Decoding ML, DL and AI
Every newbie entering into the magical world of machine intelligence, is confused in terms like Machine Learning, Deep Learning, and Artificial Intelligence. In this article, we are going to dive in these categories of machine intelligence techniques revolutionizing the world by solving complex problems in the field of healthcare, finance, education, transportation, decision analysis and support systems.
Artificial intelligence is the broadest and oldest term and a superset of Machine Learning and Deep Learning. Deep learning are specific machine learning algorithms which uses neural networks.
Let’s discuss each one by one -
Artificial Intelligence (AI)
AI is an intelligent computer system which is able to learn and comprehend complex problems which require human intelligence in general. Although, it was initially developed to mimic human behavior, but it does not necessarily need to confine to humans or biologically observable intelligence in the modern world.
There are infinite applications of AI which includes personalized shopping, AI Chatbots (you must have heard of ChatGPT, also used in the research of this article and deserves a citation), self-driving cars, face recognition, personalized advertisement (Did you hear about the Zomato ad with Hritik Roshan saying the exact name of the city and restaurants according to the location of viewers), recommendation systems used by streaming platforms and many more, I just said infinite in the starting.
Machine Learning (ML)
ML is a subset of Artificial Intelligence which consists of algorithms which use the dataset to find patterns in the training data, and generalizes it to a model which can be used on unknown dataset. Statistical techniques are applied while feature generation and optimizing the error function.
ML model learning can be -
- Supervised (Labelled Data)
- Regression
- Logistic Regression
- Classification
- Naive Bayes Classifiers
- K-NN (k nearest neighbors)
- Decision Trees
- Support Vector Machine
2. Unsupervised (Unlabeled Data)
- Clustering algorithms
- Principal Component Analysis (PCA)
- Singular Value Decomposition
Deep Learning (DL)
DL is a subset of ML which is inspired by the functionality of human brain, it uses layer of nodes which acts as neurons of brain and connection between them. Functionality of node is basically a linear regression, a linear combination of feature vectors with a bias.
Deep word here is used for more than 3 hidden layers inside a neural network. Changing the type, number of layers, neurons and activation functions can change the complexity of the DL model. We need to find these parameters to properly tune the DL model.
Deep Learning vs Machine Learning
- DL models require a lot more data than ML.
- DL models takes a lot of computational power and time to build the model.
- DL models are more accurate than ML.
- DL models requires less human intervention as feature extraction and processing is done inside the neural network itself.
- DL models are less interpretable than ML.
References