A quiz on the machine learning experiment

Adrian F. Clark

1. What is the accuracy of MLP on digit '4' of MNIST?

   • 0.97
   • 0.99
   • 0

   • 1

   MLP (and some other techniques) report a perfect score on MNIST, so the accuracy is 1.

2. How do you tell ml.py where to find any images it needs?

   • using the -data command-line qualifier

   • by editing the task file
   • using the -dataset command-line qualifier
   • by editing the program

   ml.py supports the '-data' qualifier for you to tell it where the images are to be found.

3. What is the main purpose of this experiment?

   • to try out deep learning
   • to learn about machine learning

   • to give experience in comparing the performances of algorithms

   • to see whether machine learning is better than hand-crafting solutions

   This experiment uses a variety of machine learning approaches, spanning conventional and modern, and its real purpose is for you to see which is better and how the training database affects performance.

4. Which of the supported learners is the slowest to run?

   • eigen

   • WISARD
   • MLP
   • CNN

   At least with the sketchy implementation in ml.py, eigen is by far the slowest of the learners to run...though it is probably the fastest to train.

5. Which of the following is NOT supported by ml.py?

   • support vector machine
   • multi-layer perceptron
   • convolutional neural network

   • genetic programming

   Genetic programming (GP) is not yet supported by ml.py. I use GP a lot in my own research and so one day will get round to interfacing it to the program.

6. When comparing the performances of RF and SVM on MNIST, for how many digits is the performance difference statistically significant?

   • 0
   • 9

   • 7

   • 10

   RF out-performs SVM on all digits of MNIST but those for 0, 1 and 6 are below the significance threshold of 1.96.

7. How many different classes are then in the ORL5 task?

   • 5
   • 10
   • 6

   • 40

   There are 40 subjects in the ORL database so this is a 40-class problem.

8. The eigen learner can use 'manhatten' to match feature vectors. What does this compute?

   • the correlation coefficient of the feature vectors

   • the sum of absolute differences between features

   • the angle between the feature vectors
   • the sum-squared differences between features (aka Pythagoras's theorem)

   The Manhatten or 'city block' distance is the sum of the absolute differences. It crops up quite often when assessing the similarity of things.

9. On which digit or digits of MNIST does eigen out-perform SVM?

   • 0

   • 0 and 8
   • 1 and 7
   • all of them

   eigen out-performs SVM on digit zero, though the performance difference is not significant.

10. What is a 'knowledge base' in ml.py?

    • a compiled program
    • a type of data structure
    • a set of rules

    • a trained machine learning system

    ml.py refers to a trained machine learning system as a 'knowledge base'.