Description

1. When do we use semi-supervised learning? What is self-training?
2. What is the logic behind active learning, and what are some methods to choose instances for the oracle?
3. One of the strategies for Query sampling was query-by-committee (QBC). Using the equation below, which captures vote entropy, determine the instance that our active learner would select first.
𝑥!∗” = argm𝑎𝑥(−*𝑉(𝐶𝑦%)𝑙𝑜𝑔’ 𝑉(𝐶𝑦%))
$
&!
Respectively 𝑦!, 𝑉(𝑦!), 𝑎𝑛𝑑 𝐶 are the possible labels, the number of “votes” that a label receives from the classifiers, and the total number of classifiers.

classifier Instance 1 Instance 2 Instance 3
𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒚𝟏 𝒚𝟐 𝒚𝟑 𝒚𝟏 𝒚𝟐 𝒚𝟑
𝐶$ 0.2 0.7 0.1 0.2 0.7 0.1 0.6 0.1 0.3
𝐶% 0.1 0.3 0.6 0.2 0.6 0.2 0.21 0.21 0.58
𝐶& 0.8 0.1 0.1 0.05 0.9 0.05 0.75 0.01 0.24
𝐶’ 0.3 0.5 0.2 0.1 0.8 0.1 0.1 0.28 0.62
4. Given the following univariate dataset, calculate a statistical model based on the assumption that your data is coming from a normal distribution. Determine whether the instance x=1.2 is anomalous or not if we use the boxplot test?
X = {2, 2.5, 2.6, 3, 3.1, 3.2, 3.4, 3.7, 4, 4.1,4.8}
5. Given the following univariate dataset, determine the outlier score for instances (x=0.5) and (x=4) using the following strategies:
Dataset = {1,1.05, 1.1, 1.15, 1.2, 1.21, 1.3, 1.4, 1.45, 1.5, 4.55, 5.6, 6.8, 7.58, 8.6, 9.7, 10.3, 11.4, 12.3,13.5}

(a) Inverse Relative density using 2-NN (Manhattan distance)
(b) Distance to 2nd nearest neighbor (Manhattan distance)
6. In Assignment 1 we worked with the ‘animals’ dataset. Suggest a suitable method to detect anomalies among animal instances. Would you use a supervised, semi-supervised or unsupervised approach? Can you think of a way to make anomaly detection more reliable?