# Machine learning(1): Quiz1

1. For which one of the following tasks machine learning is NOT suited?
a. Identifying bird species from audio recordings of birds’ songs.
b. Optical character recognition (OCR).
c. Determining whether a student passes or fails the year given the individual modules’ marks.
d. Detection of pedestrians and cyclists in images and videos.
Analysis:
a. For this answer, the songs of birds are signs of each bird. And, different bird species have different voices and songs. So, we could use the technology of machine learning to solve it.
b. As we all know, the technology of machine learning is based on OCR.
c. For this answer, we could also use a simple algorithm to solve it. There is no necessary to use machine learning to work out it.
d. For this answer, the image of pedestrians and cyclists has a lot of specials points, such as clothing, speed of walking and riding, even the equipment(bikes). These data are like a label on pedestrians and cyclists, which is a necessary conditioner of supervised learning.

2. Consider the (1-based) vector y=[0.7,2.2,1.4,1.1,1.5], compute iˆ=arg(mini)y(i).
Analysis:
The task’s purpose is to calculate the minimum index of y(i). We could see that the value of y(1) is 0.7 that is the smallest value. So, the value of i is 1.

3. A company is interested in customer segmentation, i.e. in finding natural groupings of its customers. In other words, they want a model to allocate customers with similar attributes to the same group. They can later use this information to decide group-specific strategies, services or products. This is an example of what kind of learning problem?
a. Supervised learning.
b. It’s not a learning problem.
c. Reinforcement learning.
d. Unsupervised learning.
Analysis:
For this situation, we could see"they want a model to allocate customers with similar attributes to the same group", which means that these customers have no labelled on themself i.e. no labelled data. No labelled data is a sign of Unsupervised learning. Then, let’s see the next sentence, “They can later use this information to decide group-specific strategies, services or products”, which means that they could not use a simple algorithm to solve it. So, that is a machine learning problem.

4. Analysis:
We could see that θ is a value of side in this set. Looking at this table, the maximum value of θ in F set is 2.1 and the minimum value of T set is 3.1, especially, if x≥θ. So, the value of θ is between 2.1 and 3.1(θ could get the 3.1).

5. A biomedical research group is working on methods to automatically and reliably diagnose a large number of images of skin moles. The system uses image processing techniques to identify major features (size, asymmetry, border irregularity, colour and others) from digital images of the skin. These features are then used by a machine-learning algorithm to make predictions. Initially, the algorithm is trained using a set of training examples constituted by a vector of features and by the diagnosis performed by a dermatologist. This is an example of what kind of learning problem?
a. It’s not a learning problem.
b. Unsupervised learning.
c. Supervised learning.
d. Reinforcement learning.
Analysis:
We could see the sentence, “These features are then used by a machine-learning algorithm to make predictions.”, which means that it uses some machine learning algorithm, so it is a machine learning problem.
Then, we could also see that “The system uses image processing techniques to identify major features (size, asymmetry, border irregularity, colour and others) from digital images of the skin”, which means that the data is labelled and has already classified. That is a necessary condition of supervised learning. Finally, the sentence"Initially, the algorithm is trained using a set of training examples constituted by a vector of features and by the diagnosis performed by a dermatologist.", means that it has an initial training set whose element is labelled. That is also a necessary condition of supervised learning. All in all, this problem is a supervised learning problem.

6. A machine learning algorithm is used to simulate human-like intelligence in non-player characters (NPCs) of a combat video game. The goodness of a policy (i.e. the rule deciding the appropriate move at each stage of the game) is determined solely by the number of wins in a set of game plays against a human player. This is an example of what kind of learning problem?
a. Supervised learning.
b. Unsupervised learning.
c. Reinforcement learning.
d. It’s not a learning problem.
7. 