Learn about different types of machine learning.
- [Instructor] There are two major types of machine learning, supervised and unsupervised. Supervised machine learning is aided by examples of inputs and their desired outputs. In this situation a machine learning or ML program tweaks its algorithm until it can produce best results based on the given examples. Here the roll of the algorithm is to capture a general rule that best maps the inputs to their corresponding outputs.
We separate the example dataset under supervised ML into training and test datasets. As their names suggest the training dataset is only used for teaching an ML algorithm, while the test dataset is for evaluating the effectiveness of the same ML algorithm after training is over. We keep our test datasets much smaller then their counterpart for training because our focus needs to be on training rather then testing to accomplish as high accuracy as possible.
Let's say that your ML goal is to automate the process of differentiating oranges, apples and bananas. In this case you can provide a training dataset consisting of various images of apples, oranges and bananas correctly labeled as apples, oranges and bananas. To eventually test how accurately your ML algorithm is able to discern apples, oranges and bananas you need to set aside a small fraction of the example dataset consisting of correctly labeled images for testing purposes.
Unsupervised ML does not require any of these steps. Your dataset is not labeled at all it is solely the responsibility of your ML implementation to find a particular pattern or structure in the dataset. Clustering is a good example of unsupervised learning. For example your ML program can provide a visual representation of clusters of where a disease occurs most frequently in a geographical area.
You realize that all the clusters are near water sources which implies that the disease is water born. To summarize, supervised learning requires training and human interventions, while unsupervised learning doesn't require neither training nor human input.
- Enabling technologies in data science
- Cloud computing and virtualization
- Installing and working with Proxmox, Hadoop, Spark, and Weka
- Managing virtual machines on Proxmox
- Distributed processing with Spark
- Fundamental applications of machine learning
- Distributed systems and distributed processing
- How Hadoop, Spark, and Weka can work together