About Deep Learning
Deep learning is a sub domain of machine learning that deals mostly with unsupervised learning of machines. Deep learning algorithms are capable of understanding the unstructured or unlabelled data as they are inspired by the working of brains cells called neurons. Human brain is the most complex structure built by nature, with billions of neurons stacked in multiple layers with each other through trillions of connections. These neurons play key role in building the human intelligence.
For machines, the deep learning algorithms are known as deep neural networks or artificial neural networks or ANN. An artificial neural network consists of arrays of algorithms in which the initial layer or input layer takes the input from surrounding environment. This input is then segregated in multiple units like pixels in case of an images or frequency in case of sound signals. These units are then coupled with random weights and are then passed on to the next hidden layer of algorithms and the process continues to have the final output.
Artificial neural network keeps a check on the output value and adjust the weights in entire network till the time the output becomes same as the input. This adjustment of weights results in self learning of the deep learning algorithms. Its is also a form of reinforcement learning techniques.
Deep learning requires immense computing power and a lot of time. Such capabilities are not available in normal computer processers with few cores. It requires multiple core processors working simultaneously, which is now possible with the help of technological advancement in the form of GPU or graphics processing unit.
Difference between Deep Learning and Machine Learning
Both Machine learning and Deep learning are part of artificial intelligence that are focused on reducing the human effort and intervention while training a machine. But key differentiator is the type of data that is being used as input.
In machine learning, some level of human intervention is always required and the algorithms mostly use structured or labeled data to train the algorithm.
Applications of Deep Learning
- Speech recognition & Natural language processing.
- Vision & Image processing – Face detection, motion detection and surveillance.
- Entertainment – recommendation on OTT applications like Netflix, YouTube, Spotify etc. based on usage.
- Fraud Detection – by analyzing user data and anomaly detection.
- Demand Forecasting – by analyzing consumption patterns.
- Self-Driving Cars