NEURAL NETWORKS
When it comes to machine learning or deep learning , artificial neural network perform really well. Artificial neural network are used in various classification task like audio, words and images.
The word neural is derived from the word "NEURONS". Let us see first what are neurons... The nerve cells that send and receive electrical signals over long distance WITHIN the body. The neurons has three parts cell body, dendrites and axon.
When it comes to machine learning or deep learning , artificial neural network perform really well. Artificial neural network are used in various classification task like audio, words and images.
The word neural is derived from the word "NEURONS". Let us see first what are neurons... The nerve cells that send and receive electrical signals over long distance WITHIN the body. The neurons has three parts cell body, dendrites and axon.
The Artificial Neutral Networks is a information producing paradigm (pattern) which is provoked by the biological nervous system, like brain process information or data. The paradigm is the specific structure of the processing system.
Network is the advancement of the computers. It evolved in and after 1960's . The first artificial neuron was produced by neuro-physiologist Warren McCulloch and the logician Walter Pits in 1943. But the technology at that time does not allow to do much .
Neural Networks take a different approach to problem solving than conventional computers.
Now you would be thinking what is conventional computers.?
Lets us have a glance, computer hardware does not execute the user source program .So, to execute the source programs there are some compilers which converts the high level language to low level language which is executed by the hardware.
conventional computers follows the the specific approach or the set of instructions in order to solve the problem.
And the solution of this conventional computers are predictable.
Where as Neural Networks works same like as our HUMAN brain does.!
this network is composed of highly interconnected processing entity(NEURONS) working parallel through out the network. there are some tasks or problem which need both the algorithmic methods and neural networks.The artificial neuron is the device which have several inputs and only one output.
Different types of neural network are used for different purposes.
It was the first and simplest form of artificial neural network devices. Here, in this network information moves only in one direction that is forward direction, through the input nodes, hidden nodes, and output nodes. there is no cycle and loops in this network.
This Neural network is composed of more than one neural network connected by one intermediary.In the mid of 1980's, there was a concept of ensemble learning where a collection of "weak" and "simple" learner can out-space the deep learning model. In the modular neural network, there is the principle of " DIVIDE AND CONQUER" which breaks large problems into the variable parts and solve the problem.
Convolutional neural network consist of various input layers and output layer, as well as hidden layer. It is the subclass of deep neural network and mostly used as the analyzing visual imagery, i.e it is used for image recognition and classification. They are highly proficient in areas like identification of faces, objects, robots, etc.
There are more neural network such as Radial Neural network, Recurrent Neural Network, Kohonen Self Organizing Neural Networks.
Deep neural network are progressing rapidly within many fields we have seen this technology diagnose medical conditions with more accuracy then the trained artificial intelligence experts.
. "Audible" is one of the application product which play the audiobook. audible massive range of audiobooks and one of the best way to bench content while you enjoy your holidays or any activity.
"Super intelligence" by Nick Bostrom the book(audio also available) goes through the history of Artificial Intelligence and the path that we might get the super intelligence. It is a good read and i recommend it.
Let us see some interesting example of neural networks which will actually clear your concepts about neural network.
1. Restore black and white photos to colors. It happen just because of neural networks. In old days it was done by some specialist but there was a way to do it instantly and it is neural network. due to this it restore automatically by the information the machine have.
2. Pixel enhancing CID style. you know that the thing that they do in movies and tv crime shows. the grainy image be enhanced and somehow magically it becomes HD .It is just because of Neural net.
Google brain researches have trained a deep learning neural network to very low resolution image of faces and predict what those faces will they lookalike. They called this method as pixel recursive super resolution and it enhance the resolution of photos dramatically in the image below you can see.
4. Lip reading. Lip net is the neural network developed by oxford university and google deep mind scientist. This network can watch silent video of the person talking and convert it into the text. Lip net has raised to 95% accuracy in reading people lips. In an average lip reader has accuracy of 50 to 60% .
What kind's of content would you like to see more on this blog? Let me know in the comments.
Thank you for reading this blog...
Neural Networks take a different approach to problem solving than conventional computers.
Now you would be thinking what is conventional computers.?
Lets us have a glance, computer hardware does not execute the user source program .So, to execute the source programs there are some compilers which converts the high level language to low level language which is executed by the hardware.
conventional computers follows the the specific approach or the set of instructions in order to solve the problem.
And the solution of this conventional computers are predictable.
Where as Neural Networks works same like as our HUMAN brain does.!
this network is composed of highly interconnected processing entity(NEURONS) working parallel through out the network. there are some tasks or problem which need both the algorithmic methods and neural networks.The artificial neuron is the device which have several inputs and only one output.
Different types of neural network are used for different purposes.
Various Types of Neural Networks
- Feed-forward neural network (Artificial Neural Network):
It was the first and simplest form of artificial neural network devices. Here, in this network information moves only in one direction that is forward direction, through the input nodes, hidden nodes, and output nodes. there is no cycle and loops in this network.
- Modular Neural Network
This Neural network is composed of more than one neural network connected by one intermediary.In the mid of 1980's, there was a concept of ensemble learning where a collection of "weak" and "simple" learner can out-space the deep learning model. In the modular neural network, there is the principle of " DIVIDE AND CONQUER" which breaks large problems into the variable parts and solve the problem.
- Convolutional neural network
Convolutional neural network consist of various input layers and output layer, as well as hidden layer. It is the subclass of deep neural network and mostly used as the analyzing visual imagery, i.e it is used for image recognition and classification. They are highly proficient in areas like identification of faces, objects, robots, etc.
There are more neural network such as Radial Neural network, Recurrent Neural Network, Kohonen Self Organizing Neural Networks.
Deep neural network are progressing rapidly within many fields we have seen this technology diagnose medical conditions with more accuracy then the trained artificial intelligence experts.
. "Audible" is one of the application product which play the audiobook. audible massive range of audiobooks and one of the best way to bench content while you enjoy your holidays or any activity.
"Super intelligence" by Nick Bostrom the book(audio also available) goes through the history of Artificial Intelligence and the path that we might get the super intelligence. It is a good read and i recommend it.
Let us see some interesting example of neural networks which will actually clear your concepts about neural network.
1. Restore black and white photos to colors. It happen just because of neural networks. In old days it was done by some specialist but there was a way to do it instantly and it is neural network. due to this it restore automatically by the information the machine have.
2. Pixel enhancing CID style. you know that the thing that they do in movies and tv crime shows. the grainy image be enhanced and somehow magically it becomes HD .It is just because of Neural net.
Google brain researches have trained a deep learning neural network to very low resolution image of faces and predict what those faces will they lookalike. They called this method as pixel recursive super resolution and it enhance the resolution of photos dramatically in the image below you can see.
3. Generating new images. Generating pics to pics there is deep learning neural networks that generate the new image based on the input, but neural net was trained to perform multiple but specific tasks either create real street screen from colored shape or make a photorealistic image just from an outline . this is done by the neural networks.
4. Lip reading. Lip net is the neural network developed by oxford university and google deep mind scientist. This network can watch silent video of the person talking and convert it into the text. Lip net has raised to 95% accuracy in reading people lips. In an average lip reader has accuracy of 50 to 60% .
What kind's of content would you like to see more on this blog? Let me know in the comments.
Thank you for reading this blog...
Comments
Post a Comment