Hands-On Artificial Intelligence for IoT by Amita Kapoor

Author:Amita Kapoor
Language: eng
Format: epub
Publisher: Packt
Published: 2019-01-31T19:41:31+00:00

Yay! Now, you have the best LSTM network for predicting wind power.

Summary

This chapter introduced an interesting nature-inspired algorithm family: genetic algorithms. We covered various standard optimization algorithms, varying from deterministic models, to gradient-based algorithms, to evolutionary algorithms. The biological process of evolution through natural selection was covered. We then learned how to convert our optimization problems into a form suitable for genetic algorithms. Crossover and mutation, two very crucial operations in genetic algorithms, were explained. While it is not possible to extensively cover all the crossover and mutation methods, we did learn about the popular ones.

We applied what we learned on three very different optimization problems. We used it to guess a word. The example was of a five-letter word; had we used simple brute force, it would take a search of a 615 search space. We used genetic algorithms to optimize the CNN architecture; again note that, with 19 possible bits, the search space is 219. Then, we used it to find the optimum hyperparameters for an LSTM network.

In the next chapter, we will talk about another intriguing learning paradigm: reinforcement learning. This is another natural learning paradigm, in the sense that in nature we normally do not have supervised learning; rather, we learn through our interactions with the environment. In the same manner, here the agent is not told anything except the rewards and punishments it receives from the environment after its action.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.