1 of 27

1

How do computers make decisions?

Unit 3, Module 3.1

2 of 27

Kinds of decisions computers can make

  • Classification
  • Prediction
  • Recommendation
  • Planning and scheduling

​

An algorithm that performs one of these decision-making tasks is called a reasoner.

There can be multiple reasoning algorithms for the same task. For example, there are many classification algorithms, i.e., many types of reasoners that do classification.

2

3 of 27

What is a classifier?

  • A classifier is a reasoner that looks at an input and decides what category it fits in.
  • The categories are called classes.

3

4 of 27

Prediction Problems

A predictor is a reasoner that estimates a numerical value.

  • If the predictions are about the present they’re just called estimates.
  • If the predictions are about the future they are called forecasts.

​

4

5 of 27

Recommender Systems

Recommender systems suggest items from a large set based on knowledge about a person’s preferences and the preferences of others who resemble them:

  • Things this person liked in the past
  • People who liked the same things this person liked
  • Things these other people liked that this person hasn’t seen yet
    • RECOMMEND those things to this person

​

​

5

6 of 27

Planning & Scheduling

Planning and scheduling problems find efficient ways to accomplish complex tasks in time sensitive situations.

​

6

7 of 27

Name the Reasoner

7

Guess which kind of reasoner is used for each of 15 images

Student Activity

8 of 27

YouTube

8

9 of 27

Auto Filters in Email

9

10 of 27

Movies (ticket sales)

10

11 of 27

Tesla Range before recharging

11

12 of 27

Handwriting Recognition

12

13 of 27

Package Delivery

13

14 of 27

Advertisement

14

15 of 27

Carmax

15

16 of 27

Waze

16

17 of 27

License Plate Reader

17

18 of 27

Tiktok Discover

18

19 of 27

Face Tagging

19

20 of 27

Amazon Warehouses

20

21 of 27

Hurricane Tracker

21

22 of 27

Sneaker Resale

22

23 of 27

Think about how your life is being affected by the decisions computers are making for you.

​

23

Reflection & Discussion

24 of 27

Make Your Own Image Classifier

24

Student Activity

25 of 27

25

26 of 27

Try It: Thumbs up vs Peace sign Classifier

Goal: Train a classifier to identify peace sign, thumbs up, or neither of these

Step 1: Open teachable machine - https://teachablemachine.withgoogle.com/

Step 2: Watch the Getting Started video (scroll down on the page for video).

Step 3: Create a peace sign class, and train it with peace sign gestures

Step 4: Create a thumbs up class, and train it with thumbs up gestures

Step 5: Create a none of the above class, and train it with no hand gestures or no peace or thumbs up sign

Step 6: Build model,

Step 7: Test out model yourself and improve it

Step 8: Then have a friend test your model and improve it

    • ​

26

27 of 27

How to Improve your Classifier

27

Reflection

  • How well does your classifier do? Does it classify the input correctly?

​

  • If your model doesn’t classify your test input correctly, check to see if you have any of the following issues
  • All your training images look the same: not enough variation
  • Not enough images (less than 30)
  • The test images don’t look like the training images: training set is not representative of the test set
  • You don’t have a null (“none of the above”) class where no gesture is being made
  • The differences between the two classes are very subtle