Biological and Artificial Neural Network

So, we know that neural networks are processing systems, which are used to process data and information, and this neural network, okay, the design of neural networks, is inspired by the structure and functioning of the human brain. Okay, so the word itself, neurons, okay, the neurons, which are the main functional and structural unit of the human brain, is found in this neural network, this noodle part. So, basically, what we are trying to do with the artificial neural network is to replicate the functions of the human brain. We are trying to create an artificial human brain. This is where artificial intelligence comes into play. A neural network is an important component, an important part of it. So, we know that the human brain consists of a large number of neurons. So, approximately, it has 10 to the power of 11 neuron cells, which are connected to each other. They are interconnected. So, a biological neuron, it looks something like this—a neuron cell. Now, it has various structural parts, and we are not going to discuss biology here, but still, we should discuss these things because then we can interrelate how the biological kingdom is related to the artificial neuron. Okay, because all of these structural and functional units are present in the artificial neuron, with just the characteristics and features that we have in the artificial neuron being inspired by these functional units. So, we should have a good understanding of each of the functional units—just basic things. These branches will give the tree-like structures that are present on the cell body and on the axon. So, the dendrites, they're connected to the cell body and are used for carrying the signal or information. Okay, these are basically the action potentials of the human brain. We have discussed these things in the bio, the biomedical instrumentation part, in the EEG signal, where the neurons fire action potentials when they are excited by an external stimulus. But that is a different thing—you can check out that video too. So, the dendrites, they carry the action potential of the human brain, which is the electroencephalogram. So, the main parts of the biological neuron, here, as we can see, we have the soma or cell body. Then, we have the axon, the synapse, and the dendrites.So, the next one is the axon, which is a single long connection found on the cell body or soma, which also carries the signal, information, or the EEG potential. Then, we have the synapse. Okay, now the soma or cell body is the main structural unit of the neuron, which carries the nucleus, this portion, and the synapse. Okay, these are the structures which are present on the dendrites and also on the axon. Now, this synapse is an important part because the interconnected neurons, okay, this neuron is connected to the other neuron with the help of these synaptic connections. Okay, this synapse and the exchange of signals, the exchange of action potential in the case of biological neurons, takes place through these synaptic connections. And all of these things, all of these functional and structural things, are studied and are tried to replicate in the artificial neural network. So, this is the biological neural level. An artificial neural network, it looks something like this: a group of these nodes, artificial neurons, which are interconnected. Now, this is a simple neural network. The actual neural networks are much more complicated and complex in terms of their structure, their architecture, and also their functioning. So, this artificial neural network, it is the node or the neuron to which the input signals are applied. Here, the inputs to this neuron are X1, X2, and X3. These are the connecting links, okay, which connect this neuron to the other neurons, and these arrowheads, W1, W2, W3, are the weights associated with each connection, each connecting link, which alter or change the input signal. Okay, and this X_net is the net input to the whole neuron or artificial node. The net input is given by the product of the individual inputs and the weight associated with that connecting link. Here, the net input to this node or neuron is given by X1 * W1 + X2 * W2 + X3 * W3. So, the signals are multiplied with their respective weights, okay, the weight associated with the links in which the signal travels. And then we have F(X_net), okay, which is the activation function.Right now, we just understand the activation function. It determines the output of this node. So, various types of functions are used. It can be a step function, it can be an exponential function, it can be a sigmoidal function—rather, a lot of functions which determine the input-output relationship, how the input is manipulated to generate the output, and it is determined by the activation function. There are various mathematical functions which are utilized to get the activation function. Then, we have the output, which can be of branch type, which can be given to various other nodes. So, this is an artificial neuron. Now, how is this artificial neuron related to this biological neuron? So, the main parts of this artificial neural network, as I said, this is the input. Okay, then we have the weight associated with each input, that is W1, W2, W3, and the inputs are X1, X2, X3. Then, we have the activation function of the neuron, which is F(X_net), and then we have the output. Now, how is this artificial neuron related to the biological neuron?