Update: You can download a machine learning model to diagnose COVID-19 at https://drive.google.com/file/d/1lq_yx_PSX5FU476cb0sp2olQhzSbfSqG/view?usp=sharing
Only for academic purposes until further notice.
What makes humans one of the most intelligent species on Earth is their ability to think! Now this ability is in itself attributed to the HUMAN BRAIN. This is a very interesting organ composed of various cells called the “Neurons”. These are responsible for the swift working of our nervous systems.
How do Neurons work?
The unique structure of neurons gives them the ability to send and receive multiple signals within the shortest possible time lapse due to neurotransmitters within the synapse. This idea itself inspired the formation of neural networks in computers to drastically increase their speed and make them work like a human brain.
What are neural networks and how do they work?
Computer networks that try to vaguely copy the synaptic connections of the neurons in an animal brain are termed as “Artificial Neural Networks” (ANN) or “Connectionist Systems”. The concept was adopted for the first time by Warren McCulloch and Walter Pitts in 1943. Following which, multiple researches have taken place.
These are based on a series of interconnected nodes which try to act like neurons. Each of the nodes is capable of receiving, processing and transmitting data, thus forming and integrated and interconnected system for the unidirectional data flow. Hence, data can be received from multiple nodes at once, be processed and be transmitted to multiple nodes at once as well. The connections between these nodes are termed as edges. The learning proceeds of these nodes and edges are termed as weights. This weight is what increases or decreases the signal’s strength at a connection.
Furthermore, these nodes are arranged in multiple layers amidst which a signal might travel multiple times before the final output, thus increasing accuracy. In between the flow of data from input to output, it is processed via different nodes at a time leading to various permutations and combinations and thus helping the machine to learn. For instance, ANN would recognize the image of a car based on its previous memory of a car and not upon its features.
How do we know if a machine is learning?
A machine is said to have been learning based on the hyper-parameter values. It includes the learning, the number of hidden layers involved in processing and the batch size. The value is determined before the learning begins and thus the process of deep learning could be quantified.
Applications
Given the backdrop of Deep learning, multiple applications have been derived including in Games, Quantum Computing, Artificial Intelligence, Face Identification, learning about the Spread of Cancer and other medical diagnosis, Drones and auto-drives among many others. Thus, it is not bound by a single field and transits across disciplines leading us to the future.
Conclusion
Technology is evolving and is day by day getting closer to or even better than the humans. The future computations with the help of these machines shall lead the humanity into the new area of knowledge and help us solve the mysteries of the universe. In this humongous process, the ANN is one of the giant steps.
Comments