model_code.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn as nn\n",
    "from torch import optim as optim\n",
    "from torch.utils import data as Data\n",
    "from torchvision.datasets import __file__\n",
    "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "#     ''' \n",
    "#     运行环境:PyTorch1.8.0 & Python3.8.1 & Cuda11.1 & Cudnn8.0\n",
    "#     CPU：i7-10070F  RAM:16G 2600 GPU:RTX3070 \n",
    "#     '''\n",
    "class attention1D_layer(nn.Module):\n",
    "#     ''' \n",
    "#     注意力机制模块\n",
    "#     input   x= [batch_size,channels(信号通道数),points(采样点数目)] 多通道一维信号\n",
    "#     return out = [b*x,b] \n",
    "#     其中:b*x 通道权重乘输入信号 ， b为基于神经得到的各采样点权重\n",
    "       \n",
    "#     '''    \n",
    "    def __init__(self,features):\n",
    "        super(attention1D_layer,self).__init__()\n",
    "        self.features = features\n",
    "        # 一维全局平均池化获得全部通道数均值\n",
    "        self.GobalPooling = nn.AdaptiveAvgPool1d(1)\n",
    "        \n",
    "        # 使用一个带Leakly的线性激活函数的BP网络获取采样点权重值 [channels →channels/4 → channels]\n",
    "        self.Linear1 = nn.Sequential(\n",
    "            nn.Linear(features,int(features/4)),\n",
    "            nn.ReLU6())\n",
    "        self.Linear2 = nn.Sequential(\n",
    "            nn.Linear(int(features/4),features),\n",
    "            nn.ReLU6())\n",
    "        \n",
    "    #前向传播函数\n",
    "    def forward(self,x):    #传播前，进行维度交换  [batch_size,channels,points]→[batch_size,points,channels]\n",
    "        b = self.GobalPooling(x.transpose(1,2))\n",
    "        #全局池化后释放最后一个维度通道数(channels)\n",
    "        b = b.squeeze(-1)\n",
    "        #送入BP网络获得b\n",
    "        b = self.Linear2(self.Linear1(b))\n",
    "        b = torch.mean(b,0).reshape(1,1,self.features)\n",
    "        #b = nn.Softmax(b)\n",
    "        return b*x,b.squeeze(0).squeeze(0)\n",
    "        \n",
    "class Inception_layer(nn.Module):\n",
    "# '''\n",
    "#     多尺度卷积模块\n",
    "#     使用4种模块卷积核处理数据来获取不同大小的感受野 \n",
    "#     参考论文:https://arxiv.org/abs/1512.00567 及 InceptionE InceptionV3 InceptionV4\n",
    "# '''\n",
    "    def __init__(self,in_features,out_features):\n",
    "        super(Inception_layer,self).__init__()\n",
    "        self.in_features = in_features\n",
    "        self.Conv1x3 = nn.Sequential(\n",
    "            nn.Conv1d(in_features,10,kernel_size=1,stride=1,padding=0),\n",
    "            nn.ReLU(),\n",
    "            nn.Dropout(0.1),\n",
    "            nn.Conv1d(10,out_features,kernel_size=3,stride=1,padding=1),\n",
    "            nn.ReLU()    \n",
    "        )\n",
    "        self.Conv1x5 = nn.Sequential(\n",
    "            nn.Conv1d(in_features,10,kernel_size=1,stride=1,padding=0),\n",
    "            nn.ReLU(),\n",
    "            nn.Dropout(0.1),\n",
    "            nn.Conv1d(10,out_features,kernel_size=5,stride=1,padding=2),\n",
    "            nn.ReLU(),\n",
    "        )\n",
    "        self.Conv3x5 = nn.Sequential(\n",
    "            nn.Conv1d(in_features,10,kernel_size=3,stride=1,padding=1),\n",
    "            nn.ReLU(),\n",
    "            nn.Dropout(0.1),\n",
    "            nn.Conv1d(10,out_features,kernel_size=5,stride=1,padding=2),\n",
    "            nn.ReLU()\n",
    "        )\n",
    "        self.short_cut = nn.Sequential(\n",
    "            nn.Conv1d(in_features,out_features,kernel_size=1,stride=1,padding=0)\n",
    "        )\n",
    "    def forward(self,x):\n",
    "        part_1 = self.Conv1x3(x)\n",
    "        part_2 = self.Conv3x5(x)\n",
    "        part_3 = self.Conv1x5(x)\n",
    "        x = self.short_cut(x)\n",
    "        # 返回四种尺度信号词义\n",
    "        return part_1,part_2,part_3,x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ResPart(nn.Module):\n",
    "# '''\n",
    "#     恒等连接结构 :  by conv1→conv3→conv1\n",
    "#     考虑输入通道数是否需要调整来满足残差相接: \n",
    "#         F(x) = adjust(x) + f(x)\n",
    "#     参考论文结构:https://arxiv.org/pdf/1512.03385.pdf\n",
    "# '''\n",
    "    def __init__(self,in_features,out_features):\n",
    "        self.in_features = in_features\n",
    "        self.out_features = out_features\n",
    "        super(ResPart,self).__init__()\n",
    "        self.in_features = in_features\n",
    "        self.out_features = out_features\n",
    "        if in_features != out_features:\n",
    "            self.conv = nn.Sequential(\n",
    "                nn.Conv1d(in_features,out_features,kernel_size=1,stride=1,padding=0),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1),\n",
    "                nn.Conv1d(out_features,out_features,kernel_size=3,stride=1,padding=1),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1),\n",
    "                nn.Conv1d(out_features,out_features,kernel_size=1,stride=1,padding=0),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1)\n",
    "             )\n",
    "            self.adjust_x = nn.Conv1d(in_features,out_features,kernel_size=1,stride=1,padding=0)\n",
    "        else:\n",
    "            self.conv = nn.Sequential(\n",
    "                nn.Conv1d(in_features,out_features,kernel_size=1,stride=1,padding=0),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1),\n",
    "                nn.Conv1d(out_features,out_features,kernel_size=1,stride=1,padding=0),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1),\n",
    "                nn.Conv1d(out_features,out_features,kernel_size=1,stride=1,padding=0),\n",
    "                nn.ReLU(),\n",
    "                nn.Dropout(0.1)\n",
    "            )\n",
    "    def forward(self,x):\n",
    "        if self.in_features !=  self.out_features:\n",
    "            short_cut = self.adjust_x(x)\n",
    "            x = self.conv(x)\n",
    "            return short_cut.add(x)\n",
    "        else:\n",
    "            return self.conv(x).add(x) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class sbd_mode(nn.Module):\n",
    "    def __init__(self,in_channels,goal_channel=1):\n",
    "        super(sbd_mode,self).__init__()\n",
    "        self.in_channels = in_channels\n",
    "        self.goal_channel = goal_channel\n",
    "        self.ResBody_layer1 = ResPart(in_channels,100)\n",
    "        self.InCeption_layer = Inception_layer(100,100)\n",
    "        self.ResBody_layer2 = ResPart(400,100)\n",
    "        self.attention_layer = attention1D_layer(100)     \n",
    "        self.conv = nn.Conv1d(100,self.goal_channel,kernel_size=1)\n",
    "        self.fully = nn.Sequential(\n",
    "            nn.Linear(100,200),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(200,100))\n",
    "        \n",
    "        \n",
    "    #\n",
    "    def forward(self,x):\n",
    "        x = self.ResBody_layer1(x)\n",
    "        #print(x.shape)\n",
    "        p1,p2,p3,x = self.InCeption_layer(x)\n",
    "        x = torch.cat((p1,p2,p3,x),axis=1)\n",
    "        #print(x.shape)\n",
    "        x = self.ResBody_layer2(x)\n",
    "        x,b = self.attention_layer(x)\n",
    "        x = self.conv(x)\n",
    "        x = x.squeeze(1)\n",
    "        return self.fully(x)\n",
    "model = sbd_mode(in_channels=3).to(device)   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sbd_mode(\n",
      "  (ResBody_layer1): ResPart(\n",
      "    (conv): Sequential(\n",
      "      (0): Conv1d(3, 100, kernel_size=(1,), stride=(1,))\n",
      "      (1): ReLU()\n",
      "      (2): Dropout(p=0.1, inplace=False)\n",
      "      (3): Conv1d(100, 100, kernel_size=(3,), stride=(1,), padding=(1,))\n",
      "      (4): ReLU()\n",
      "      (5): Dropout(p=0.1, inplace=False)\n",
      "      (6): Conv1d(100, 100, kernel_size=(1,), stride=(1,))\n",
      "      (7): ReLU()\n",
      "      (8): Dropout(p=0.1, inplace=False)\n",
      "    )\n",
      "    (adjust_x): Conv1d(3, 100, kernel_size=(1,), stride=(1,))\n",
      "  )\n",
      "  (InCeption_layer): Inception_layer(\n",
      "    (Conv1x3): Sequential(\n",
      "      (0): Conv1d(100, 10, kernel_size=(1,), stride=(1,))\n",
      "      (1): ReLU()\n",
      "      (2): Dropout(p=0.1, inplace=False)\n",
      "      (3): Conv1d(10, 100, kernel_size=(3,), stride=(1,), padding=(1,))\n",
      "      (4): ReLU()\n",
      "    )\n",
      "    (Conv1x5): Sequential(\n",
      "      (0): Conv1d(100, 10, kernel_size=(1,), stride=(1,))\n",
      "      (1): ReLU()\n",
      "      (2): Dropout(p=0.1, inplace=False)\n",
      "      (3): Conv1d(10, 100, kernel_size=(5,), stride=(1,), padding=(2,))\n",
      "      (4): ReLU()\n",
      "    )\n",
      "    (Conv3x5): Sequential(\n",
      "      (0): Conv1d(100, 10, kernel_size=(3,), stride=(1,), padding=(1,))\n",
      "      (1): ReLU()\n",
      "      (2): Dropout(p=0.1, inplace=False)\n",
      "      (3): Conv1d(10, 100, kernel_size=(5,), stride=(1,), padding=(2,))\n",
      "      (4): ReLU()\n",
      "    )\n",
      "    (short_cut): Sequential(\n",
      "      (0): Conv1d(100, 100, kernel_size=(1,), stride=(1,))\n",
      "    )\n",
      "  )\n",
      "  (ResBody_layer2): ResPart(\n",
      "    (conv): Sequential(\n",
      "      (0): Conv1d(400, 100, kernel_size=(1,), stride=(1,))\n",
      "      (1): ReLU()\n",
      "      (2): Dropout(p=0.1, inplace=False)\n",
      "      (3): Conv1d(100, 100, kernel_size=(3,), stride=(1,), padding=(1,))\n",
      "      (4): ReLU()\n",
      "      (5): Dropout(p=0.1, inplace=False)\n",
      "      (6): Conv1d(100, 100, kernel_size=(1,), stride=(1,))\n",
      "      (7): ReLU()\n",
      "      (8): Dropout(p=0.1, inplace=False)\n",
      "    )\n",
      "    (adjust_x): Conv1d(400, 100, kernel_size=(1,), stride=(1,))\n",
      "  )\n",
      "  (attention_layer): attention1D_layer(\n",
      "    (GobalPooling): AdaptiveAvgPool1d(output_size=1)\n",
      "    (Linear1): Sequential(\n",
      "      (0): Linear(in_features=100, out_features=25, bias=True)\n",
      "      (1): ReLU6()\n",
      "    )\n",
      "    (Linear2): Sequential(\n",
      "      (0): Linear(in_features=25, out_features=100, bias=True)\n",
      "      (1): ReLU6()\n",
      "    )\n",
      "  )\n",
      "  (conv): Conv1d(100, 1, kernel_size=(1,), stride=(1,))\n",
      "  (fully): Sequential(\n",
      "    (0): Linear(in_features=100, out_features=200, bias=True)\n",
      "    (1): ReLU()\n",
      "    (2): Linear(in_features=200, out_features=100, bias=True)\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## train \n",
    "import hiddenlayer as h1\n",
    "optimizer = optim.Adam(model.parameters())\n",
    "loss_func = nn.MSELoss()\n",
    "history1 = h1.History()\n",
    "canvas1 = h1.Canvas()\n",
    "for epoch in range(500):\n",
    "    all_loss = 0\n",
    "    for step,(b_x,b_y) in enumerate(data_gen):\n",
    "        b_x = b_x.to(device)\n",
    "        b_y = b_y.to(device)\n",
    "        pre = model(b_x)\n",
    "        loss = loss_func(pre,b_y)\n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        all_loss += loss.item()\n",
    "    epo_loss = all_loss / 10\n",
    "    history1.log(epoch,train_loss=epo_loss)\n",
    "    \n",
    "    with canvas1:\n",
    "        canvas1.draw_plot(history1['train_loss'])   "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}