Machine Learning Notes-1 (Introduction and Learning Types)
“It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of light, it was the season of darkness, it was the spring of hope, it was the winter of despair.” Charles Dickens, A Tale of Two Cities
Machine learning is a computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E. (Tom Mitchell, 1997)
Machine learning is field of study that gives computers the ability to learn without being explicitly programmed. (Arthur Samuel 1959)
Lets first clear the confusion around the buzzwords of artificial intelligence, machine learning and deep learning. I think the following image will help to distinguish these three concepts from each other.
Artificial intelligence is any technique which enables computers to mimic human behavior.
Machine learning is artificial intelligence techniques that give computers the ability to learn without being explicitly programmed to do so.
Deep learning is a subset of machine learning techniques which make the computation of multi-layer neural networks feasible. Extracting patterns from data using neural networks.
Eg. in a neural network at each iteration the weights of the model are updated, each new set of weights is a new hypothesis offered by the ML algorithm.
- Supervised Learning
- Unsupervised Learning
- Semi-supervised Learning
- Reinforcement Learning
Supervised Learning : The training data fed to the algorithm includes the desired solutions (labels). Classification and regression problems fits to supervised learning. E.g. handwritten digits recognition system, spam mail filtering system are classification problems which can be solved by supervised learning. Predicting a house price using its room number, age, area it is in etc. is a a problem of regression.
Classification is about predicting a label and regression is about predicting a quantity.
Unsupervised Learning : The training data is not labeled, the algorithm tries to learn without a supervision. Usually clustering, anomaly detection problems are such learning problems.
Semi-supervised Learning : It is a combination of supervised and unsupervised learning. When the data set has a number of labeled data and majority of unlabeled data the unlabeled data is tried to be labeled using a model trained on labeled data. That newly labeled data can be called pseudo labeled. At final both labeled and pseudo labeled data is used to train a model together to achieve more accurate results. Another approach is to cluster the unlabeled data and use the labeled data to label clusters.
Reinforcement Learning : It is a type of learning technique enables an agent to learn in an interactive environment by trial and error and using the feedback of its actions. These feedback signals are usually either a reward or a punishment. It is more based on gaining experience in the environment.