Artifical Neural Network (ANN) is a concept of knowledge engineering in the field of artificial intelligence that is designed by adopting the human nervous system, whose processing is primarily in the brain. The smallest part of the human brain is a nerve cell called the basic unit of information processing or neuron. There are about 10 billion neurons in the human brain and about 60 trillion connections between neurons in the human brain. By using these neurons simultan, the human brain can process information in parallel and quickly, even faster than the current fastest computer. A neuron consists of the following elements of the cell body, a number of fibers that channel information to neurons and a single fiber coming out of a neuron called an axon.
Any external signal received by the dendrite will pass through the synapse to be forwarded to the neuron, then processed in the soma. After the ends will be removed through the axon to be re-processed by another neuron or out as the final result of brain process.
By analogy of the working system of the human brain, ANN consists of a processing unit called a neuron (an axon in the human brain) that contains activation enhancers and functions, a number of weights (synapses in the human brain), a number of input vectors (dendrites in the human brain). The activation function is useful for regulating the output given by neurons.
Characteristics of ANN seen from the relationship patterns between neurons, the method of determining the weight of each connection, and the activation function. The picture above explains the basic ANN structure, which in reality is not just as simple as it is.
Neural network is built from mnodes/unitsnits connected by the link directly. Links from one unit to another are used to propagate activation from the first unit to the next. Each link has numerical weights. This weight determines the strength and the marker of a connectivity.
The process in ANN starts from the input received by the neuron along with the weighted value of each input. After entering the neuron, the existing input value will be summed by a function of the propagation (summing function), which can be seen as in the image with the symbol of sigma (Σ). The sum result will be processed by the activation function of each neuron, here will be compared with the sum result with a certain threshold (threshold value). If the value exceeds the threshold, then the activation of the neuron will be canceled, on the contrary, if it is still below the threshold value, the neuron will be activated. Once active, the neuron sends the output value through its output weight to all the neurons associated with it. This process will continue to repeat in the next inputs.
ANN consists of many neurons in it. These neurons will be grouped into several layers. The neurons present in each layer are linked to neurons in other layers. This certainly does not apply to the input and output layers, but only the layers are in between. The information received at the input layer goes to the layers in ANN one by one until it reaches the last layer / output layer. Layers that lie between input and output are referred to as hidden layers. However, not all ANNs have hidden layers, there are also only input and output layers.
Any external signal received by the dendrite will pass through the synapse to be forwarded to the neuron, then processed in the soma. After the ends will be removed through the axon to be re-processed by another neuron or out as the final result of brain process.
Advertisement
By analogy of the working system of the human brain, ANN consists of a processing unit called a neuron (an axon in the human brain) that contains activation enhancers and functions, a number of weights (synapses in the human brain), a number of input vectors (dendrites in the human brain). The activation function is useful for regulating the output given by neurons.
Characteristics of ANN seen from the relationship patterns between neurons, the method of determining the weight of each connection, and the activation function. The picture above explains the basic ANN structure, which in reality is not just as simple as it is.
- Input, works like dendrite
- Output, functions like an axon
- The activation function acts like a synapses
Neural network is built from mnodes/unitsnits connected by the link directly. Links from one unit to another are used to propagate activation from the first unit to the next. Each link has numerical weights. This weight determines the strength and the marker of a connectivity.
The process in ANN starts from the input received by the neuron along with the weighted value of each input. After entering the neuron, the existing input value will be summed by a function of the propagation (summing function), which can be seen as in the image with the symbol of sigma (Σ). The sum result will be processed by the activation function of each neuron, here will be compared with the sum result with a certain threshold (threshold value). If the value exceeds the threshold, then the activation of the neuron will be canceled, on the contrary, if it is still below the threshold value, the neuron will be activated. Once active, the neuron sends the output value through its output weight to all the neurons associated with it. This process will continue to repeat in the next inputs.
ANN consists of many neurons in it. These neurons will be grouped into several layers. The neurons present in each layer are linked to neurons in other layers. This certainly does not apply to the input and output layers, but only the layers are in between. The information received at the input layer goes to the layers in ANN one by one until it reaches the last layer / output layer. Layers that lie between input and output are referred to as hidden layers. However, not all ANNs have hidden layers, there are also only input and output layers.
No comments