Fast AI

Question 1

What did Frank Rosenblatt build?

Answer 1

perceptron

Question 2

What was Marvin Minsky’s innovation?

Answer 2

Using multiple layers could solve for limitations

Question 3

What is the general method for updating weights in a neural net?

Answer 3

Stochastic gradient descent

Question 4

What is ImageNet?

Answer 4

The ImageNet project is a large visual database designed for use in visual object recognition software research. More than 14 million images have been hand-annotated by the project to indicate what objects are pictured and in at least one million of the images, bounding boxes are also provided

Question 5

What’s a tensor?

Answer 5

collections of numbers (vector, matrix, and higher dimensions)

Question 6

Explain gradient descent

Answer 6

At each step we try and work out the gradient and based on updating the parameters to minimise the loss

Question 7

What function can be used to describe any input-output? In neural nets

Answer 7

Combinations of Rectified Linear functions

Question 8

Describe deep learning mathematically…

Answer 8

Combinations of ReLu
Gradient descent
Inputs and outputs

Question 9

When i adjust parameters (in line with gradient descent), the amount i adjust by is called the what?

Answer 9

The learning rate

Question 10

the foundational mathematical operation in deep learning is …

Answer 10

Matrix multiplication

Question 11

Describe solving the Titanic (kaggle) problem using regression in Excel

Answer 11

• Normalize all the variables
• Random coefficients
• Loss for each row, and then the mean of the sum
• “solve” function to minimimise the Total Loss by changing the coefficients

Question 12

What is the method known where you split data by one variable to determine the outcome?

Answer 12

1R

Question 13

How is the 1R made more sophisticated?

Answer 13

Turning it into a Decision Tree

Question 14

What is the best package to use in Python for non-deep (aka classical) machine learning?

Answer 14

Scikit Learn

Question 15

What’s the name of the measure for how good a split is?

Answer 15

Gini
(how likely if you pick from a sample and then pick another that you’ll get the same thing)

Question 16

What’s the main way of improving decision trees?

Answer 16

Random Forest

Question 17

What’s the idea behind “bagging” (used in Random Forest)?

Answer 17

Take average predicition of uncorrelated models - that should reduce the error to zero

Question 18

What’s a good first model to use when given a data set?

Answer 18

Random Forest - get to see which columns are the most important

Question 19

What is an Out of Bag Error?

Answer 19

Predictions made on the rows not used in the particular decision tree (subset)

Question 20

What is gradient boosting?

Answer 20

Boosting is one kind of ensemble Learning method which trains the model sequentially and each new model tries to correct the previous model

Question 21

what is TTA (machine learning)?

Answer 21

Test Time Augmentation
(passing along to the model multiple augmented images and getting back the average result)

Question 22

What does Jeremy Howard recommend for GPU?

Answer 22

Nvidia
RTX cards are for consumers (NOT for data centers) e.g. 3080 is about $1,000

Question 23

Recommendation systems are known as

Answer 23

Collaborative Filtering

Question 24

What’s a convolution?

Answer 24

sliding matrix (/kernel) of values from o.g. image dot-product-ed with a filter to pick out features such as vertical edges, horizontal edges etc

Question 25

How is overfitting avoided in CNNs?

Answer 25

By introducing a drop out layer - some of the pixels are zeroed (randomly)