-
Book Overview & Buying
-
Table Of Contents
How to Prove Anything
By :
Mixing and matching positive and negative behaviors creates a particularly large search space in our ongoing wet food optimization problem. Because of the large gaps in rewards between positive or negative behavior and results, two methods were attempted: a Genetic Algorithm (GA) approach and a Reinforcement Learning (RL) approach. Unfortunately for both approaches, the methods needed to be implemented live with real cats and real humans. As shown in [18], humans can’t be simulated, and even the best models can only detect when they come home from work [19].
In order to translate data into a form that can be easily understood by either of the experimental algorithms, each cat behavior will be denoted by a well-defined function with one or more parameters, resulting in a positive or negative value to denote our communicated happiness. For example,
represents a meow of V volume in decibels, D duration seconds, and F centroid...
Change the font size
Change margin width
Change background colour