Exploring different approaches to Machine Learning and testing them in architectural contexts.
A genetic algorithm is a search heuristic that is inspired by Charles Darwin’s theory of natural evolution. This algorithm reflects the process of natural selection where the fittest individuals are selected for reproduction in order to produce offspring of the next generation.
- Takes a large number of attempts from multiple agents to get optimised.
- Does not track what the fittest agent did right.
- Unable to adapt well to changing conditions.
Cartpole problem is one of the most common testbeds for machine reinforcement learning algorithms. In this particular demonstration, the agent uses random and noise methods to optimise its behaviour.
- Even in its most trivial form, reinforcement learning has potential.
Maze Seeking Agent
The agent’s purpose here is to reach the target located in the center of the gridworld. It uses a Q-learning algorithm to do so.
It learns from its own past experiences of navigating the maze, and builds up its memory as it goes. Eventually, it gets close to perfect. But it can also adapt itself to new conditions. This is demonstrated by setting up boulders while the agent is still moving.
This study also demonstrates the dilemma of exploration vs exploitation, something which is still an unsolved problem in the discipline of machine learning.
- Resilient to changing conditions, which is good since architectural environments are constantly changing and imperfect.
- Still lacks an human interface.
This simple gadget demonstrates how a machine can learn from human interactions and passively start augmenting the same. The idea is that eventually the gadget will gain so much experience, that there’ll be no need to manually modify its states anymore.
In theory, once there is a sufficient number of such gadgets talking to each other, their combined intelligence will get much more sophisticated and powerful. For example, you pulling your towel out of the cabinet probably means that bathroom lights and geyser needs to be turned on.
- Makes you question if human patterns can be predicted adequately, and how much sensors would be needed to do so.
- This gets even more complicated when more than one person gets in the equation. eg - a couple or a family.
- It is also important identify what data is relevant for which gadget. Deep Learning methods can be used for that purpose and are worth exploring.