- Question 1 Data Model And Evaluation 1 (71.02 KiB) Viewed 17 times
Question 1: Data, Model, and Evaluation
-
answerhappygod
- Site Admin
- Posts: 899603
- Joined: Mon Aug 02, 2021 8:13 am
Question 1: Data, Model, and Evaluation
Question 1: Data, Model, and Evaluation
Section A: Short Questions Question 1: Data, Model, and Evaluation [10 marks] a) Scientists have found 11 different coronavirus strains to date. Suppose that they have a set of data presenting the characteristics of coronavirus, and they will use a neural network to determine which variant a given sample belongs to. They have two options: train a separate neural network for each strain or train a single neural network with one output neuron for each strain but a common hidden layer. Which method is better? Justify your answer. [3 marks] b) We would like to apply a classification model to check if a patient is infected with the coronavirus. If the model is 85% confident that the patient is infected, then we will do extra tests to collect additional clinical features. Which classification methods do you recommend in this case: neural networks, decision tree, or naive Bayes? Justify your answer. [3 marks] c) The following figure shows testing errors (blue line) and training errors (red line) through the iterations. There is a large gap between the two lines. What is the cause of this phenomenon? How do you address the problem? Could you please draw an ideal learning curve? [4 marks] testing error error training error iteration
Section A: Short Questions Question 1: Data, Model, and Evaluation [10 marks] a) Scientists have found 11 different coronavirus strains to date. Suppose that they have a set of data presenting the characteristics of coronavirus, and they will use a neural network to determine which variant a given sample belongs to. They have two options: train a separate neural network for each strain or train a single neural network with one output neuron for each strain but a common hidden layer. Which method is better? Justify your answer. [3 marks] b) We would like to apply a classification model to check if a patient is infected with the coronavirus. If the model is 85% confident that the patient is infected, then we will do extra tests to collect additional clinical features. Which classification methods do you recommend in this case: neural networks, decision tree, or naive Bayes? Justify your answer. [3 marks] c) The following figure shows testing errors (blue line) and training errors (red line) through the iterations. There is a large gap between the two lines. What is the cause of this phenomenon? How do you address the problem? Could you please draw an ideal learning curve? [4 marks] testing error error training error iteration