十月 11, 2018
發表者: 外匯維基團隊
類別: 外匯交易系統

※ Neural Network ─ Hull Transferring Common (健康管理協會) & 偏差尺度傳輸公共 (迪斯曼公司) ※

☛ Makes use of HMA algorithm however this one is a variation from low-lag to zero-lag
after which... 與下一個融合:
↓
☛ Jurik 過濾器/平滑和定制 MA 品種
☛ Mixed with Deviation-Scaled Transferring Common Algorithm
☛ 更高 & 最棒的方法 (更高 & APB計算)

意識到: Greatest use with Volumes on Primary Chart indicator ── advisable for indicator set off/replace motion

Transient principle of Neural Networks:

The Neural community is an adjustable mannequin of outputs as capabilities of inputs. It consists of a number of layers:

enter layer, which consists of enter knowledge
hidden layer, which consists of processing nodes known as neurons
output layer, which consists of 1 or a number of neurons, whose outputs are the community outputs.

All nodes of adjoining layers are interconnected. These connections are known as synapses. Each synapse has an assigned scaling coefficient, by which the info propagated by means of the synapse is multiplied. These scaling coefficient are known as weights (w[我][j][k]). 在一個 Feed-Ahead Neural Network (FFNN) the info is propagated from inputs to the outputs. Right here is an instance of FFNN with one enter layer, one output layer, and two hidden layers:

連接圖片

The topology of a FFNN is commonly abbreviated as follows: <# of inputs> - <# of neurons within the first hidden layer> - <# of neurons within the second hidden layer> -...- <# of outputs>. The above community might be known as a 4-3-3-1 community.
The info is processed by neurons in two steps, correspondingly proven throughout the circle by a summation signal and a step signal:

All inputs are multiplied by the related weights and summed
The ensuing sums are processed by the neuron's activation operate, whose output is the neuron output.

It's the neuron's activation operate that offers non-linearity to the neural community mannequin. 沒有它, there isn't a motive to have hidden layers, and the neural community turns into a linear auto-regressive (AR) mannequin.

✜ 偏差尺度傳輸公共 (迪斯曼公司) ✜

全新的 迪斯曼公司 made by John Ehlers and featured within the July 2018 situation of TASC journal.

The DSMA is a knowledge smoothing method that acts as an exponential transferring common with a dynamic smoothing coefficient. The smoothing coefficient is mechanically up to date primarily based on the magnitude of value modifications. Within the Deviation-Scaled Transferring Common, 選擇與暗示的通常偏差作為該幅度的度量. The ensuing indicator offers substantial smoothing of the info even when value modifications are small whereas shortly adapting to those modifications.

The writer explains that because of its design, it has minimal lag but is ready to present appreciable smoothing. 儘管如此, 神經網路 - 健康管理協會 & DSMA indicator is fused with Jurik filters/smoothing mixed with zero-lag HMA system.