Supervised Learning Machine Based On Monte Carlo And Application Development

Neural Network can be traced back to 1940s with the presence of M-P neuron model. Even though it stuck during the medium term, people still have confidence that this abstract and simulated methods derived from human brains’ ability for features extracting would finally bring up large impact. Meanwhile neural network is developing rapidly. By far the significance of neural network has already obtained a broad recognition and is currently on a further creative developing. It is widely spread and applied in many areas including machine learning, pattern recognition, artificial intelligence, bioinformatics and even robotics.And also there would two different groups based on the layers of hidden layer in neural network—shallow neural network and deep neural network. They are all been deployed and compared by more and more people nowadays.Neural network was limited due to the computer hardware conditions and algorithms development. But in 2006, with the impulse of a revised RBM algorithm proposed by Prof. Geoffrey E. Hinton and the available of more complex computer hardware, deep neural network has been a hot area that more and more research and application are undertaken.Since it is much more similar to the multi-layers working mechanism of our human brains, deep neural network is more generally accepted by people. The most important part of neural network is the multi hidden layers. And that is also the main cause about why it has a more powerful ability of abstracting features than shallow neural network. In the moment deep neural network is being developing in a fascinating speed especially when it comes to deal with big data.In this paper we will discuss about a supervised learning machine based on Monte Carlo method. It is a shallow neural network and is being studied on time series prediction and pattern recognition. The advantage of shallow neural network is its flexibility in its designing algorithms. Not only the training is secured but also the mechanism is flexible thanks to Monte Carlo method. As mentioned before we will conduct two different experiments which are time series prediction and classification of bioinformatics. For the part of classification of bioinformatics we will undertake a further discussion about comparison between the supervised learning machine based on Monte Carlo and deep neural network. We should take a deep consideration about how to choose between deep and shallow neural network in the mean time when shallow neural network maintains some specific advantages.