PSP design focuses on the optimal use of capital equipment as its primary motivation, the essential competitive advantage of digital presses and workflow is its ability to adapt to different types of content with highest robustness to failure and minimal component-level change; these characteristics are also the same for neural networks. By generalizing the fulfillment order with a representative neural network, we can automatically identify redundancy between jobs and optimize the infrastructure for a particular content mix. By adaptively changing the neural network in the face of different job fulfillment demands, the neural network can also indicate how to transform the current PSP infrastructure to handle a new mix of jobs requests. We apply a structural learning technique based on a subset of Hidden Markov Models, Directed Acyclic Graphics, and then map these neural structures into print shop infrastructure. We will demonstrate our results with real world PSP data, and compare and contrast the current real world PSP design with its neurally designed counterpart.
Document Type: Research Article
Publication date: January 1, 2010
For more than 25 years, NIP has been the leading forum for discussion of advances and new directions in non-impact and digital printing technologies. A comprehensive, industry-wide conference, this meeting includes all aspects of the hardware, materials, software, images, and applications associated with digital printing systems, including drop-on-demand ink jet, wide format ink jet, desktop and continuous ink jet, toner-based electrophotographic printers, production digital printing systems, and thermal printing systems, as well as the engineering capability, optimization, and science involved in these fields.
Since 2005, NIP has been held in conjunction with the Digital Fabrication Conference.
- Information for Authors
- Submit a Paper
- Subscribe to this Title
- Membership Information
- Terms & Conditions
- ingentaconnect is not responsible for the content or availability of external websites