官术网_书友最值得收藏!

Getting started with deep feedforward neural networks

A deep feedforward neural network is designed to approximate a function, f(), that maps some set of input variables, x, to an output variable, y. They are called feedforward neural networks because information flows from the input through each successive layer as far as the output, and there are no feedback or recursive loops (models including both forward and backward connections are referred to as recurrent neural networks).

Deep feedforward neural networks are applicable to a wide range of problems, and are particularly useful for applications such as image classification. More generally, feedforward neural networks are useful for prediction and classification where there is a clearly defined outcome (what digit an image contains, whether someone is walking upstairs or walking on a flat surface, the presence/absence of disease, and so on).

Deep feedforward neural networks can be constructed by chaining layers or functions together. For example, a network with four hidden layers is shown in the following diagram:

Figure 4.1: A d eep feedforward neural network

This diagram of the model is a directed acyclic graph. Represented as a function, the overall mapping from the input, X, to the output, Y, is a multilayered function. The first hidden layer is H1=f(1)(X, w1 a1), the second hidden layer is H2=f(2)(H1, w2 a2), and so on. These multiple layers can allow complex functions and transformations to be built up from relatively simple ones.

If sufficient hidden neurons are included in a layer, it can approximate to the desired degree of precision with many different types of functions. Feedforward neural networks can approximate non-linear functions by applying non-linear transformations between layers. These non-linear functions are known as activation functions, which we will cover in the next section.

The weights for each layer will be learned as the model is trained through forward- and backward-propagation. Another key piece of the model that must be determined is the cost, or loss, function. The two most commonly used cost functions are cross-entropy, which is used for classification tasks, and mean squared error (MSE), which is used for regression tasks.

主站蜘蛛池模板: 屏东市| 辽中县| 桓台县| 勃利县| 梁平县| 田东县| 林州市| 庆安县| 读书| 清原| 乌兰察布市| 常州市| 福海县| 云南省| 安庆市| 罗田县| 台湾省| 吉林市| 河曲县| 灵璧县| 苍溪县| 乌拉特中旗| 商水县| 延寿县| 康定县| 冕宁县| 彭山县| 万宁市| 东乡| 交口县| 连平县| 梓潼县| 嫩江县| 凤凰县| 罗源县| 楚雄市| 仙桃市| 湟源县| 宁晋县| 安岳县| 云阳县|