https:\/\/github.com\/JunreyCen\/…<\/a>\u7684\u4ee3\u7801\u8fdb\u884c\u5b9e\u9645\u8fd0\u884c\u6d4b\u8bd5\uff0c\u4e24\u8005\u5b9e\u9645\u8868\u73b0\u4e0a\u5dee\u5f02\u4e0d\u5927\uff0c\u6548\u679c\u90fd\u633a\u597d\u7684\u3002\u60a8\u5728\u4f7f\u7528\u793a\u4f8b\u7684\u65f6\u5019\u8981\u6ce8\u610f\u4e00\u4e0b\uff0c\u5bf9\uff1a<script src="https:\/\/cdn.jsdelivr.net\/npm\/vue"><\/script>\u00a0\u505a\u4e2a\u5c0f\u4fee\u6539\uff1a<script src="http:\/\/static.runoob.com\/assets\/vue\/1.0.11\/vue.min.js"><\/script>\u5373\u53ef\u3002\u00a0\u795e\u7ecf\u7f51\u7edc\u7684\u62df\u5408\u00a0\u867d\u7136\uff0c\u5230\u8fd9\u91cc\u4f3c\u4e4e\u5df2\u7ecf\u5b8c\u6210\u4e86\u53ef\u5fae\u7f16\u7a0b\u89e3\u51b3\u5f39\u6027\u963b\u5c3c\u51fd\u6570\u62df\u5408\u7684\u95ee\u9898\uff0c\u5e76\u627e\u5230\u4e86\u7a0b\u5e8f\u7684\u7b97\u6cd5\u89e3\u51b3\u65b9\u6848\uff0c\u4f46\u662f\uff0c\u56de\u5230\u667a\u80fd\u5316\u7f16\u7a0b\u601d\u60f3\u91cc\uff0c\u6211\u8111\u6d77\u91cc\u6d6e\u73b0\u4e00\u4e2a\u95ee\u9898\uff1a\u56fe\u8ba1\u7b97\u548c\u795e\u7ecf\u7f51\u7edc\u80fd\u5426\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u5462\uff1f\u5e26\u7740\u8fd9\u4e2a\u95ee\u9898\uff0c\u6211\u53c8\u5b9a\u4e49\u4e86\u4e00\u4e2a\u56fe\u8ba1\u7b97\u7684\u795e\u7ecf\u7f51\u7edc\uff1a
model = tf.keras.Sequential()<\/p>\n\u6dfb\u52a0\u5c42<\/h2>\n\u6ce8\uff1ainput_dim(\u8f93\u5165\u795e\u7ecf\u5143\u4e2a\u6570)\u53ea\u9700\u8981\u5728\u8f93\u5165\u5c42\u91cd\u89c6\u8bbe\u7f6e\uff0c\u540e\u9762\u7684\u7f51\u7edc\u53ef\u4ee5\u81ea\u52a8\u63a8\u65ad\u51fa\u8be5\u5c42\u7684\u5bf9\u5e94\u8f93\u5165<\/h2>\n
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))
model.add(tf.keras.layers.Dense(units=10, input_dim=1, activation=’selu’))<\/p>\n
\u795e\u7ecf\u5143\u4e2a\u6570 \u8f93\u5165\u795e\u7ecf\u5143\u4e2a\u6570 \u6fc0\u6d3b\u51fd\u6570<\/h2>\n
model.add(tf.keras.layers.Dense(units=1, activation=’selu’))<\/p>\n
\u8f93\u51fa\u795e\u7ecf\u5143\u4e2a\u6570<\/h2>\n2 \u8bbe\u7f6e\u4f18\u5316\u5668\u548c\u635f\u5931\u51fd\u6570<\/h2>\n
model.compile(optimizer=’adam’, loss=’mse’)<\/p>\n
\u4f18\u5316\u5668 \u635f\u5931\u51fd\u6570(\u5747\u65b9\u8bef\u5dee)<\/h2>\n3 \u8bad\u7ec3<\/h2>\n
history = model.fit(t_x, t_y,epochs=2000)<\/p>\n
4 \u9884\u6d4b<\/h2>\n
y_pred = model.predict(t_x)\u8fd9\u91cc\u6211\u7528 TensorFlow \u7684\u00a0keras\u00a0API \u5b9a\u4e49\u4e86\u5341\u4e2a\u00a0Dense\u00a0\u5c42\uff0c\u6bcf\u5c42\u6709\u5341\u4e2a\u795e\u7ecf\u5143\u3001\u4e00\u4e2a\u8f93\u5165\u7ef4\u5ea6\u548c\u00a0selu\u00a0\u975e\u7ebf\u6027\u6fc0\u6d3b\u51fd\u6570\uff0c\u518d\u5b9a\u4e49\u4e86\u4e00\u4e2a\u795e\u7ecf\u5143\u540c\u6837\u6fc0\u6d3b\u51fd\u6570\u7684\u00a0Dense\u00a0\u5c42\u4f5c\u4e3a\u8f93\u51fa\u3002\u8fd9\u4e2a\u5b9a\u4e49\u662f\u901a\u8fc7\u5b9e\u9a8c\u53d1\u73b0\u8fc7\u591a\u6216\u8fc7\u5c11\u7684\u795e\u7ecf\u5143\u90fd\u65e0\u6cd5\u63d0\u5347\u8bad\u7ec3\u6548\u679c\uff0c\u592a\u591a\u7684\u5c42\u4e5f\u65e0\u6cd5\u66f4\u597d\u7684\u4f18\u5316\u51c6\u786e\u7387\u3002\u6a21\u578b\u7684\u4f18\u5316\u5668\u5982\u524d\u9762\u7684\u793a\u4f8b\u4e5f\u662f\u00a0adam\uff0c\u635f\u5931\u51fd\u6570\u4e5f\u662f\u5e73\u5747\u65b9\u5dee\u635f\u5931\u51fd\u6570\u00a0mse\u3002\u8bad\u7ec3\u8fc7\u7a0b\uff1a
1\/1 [==============================] – 1s 573ms\/step – loss: 74593.5156
Epoch 2\/2000
1\/1 [==============================] – 0s 38ms\/step – loss: 74222.5469
Epoch 3\/2000
1\/1 [==============================] – 0s 31ms\/step – loss: 74039.7734
…
Epoch 1994\/2000
1\/1 [==============================] – 0s 26ms\/step – loss: 0.3370
Epoch 1995\/2000
1\/1 [==============================] – 0s 25ms\/step – loss: 0.3370
Epoch 1996\/2000
1\/1 [==============================] – 0s 24ms\/step – loss: 0.3370
Epoch 1997\/2000
1\/1 [==============================] – 0s 26ms\/step – loss: 0.3370
Epoch 1998\/2000
1\/1 [==============================] – 0s 24ms\/step – loss: 0.3370
Epoch 1999\/2000
1\/1 [==============================] – 0s 25ms\/step – loss: 0.3369
Epoch 2000\/2000
1\/1 [==============================] – 0s 24ms\/step – loss: 0.3369
\u53ef\u4ee5\u770b\u5230\uff0c\u6574\u4e2a\u795e\u7ecf\u7f51\u7edc\u7684\u8bad\u7ec3\u8fc7\u7a0b\u6301\u7eed\u7a33\u5b9a\u7684\u63d0\u5347\uff0c\u4e00\u76f4\u5230\u8fd1 2000 \u4e2a\u00a0Epoch\u00a0\u7684\u65f6\u5019\u00a0loss\u00a0\u635f\u5931\u8fd8\u5728\u7ee7\u7eed\u4e0b\u964d\uff0c\u53ef\u89c1\u7ee7\u7eed\u8bad\u7ec3\u8fd8\u80fd\u8fdb\u4e00\u6b65\u63d0\u5347\u62df\u5408\u6548\u679c\uff0c\u66f4\u597d\u7684\u903c\u8fd1\u771f\u5b9e\u6570\u636e\u3002\u6309\u7167\u7ba1\u7406\uff0c\u6211\u4eec\u753b\u4e2a\u56fe\u6765\u89c2\u5bdf\u4e00\u4e0b\u6548\u679c\uff1a<\/p>\n
5 \u753b\u56fe<\/h2>\n
plt.scatter(xdata, ydata)
plt.plot(t_x, y_pred, ‘r-‘, lw=5)
plt.show()\u5982\u56fe 8 \u6240\u793a\uff0c\u8bad\u7ec3\u597d\u7684\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u7ed3\u679c\u548c\u771f\u5b9e\u6570\u636e\u5bf9\u6bd4\u51e0\u4e4e\u662f\u4e00\u81f4\u7684\uff0c\u4f46\u662f\uff0c\u5bf9\u6bd4\u4e4b\u524d 4PL \u548c\u591a\u9879\u5f0f\u7684\u51fd\u6570\u62df\u5408\uff0c\u6574\u4f53\u6548\u679c\u5e76\u4e0d\u50cf\u51fd\u6570\u7684\u66f2\u7ebf\u4e00\u6837\u5e73\u6ed1\uff0c\u6574\u4f53\u5448\u73b0\u8fc7\u62df\u5408\u7684\u72b6\u6001\u3002
![\"\u56fe\u7247\"](\"https:\/\/segmentfault.com\/img\/remote\/1460000042328425\" "\\"\u56fe\u7247\\"")
<\/p>\n
print(t_x.reshape(-1, 1)) print(model.predict(t_x.reshape(-1, 1)))<\/p>\n
\u8f93\u51fa\uff1a\u9884\u6d4b\u8f93\u51fa\uff1a<\/h2>\n
[[ 0.]
[ 500.]
[ 1000.]
[ 1500.]
[ 2500.]
[ 6000.]
[ 8000.]
[10000.]
[12000.]]
print(t_y.reshape(-1,1))<\/p>\n
\u8f93\u51fa\uff1a<\/h2>\n
[[ 0. ] [-1.7345996]
[ 90. ] [90.01992]
[160. ] [159.91837]
[210. ] [210.06012]
[260. ] [260.01797]
[347.5] [347.43182]
[357.5] [357.57867]
[367.5] [367.53287]
[377.5]] [377.4857]\u53ef\u4ee5\u770b\u5230\uff0c\u5bf9\u6bd4\u771f\u5b9e\u6570\u636e\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u7684\u7ed3\u679c\u662f\u975e\u5e38\u7cbe\u786e\u7684\uff0c\u8bef\u5dee\u6bd4\u4f7f\u7528\u591a\u9879\u5f0f\u62df\u5408\u7684\u6548\u679c\u597d\u975e\u5e38\u591a\u3002\u00a0\u81f3\u6b64\uff0c\u6211\u4eec\u4f7f\u7528\u4e86 4PL\u3001\u591a\u9879\u5f0f\u4ee5\u53ca\u795e\u7ecf\u7f51\u7edc\u4e09\u79cd\u53ef\u5fae\u7f16\u7a0b\u7684\u65b9\u6cd5\uff0c\u5bf9\u5f39\u6027\u963b\u5c3c\u51fd\u6570\u8fdb\u884c\u62df\u5408\uff0c\u903c\u8fd1\u771f\u5b9e\u6570\u636e\u7684\u6548\u679c\u867d\u7136\u6709\u4e00\u5b9a\u5dee\u5f02\uff0c\u4f46\u603b\u4f53\u90fd\u62df\u5408\u7684\u6bd4\u8f83\u597d\u3002\u540c\u795e\u7ecf\u7f51\u7edc\u7684\u65b9\u6cd5\u4e0d\u540c\uff0c4PL \u548c\u591a\u9879\u5f0f\u62df\u5408\u53ef\u4ee5\u76f4\u63a5\u63d0\u4f9b\u516c\u5f0f\u548c\u53c2\u6570\uff0c\u6211\u4eec\u53ef\u4ee5\u65b9\u4fbf\u7684\u5c06\u5176\u5e94\u7528\u5728\u7a0b\u5e8f\u7684\u7b97\u6cd5\u4e2d\u3002\u800c\u5bf9\u4e8e\u795e\u7ecf\u7f51\u7edc\uff0c\u5219\u9700\u8981\u628a\u795e\u7ecf\u7f51\u7edc\u5728\u6d4f\u89c8\u5668\u91cc\u8dd1\u8d77\u6765\uff0c\u4e0b\u9762\u5c31\u4ecb\u7ecd\u4e00\u4e0b\u8fd9\u4e2a\u8fc7\u7a0b\u3002\u00a0\u6d4f\u89c8\u5668\u6267\u884c\u793a\u4f8b\u00a0\u4e3a\u4e86\u80fd\u591f\u628a\u795e\u7ecf\u7f51\u7edc\u5728\u6d4f\u89c8\u5668\u91cc\u8dd1\u8d77\u6765\uff0c\u5e76\u8ba9 JavaScript \u53ef\u4ee5\u8c03\u7528\uff0c\u9996\u5148\u9700\u8981\u5c06\u795e\u7ecf\u7f51\u7edc\u5b58\u50a8\u6210\u6587\u4ef6\u3002model.save(‘saved_model\/w4model’)\u5b58\u50a8\u5230\u76ee\u5f55\u540e\u4f1a\u5f97\u5230\u4e24\u4e2a\u6587\u4ef6\uff1akeras_metadata.pb\u00a0\u548c\u00a0saved_model.pb\u00a0\u4ee5\u53ca\u4e24\u4e2a\u76ee\u5f55\uff1aassets\u00a0\u548c \u00a0variables\u00a0\u3002\u8fd9\u91cc\u4f7f\u7528\u7684\u662f TensorFlow \u7684\u00a0tf_saved_model\uff0c\u4e0b\u9762\u8f6c\u6362\u6a21\u578b\u7ed9 TensorFlow.js \u5728\u6d4f\u89c8\u5668\u4f7f\u7528\u7684\u65f6\u5019\u4f1a\u7528\u5230\u8fd9\u4e2a\u53c2\u6570\u3002\u4e3a\u4e86\u80fd\u591f\u8ba9\u6a21\u578b\u5728\u6d4f\u89c8\u5668\u8fd0\u884c\uff0c\u5148\u5b89\u88c5\u00a0pip install tensorflowjs\u00a0\u7528\u00a0tensorflowjs_converter\u00a0\u547d\u4ee4\u884c\u8f6c\u6362\u5de5\u5177\u8fdb\u884c\u6a21\u578b\u7684\u8f6c\u6362\uff1atensorflowjs_converter –input_format=tf_saved_model \/
–output_node_names=”w4model” \/
–saved_model_tags=serve .\/saved_model\/w4model .\/web_model\u8fd9\u91cc\u00a0–input_format\u00a0\u53c2\u6570\u00a0tf_saved_model\u00a0\u5bf9\u5e94\u4e4b\u524d\u6a21\u578b\u5b58\u50a8\u4f7f\u7528\u7684\u65b9\u6cd5\uff0c\u8981\u6ce8\u610f\u4e0d\u540c\u65b9\u6cd5\u4fdd\u5b58\u7684\u6587\u4ef6\u683c\u5f0f\u548c\u7ed3\u6784\u7684\u4e0d\u540c\u76f8\u4e92\u4e0d\u517c\u5bb9\u3002–output_node_names\u00a0\u662f\u6a21\u578b\u7684\u540d\u79f0\uff0c–saved_model_tags\u00a0\u4e2d\u00a0tag\u00a0\u662f\u7528\u6765\u533a\u522b\u4e0d\u540c\u7684\u00a0MetaGraphDef\uff0c\u8fd9\u662f\u5728\u52a0\u8f7d\u6a21\u578b\u6240\u9700\u8981\u7684\u53c2\u6570\u5176\u9ed8\u8ba4\u503c\u662f\u00a0serve\u3002\u00a0\u901a\u8fc7\u00a0tensorflowjs_converter\u00a0\u8fdb\u884c\u6a21\u578b\u8f6c\u6362\u540e\uff0c\u6211\u4eec\u5728\u00a0web_model\u00a0\u6587\u4ef6\u5939\u91cc\u4f1a\u770b\u5230\u00a0group1-shard1of1.bin\u00a0\u4ee5\u53ca\u00a0model.json\u00a0\u4e24\u4e2a\u6587\u4ef6\u3002\u8fd9\u4e24\u4e2a\u6587\u4ef6\u4e2d\uff0c\u4ee5\u00a0.json\u00a0\u540e\u7f00\u7ed3\u5c3e\u7684\u6587\u4ef6\u662f\u6a21\u578b\u5b9a\u4e49\u6587\u4ef6\uff0c\u4ee5\u00a0.bin\u00a0\u540e\u7f00\u7ed3\u5c3e\u7684\u6587\u4ef6\u662f\u6a21\u578b\u7684\u6743\u91cd\u6587\u4ef6\u3002\u60a8\u53ef\u4ee5\u628a\u6a21\u578b\u5b9a\u4e49\u6587\u4ef6\u770b\u505a 4PL \u548c\u591a\u9879\u5f0f\u7684\u51fd\u6570\uff0c\u628a\u6a21\u578b\u6743\u91cd\u6587\u4ef6\u770b\u505a\u51fd\u6570\u7684\u53c2\u6570\u3002\u00a0\u901a\u8fc7 npm \u521d\u59cb\u5316\u4e00\u4e2a node.js \u9879\u76ee\uff0c\u7136\u540e\u5728\u00a0package.json\u00a0\u914d\u7f6e\u6587\u4ef6\u91cc\u52a0\u5165\uff1a\u00a0“dependencies”: {<\/p>\n
"@tensorflow\/tfjs": "^3.18.0",\n"@tensorflow\/tfjs-converter": "^3.18.0"<\/code><\/pre>\n},\u5c06\u4f9d\u8d56\u5f15\u5165\u5230\u7a0b\u5e8f\u6587\u4ef6\u4e2d\uff0c\u8fd9\u91cc\u8981\u6ce8\u610f\u00a0loadGraphModel\u00a0\u662f\u6211\u4eec\u4ece\u00a0@tensorflow\/tfjs-converter\u00a0\u00a0\u4f9d\u8d56\u4e2d\u5bfc\u5165\u7684\uff0c\u867d\u7136 \u00a0@tensorflow\/tfjs\u00a0\u63d0\u4f9b\u00a0tf.loadGraphModel()\u00a0\u65b9\u6cd5\u52a0\u8f7d\u6a21\u578b\uff0c\u4f46\u662f\u8fd9\u4e2a\u65b9\u6cd5\u53ea\u9002\u7528\u4e8e TensorFlow.js \u4e2d\u4fdd\u5b58\u7684\u6a21\u578b\uff0c\u6211\u4eec\u901a\u8fc7 Python \u91cc\u00a0model.save()\u00a0\u65b9\u6cd5\u4fdd\u5b58\uff0c\u5e76\u7528 Converter \u8f6c\u6362\u7684\u6a21\u578b\uff0c\u5fc5\u987b\u7528\u00a0tfjs-converter\u00a0\u4f9d\u8d56\u5305\u4e2d\u63d0\u4f9b\u7684\u00a0loadGraphModel\u00a0\u65b9\u6cd5\u8fdb\u884c\u52a0\u8f7d\u3002
import * as tf from “@tensorflow\/tfjs”;
import { loadGraphModel } from “@tensorflow\/tfjs-converter”;
window.onload = async () => {
const resultElement = document.getElementById(“result”);
const MODEL_URL = “model.json”;<\/p>\n
console.time(“Loading of model”);
const model = await loadGraphModel(MODEL_URL);
console.timeEnd(“Loading of model”);<\/p>\n
const test_data = tf.tensor([<\/p>\n
[0.0],\n[500.0],\n[1000.0],\n[1500.0],\n[2500.0],\n[6000.0],\n[8000.0],\n[10000.0],\n[12000.0],<\/code><\/pre>\n]);
tf.print(test_data);
console.time(“Loading of model”);
let outputs = model.execute(test_data);
console.timeEnd(“execute\uff1a”);
tf.print(outputs);
resultElement.innerText = outputs.toString();
};\u8fd9\u91cc\u9700\u8981\u6ce8\u610f\u7684\u662f\uff0c\u7531\u4e8e\u6211\u4eec\u7684\u6a21\u578b\u5728\u9884\u6d4b\u65f6\u4f7f\u7528\u7684\u662f\u5f20\u91cf\u4f5c\u4e3a\u8f93\u5165\uff0c\u56e0\u6b64\uff0c\u9700\u8981\u7528\u00a0tf.tensor()\u00a0\u65b9\u6cd5\u6765\u8fd4\u56de\u7ecf\u8fc7\u5305\u88c5\u7684\u5f20\u91cf\u4f5c\u4e3a\u6a21\u578b\u7684\u8f93\u5165\u3002\u8fd0\u884c\u7a0b\u5e8f\u540e\uff0c\u6211\u4eec\u53ef\u4ee5\u4ece\u6d4f\u89c8\u5668\u5f00\u53d1\u8005\u5de5\u5177\u7684\u63a7\u5236\u53f0\u770b\u5230\u6253\u5370\u7684\u8c03\u8bd5\u4fe1\u606f\uff1a
[Violation] ‘load’ handler took 340ms
index.js:12 Loading of model: 67.19482421875 ms
print.ts:34 Tensor<\/p>\n
[[0 ],\n [500 ],\n [1000 ],\n [1500 ],\n [2500 ],\n [6000 ],\n [8000 ],\n [10000],\n [12000]]<\/code><\/pre>\nindex.js:28 execute: 257.47607421875 ms
print.ts:34 Tensor<\/p>\n
[[-1.7345995 ],\n [90.0198822 ],\n [159.9183655],\n [210.0600586],\n [260.0179443],\n [347.4320068],\n [357.5788269],\n [367.5332947],\n [377.4856262]]\u6211\u4eec\u4f1a\u770b\u5230\u63a7\u5236\u53f0\u8f93\u51fa\u00a0webgl\u00a0\u4ee3\u8868 TensorFlow.js \u6210\u529f\u542f\u7528\u4e86 WebGL \u7684\u52a0\u901f\u80fd\u529b\uff0c\u5728\u5176\u52a0\u901f\u4e4b\u4e0b\uff0c\u6211\u4eec\u5728\u9884\u6d4b\u7684\u65f6\u5019\u901f\u5ea6\u4f1a\u4ece 257ms \u5927\u5e45\u5ea6\u63d0\u5347\u5230 131ms\uff0c\u591a\u6b21\u9884\u6d4b\u65f6\u95f4\u7531\u4e8e\u6743\u91cd\u548c\u8ba1\u7b97\u56fe\u5df2\u7ecf\u52a0\u8f7d\u5230\u663e\u5b58\u4e2d\u901f\u5ea6\u4f1a\u66f4\u5feb\uff0c\u8fbe\u5230 78ms \u5de6\u53f3\u3002\u5f53\u7136\uff0c\u5bf9\u4e8e\u52a8\u753b\u6765\u8bf4\u8fd9\u79cd\u8017\u65f6\u8fd8\u662f\u6709\u70b9\u591a\uff0c\u60a8\u53ef\u4ee5\u56de\u987e\u672c\u4e66\u7aef\u667a\u80fd\u7684\u90e8\u5206\uff0c\u4f7f\u7528\u76f8\u5173\u7684\u5de5\u5177\u548c\u65b9\u6cd5\u5bf9\u6a21\u578b\u8fdb\u884c\u538b\u7f29\u548c\u4f18\u5316\uff0c\u53ea\u9700\u8981\u5728\u6a21\u578b\u8f93\u51fa\u7684\u65f6\u5019\u9009\u62e9 TensorFlow.js\uff0c\u6216\u8005\u7528 TensorFlow \u63d0\u4f9b\u7684\u5de5\u5177\u628a\u4f18\u5316\u540e\u7684\u6a21\u578b\u8f6c\u6210 TensorFlow.js \u7684\u6a21\u578b\u5373\u53ef\u590d\u7528\u8fd9\u4e2a\u793a\u4f8b\u4e2d\u7684\u4ee3\u7801\u3002\u00a0\u6700\u540e\uff0c\u8ba9\u6211\u4eec\u518d\u6b21\u56de\u5230\u5f39\u6027\u963b\u5c3c\u51fd\u6570\u5728 JavaScript \u91cc\u7684\u5b9a\u4e49\uff1a<\/code><\/pre>\ndamping(x, max) {<\/p>\n
let y = Math.abs(x);\n y = 0.821 * max \/ (1 + 4527.779 \/ Math.pow(y, 1.153));\n return Math.round(x < 0 ? -y : y);\n },\u628a\u8fd9\u91cc\u7684\u00a0y\u00a0\u53d8\u6210\u6a21\u578b\u9884\u6d4b\u7684\u7ed3\u679c\uff1ay = model.execute(tf.tensor([[x],]));\u00a0\u5373\u53ef\u7528\u6a21\u578b\u6765\u62df\u5408\u5f39\u6027\u963b\u5c3c\u51fd\u6570\u5e76\u8fd4\u56de\u8ba1\u7b97\u7ed3\u679c\u3002\u60a8\u53ef\u80fd\u4f1a\u95ee\uff1a\u65e2\u7136\u6211\u6709 4PL\u3001\u591a\u9879\u5f0f\u62df\u5408\u7b49\u65b9\u6cd5\uff0c\u4e3a\u4ec0\u4e48\u8981\u7528\u795e\u7ecf\u7f51\u7edc\u5462\uff1f\u662f\u7684\uff0c\u5982\u679c\u60a8\u8981\u89e3\u51b3\u7684\u95ee\u9898\u7b80\u5355\u5230\u8db3\u4ee5\u7528\u51fd\u6570\u62df\u5408\u7684\u65b9\u5f0f\u89e3\u51b3\uff0c\u540c\u65f6\uff0c\u60a8\u4e5f\u77e5\u9053\u5e94\u8be5\u7528 4PL \u8fd8\u662f\u591a\u9879\u5f0f\u5f02\u6216\u662f\u5176\u5b83\u51fd\u6570\u8fdb\u884c\u62df\u5408\u6765\u903c\u8fd1\u771f\u5b9e\u60c5\u51b5\u3002\u6ee1\u8db3\u8fd9\u4e24\u4e2a\u524d\u63d0\u6761\u4ef6\uff0c\u5f53\u7136\u53ef\u4ee5\u7528\u51fd\u6570\u62df\u5408\u7684\u65b9\u5f0f\u76f4\u63a5\u627e\u5230\u5bf9\u5e94\u7684\u51fd\u6570\u548c\u53c2\u6570\u6765\u89e3\u51b3\u95ee\u9898\u3002\u4f46\u662f\uff0c\u73b0\u5b9e\u4e2d\u6211\u4eec\u8981\u89e3\u51b3\u7684\u95ee\u9898\u5f80\u5f80\u4f1a\u66f4\u52a0\u7684\u590d\u6742\uff0c\u6bd4\u5982\u6211\u4eec\u60f3\u505a\u4e2a\u8bed\u97f3\u8f93\u5165\u4ea4\u4e92\u80fd\u529b\uff0c\u5c31\u9700\u8981\u5bf9\u8bed\u97f3\u8fdb\u884c\u97f3\u9891\u4fe1\u53f7\u89e3\u6790\u3001\u5206\u6790\u3001\u5904\u7406\u3001\u8bed\u4e49\u8bc6\u522b\u3001\u8f6c\u6587\u672c\u7b49\u7b49\u8fc7\u7a0b\uff0c\u6bcf\u4e2a\u73af\u8282\u4e2d\u8981\u89e3\u51b3\u7684\u95ee\u9898\u90fd\u65e0\u6cd5\u7b80\u5355\u7528\u4e00\u4e2a\u51fd\u6570\u548c\u5355\u53d8\u91cf\u591a\u53c2\u6570\u65b9\u5f0f\u53bb\u63cf\u8ff0\u548c\u62df\u5408\u53d8\u91cf\u95f4\u7684\u5173\u7cfb\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u53ef\u4ee5\u628a\u590d\u6742\u7684\u95ee\u9898\u7528\u795e\u7ecf\u7f51\u7edc\u53bb\u62df\u5408\uff0c\u8ba9\u795e\u7ecf\u7f51\u7edc\u901a\u8fc7\u795e\u7ecf\u5143\u3001Edge\u3001\u6743\u91cd\u3001\u53c2\u6570\u7b49\u4ee5\u81ea\u52a8\u5fae\u5206\u7684\u80fd\u529b\u53bb\u5bfb\u627e\u7b54\u6848\u3002\u6700\u540e\uff0c\u9700\u8981\u6307\u51fa\u7684\u662f\u62df\u5408\u6bd5\u7adf\u662f\u4e00\u79cd\u8fd1\u4f3c\uff0c\u65e0\u6cd5\u63ed\u793a\u95ee\u9898\u672c\u8d28\u662f\u4ec0\u4e48\uff1f\u8fd9\u5c31\u662f\u5927\u5bb6\u7ecf\u5e38\u8d28\u7591\u673a\u5668\u5b66\u4e60\u53ef\u89e3\u91ca\u6027\u7684\u6839\u6e90\u6240\u5728\u3002\u8bd5\u60f3\u4e00\u4e0b\uff0c\u5982\u679c\u6211\u4eec\u7a0b\u5e8f\u7684\u5f39\u6027\u963b\u5c3c\u7b97\u6cd5\u662f\u6839\u636e\u7269\u7406\u5b66\u5b9a\u4e49\u7684\u516c\u5f0f\u53bb\u7f16\u5199\u7684\uff0c\u90a3\u8fd9\u4e2a\u7b97\u6cd5\u5c31\u5177\u5907\u4e86\u53ef\u89e3\u91ca\u6027\u3002\u5177\u5907\u4e86\u53ef\u89e3\u91ca\u6027\uff0c\u6211\u5c31\u80fd\u591f\u63ed\u793a\u6bcf\u4e2a\u53d8\u91cf\u7684\u610f\u4e49\u4ee5\u53ca\u53d8\u91cf\u4e4b\u95f4\u7684\u5173\u7cfb\uff0c\u4ece\u800c\u4f7f\u7a0b\u5e8f\u548c\u7b97\u6cd5\u5177\u5907\u66f4\u7cbe\u786e\u3001\u7ec6\u81f4\u7684\u5904\u7406\u80fd\u529b\u3002\u800c\u51fd\u6570\u62df\u5408\u6216\u795e\u7ecf\u7f51\u7edc\u62df\u5408\u7684\u53ef\u5fae\u7f16\u7a0b\u4e2d\uff0c\u673a\u5668\u5b66\u4e60\u5230\u4e86\u4ec0\u4e48\uff1f\u6211\u4eec\u5f88\u96be\u7ed9\u51fa\u7b54\u6848\uff0c\u53ea\u80fd\u901a\u8fc7\u53cd\u590d\u5b9e\u9a8c\u6765\u63a8\u5b9a\u51fd\u6570\u6216\u795e\u7ecf\u7f51\u7edc\u80fd\u591f\u6839\u636e\u8f93\u5165\u4f30\u8ba1\u51fa\u6b63\u786e\u7684\u8f93\u51fa\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u4e0d\u80fd\u9047\u5230\u95ee\u9898\u5c31\u4e00\u80a1\u8111\u7684\u4f7f\u7528\u53ef\u5fae\u7f16\u7a0b\u53bb\u89e3\u51b3\uff0c\u5728\u89e3\u91ca\u6027\u65b9\u9762\u6709\u5f3a\u8bc9\u6c42\u7684\u573a\u666f\u4e2d\uff0c\u5c3d\u91cf\u8fd8\u662f\u8981\u4ece\u95ee\u9898\u672c\u8eab\u51fa\u53d1\u627e\u51fa\u89e3\u51b3\u7684\u9014\u5f84\u3002\u00a0\u539f\u6587\u94fe\u63a5\uff1ahttps:\/\/click.aliyun.com\/m\/1000353593\/\u672c\u6587\u4e3a\u963f\u91cc\u4e91\u539f\u521b\u5185\u5bb9\uff0c\u672a\u7ecf\u5141\u8bb8\u4e0d\u5f97\u8f6c\u8f7d\u3002\n\n<\/code><\/pre>\nhttps:\/\/segmentfault.com\/a\/1190000042328403<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"\u7b80\u4ecb\uff1a\u00a0\u7814\u7a76\u8005\u8868\u793a\uff0c\u7531\u4e8e\u5e7f\u6cdb\u7684\u79d1\u5b66\u8ba1\u7b97\u548c\u673a\u5668\u5b66\u4e60\u9886\u57df\u5728\u5e95\u5c42\u7ed3\u6784\u4e0a\u90fd\u9700\u8981\u7ebf\u6027\u4ee3\u6570\u7684\u652f\u6301\uff0c\u56e0\u6b64\u6709\u53ef\u80fd\u4ee5\u53ef\u5fae\u7f16\u7a0b\u7684\u5f62 […]<\/p>\n","protected":false},"author":0,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[],"tags":[],"class_list":["post-10006223","post","type-post","status-publish","format-standard","hentry"],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10006223","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10006223"}],"version-history":[{"count":0,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10006223\/revisions"}],"wp:attachment":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10006223"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10006223"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10006223"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}