Unsupervised learning algorithms
1. Introduction:
The models that are trained without using the supervised training set or labelled data set is called as unsupervised learning. This task is similar to grouping the data based upon similarity using different techniques.
2. Types of unsupervised learning
i. Clustering: simply from its name, its the grouping of the data based upon the similarity. easy, right?
ii. Association: It becomes easier if you think as a problem set for recommendation system. The user data and item data are associated and that relationship is used further for recommending the item.
3. Moving on unsupervised learning algorithms
a) K-means clustering
b) k-nearest neighbors
c) Anomaly detection
d) Neural networks
e) Singular value decomposition
f) spectral clustering
g) Divisine clustering
a) K-means clustering
The name itself gives hint ‘k’ which means the algorithm tries to separate the data into k clusters.
easier right > ?
dive in
But how the clusters are found ?
>> ok , lets ponder around more.
step 2 : lets randomly shuffle the data and set centroid. Then select k data points for the centroids without replacement.
step 3 : iterate until while there is no change in centroid
3.1 compute sum of squared distance between data points and all centroids
3.2 assign each data point to closest centroid point
3.3 compute centroids for the clusters by taking the average of the all data points
This approach is called as Expectation-Maximization.
E means to assign datda to closest centroid point and m-step is computing centroid of each cluster
