China
Africa
Explore chapters and articles related to this topic
It's Not Just the Algorithms, Really!
Published in James Luke, David Porter, Padmanabhan Santhanam, Beyond Algorithms, 2022
James Luke, David Porter, Padmanabhan Santhanam
Algorithms are generally implemented using a chosen programming language (e.g. java, python, etc.) as generic tools. In the brave new world of Cloud-based services, most of the Cloud providers provide implementations of the most popular AI algorithms that you can provision as services. So, how do you make a generic algorithm work for your particular application? The answer is that you give it an AI Model. An AI Model is a configuration of an AI Algorithm to perform a specific task. The form of the AI Model depends on the AI Algorithm. In a rules-based system, the Model will be a list of rules. In a neural network, the Model will describe the configuration of the network in terms of the number of neurons and the connectivity between them. For example, if we use a Neural Network to perform the X-Ray classification, the AI Model will comprise the weights and configuration of the neurons required to perform that task.
Integer Programming
Published in Michael W. Carter, Camille C. Price, Ghaith Rabadi, Operations Research, 2018
Michael W. Carter, Camille C. Price, Ghaith Rabadi
Google OR Tools offers an interface to several MIP solvers. By default, it uses COIN-OR branch and cut implementation, an open source solver from the Computational Infrastructure for Operations Research project (COIN-OR). However, one can also use other MIP solvers (such as Gurobi) with Google OR Tools wrapper. Google’s OR Tools are offered for various platforms (Windows, Mac OS and Linux) and languages (C++, Java, and Python).
Software Technology for A/V Systems
Published in Al Kovalick, Video Systems in an IT Environment, 2013
.NET has found a home with standalone as well as distributed systems. In fact, all the architectural classes in Figure 4.3 can be implemented using .NET’s tools, components, and connectivity. Many standalone A/V software applications use the .NET framework and development tools for program creation. Its tight integration with the Windows OS makes it the preferred framework for many software projects.
A multi-agent deep reinforcement learning approach for solving the multi-depot vehicle routing problem
Published in Journal of Management Analytics, 2023
We terminate the search when there is no improvement in the fitness value for ten generations or when the maximum number of generations is reached. (iii) OR-Tools: Google's operations research tools for combinatorial optimization. OR-tools implements different construction heuristics such as path cheapest arc, savings, and best insertion, each of which can be incorporated into metaheuristics, such as TS, SA, Guided local search, and Gradient descent (Perron & Furnon, 2019; Surana, 2019). The OR-tools can be adjusted to work as heuristics, metaheuristics, or a hybrid heuristics solver. In our experiments, we use the hybrid setting recommended by the developers and allow the solver to choose the best heuristics and metaheuristics automatically in each run. The search procedures will stop when the number of solutions generated during the search exceeds 1000.
Dynamic optimization for the enzymatic production of acylglycerols
Published in Chemical Engineering Communications, 2020
Laís Koop, Lorena I. Soares, Fernando Augusto Pedersen Voll, Adrian Bonilla-Petriciolet, Marcos Lúcio Corazza
For the resolution of complex dynamic optimization problems as those involved in acylglycerols production, several authors have reported the application of stochastic global optimization methods (Kapadi and Gudi, 2004; Lee et al., 2007; Liu et al., 2013; Rocha et al., 2014; Datta and Kumar, 2015; Mirvakili et al., 2015; Eswari and Venkateswarlu, 2016). These tools can be used for any optimization problem, since they do not require any assumption of its mathematical characteristics such as the continuity of the objective function and constrains. In addition, they are robust for high non-linear problems even with high number of decision variables (Rangaiah, 2010).