Robotics Design, Repair and Restoration
Gradient Decent
The Practical Way
In this chapter, we will be exploring the concept of potential dividers, and how to calculate and utilize them in everyday circuits.
Overview
1.1 - What is Gradient Decent?
Machine Learning is the autonomous process by which a machine can improve its accuracy when attempting to accomplish a task. The aim is for the device to "learn" for itself rather than a programmer providing a step-by-step solution. This is also useful when solving a problem with no clear answer.
For example, how would you teach a robot to tell the difference between pictures of cats and dogs? Would you say that dogs have long ears and cats have round faces? And how reliable will this solution be when new images are shown?
Unfortunately, clear, simple rules do not always work. Yet, even with the most simplistic depictions shown above, we can use our human ability to generalise solutions to solve the problem with a glance. This leads us to a big problem: What may be easy for us can be exceptionally difficult for a robot to solve and even more impossible to explain.
A potential method would be to analyse thousands of examples from both categories, developing rules and uncovering patterns on a more microscopic level, to formulate a generalised solution to this problem. Then, given a model with a suitable architecture, the more and slightly varying the images we provide the system, the more accurate its prediction should become.
Gradient Descent is one of many machine learning algorithms used to iteratively optimise an AI model. It involves making incremental changes to improve the model's accuracy, calculated based on its error.
1.2 - The Problem - Linear Regression
One of the simplest problems we can apply gradient descent is Linear Regression. The line blue depicts a predicted relationship between two variables, for which the red dots represent single data samples from a given dataset. The goal is to adjust the properties of the line so that it eventually "fits" the trend of the sample data, resulting in a line of best fit.
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
1.3 - The Approach
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
2 - Gradient Test
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
3 - How Do They Work?
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
Potential Divider - Lab
1 - What are potential dividers?
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
Potential Divider - Summary
1 - What are potential dividers?
To initiate the repair, we first need to strip down the robot to expose the error, as the basic checks show the fault lies within the robot. Below is a picture of the top face of the robot in question.
Chapter Quiz
Using your newly gained wisdom of potential dividers, attempt to solve the provided quiz question bellow.