Unsupervised Learning Algorithms

Well,…Unsupervised learning in itself says not supervised learning.

Intuitively speaking,most of human and animal learning is unsupervised learning. We are not given right answer every where. We just make decision we find to be right and try to do less mistakes next time.

Just like above, unsupervised learning is the process of learning by machines without having labels (right answer) and making less errors next time(clustering does it by placing data in nearest clusters). And yeah , unsupervised learning is the actual AI that has so much hype for. Supervised learning has labels(output) and features(inputs) both which make it comparatively easier to solve supervised learning problems than unsupervised learning problems.

Though unsupervised learning problem is harder to solve, it is indispensable for development of AI.

Few examples of unsupervised learning includes using anomaly detection, recommendation system, time series clustering, etc. This examples might sound formidable, so for now just take youtube’s recommendation system as its example.

Now, I will describe few advantages and disadvantages of unsupervised learning.

1. ADVANTAGES

a. As we know unsupervised learning is where labels are not provided and most of the data in the world is unlabeled, unsupervised learning is really helpful for real world data .It helps label these data by clustering process. Clustering means dividing data into different clusters with similar characters and providing them labels on our own which has made humans easier to handle unlabeled data.

b. For problems where patterns are unknown or constantly changing ,unsupervised learning shines.

c. Unsupervised learning is the core to building strong AI ,which is what we are all working at the present for. Advent of superintelligence must be preceded by unsupervised learning.

d.Most problems that could not be solved previously using supervised learning is solved using unsupervised.

e. It helps prevent overfitting problem , helps for feature selection using dimensionality reduction, helps finding outliers in the data and many more.

2. DISADVANTAGE

Albeit unsupervised learning has many applications and avails ,still it does have some drawbacks in it.

a. It is not suitable for narrowly defined specific tasks.

b. It gets trounced by supervised learning at narrowly defined tasks for which we have well-defined patterns that do not change over time and for labeled datasets.

ALGORITHMS

There are many algorithms for unsupervised learning and describing each would take a whole lot of time ,so I provide here a brief inventory of most important unsupervised learning algorithms used in industry:

1.Dimensionality reduction

a. Linear Projection Dimensionality Reduction(Linear data)

b. Manifold Learning(Non linear data)

c.Independent Component Analysis

2.Clustering

a. K-means clustering

b. Hierarchical Clustering

c. DBSCAN

3.UNSUPERVISED DEEP LEARNING

a. Restricted Boltzmann’s machines

b. Deep belief networks

NOTE: If you can solve a problem using supervised learning then you should not go for unsupervised learning because it provides complexity and is costly too. So, always first go for supervised learning then unsupervised learning. Regarding algorithms too, first use machine learning algorithms then use deep learning algorithms if the problem is not solved by machine learning algorithms.

Oh, yessss😂….finally the article is over and I hope you received a little bit of wisdom from this modicum amount of writing. I will be writing about these algorithms in the next article and hope to catch you there folks.

I hope all of you to be safe, healthy and happy in your life.

God bless y’all and good bye pals.😁.