{"id":1160,"date":"2017-07-21T12:02:00","date_gmt":"2017-07-21T03:02:00","guid":{"rendered":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/?p=1160"},"modified":"2019-03-12T14:02:59","modified_gmt":"2019-03-12T05:02:59","slug":"keras%e3%81%a7%e5%8c%96%e5%90%88%e7%89%a9%e3%81%ae%e6%ba%b6%e8%a7%a3%e5%ba%a6%e4%ba%88%e6%b8%ac%ef%bc%88%e3%83%8b%e3%83%a5%e3%83%bc%e3%83%a9%e3%83%ab%e3%83%8d%e3%83%83%e3%83%88%e3%83%af%e3%83%bc","status":"publish","type":"post","link":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/2017\/07\/21\/keras%e3%81%a7%e5%8c%96%e5%90%88%e7%89%a9%e3%81%ae%e6%ba%b6%e8%a7%a3%e5%ba%a6%e4%ba%88%e6%b8%ac%ef%bc%88%e3%83%8b%e3%83%a5%e3%83%bc%e3%83%a9%e3%83%ab%e3%83%8d%e3%83%83%e3%83%88%e3%83%af%e3%83%bc\/","title":{"rendered":"Keras\u3067\u5316\u5408\u7269\u306e\u6eb6\u89e3\u5ea6\u4e88\u6e2c\uff08\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\uff0c\u56de\u5e30\u5206\u6790\uff09"},"content":{"rendered":"<p>\u74b0\u5883: macOS Sierra 10.12.5, CPU: 3.3 GHz Intel Core i5, \u30e1\u30e2\u30ea\uff1a 8 GB. Python 3.6.1, Keras 2.0.6, TensorFlow 1.3.0rc0, RDKit 2017.03.3, mordred 0.4.0.post1.dev1.<br \/>\n\u53c2\u8003\u30b5\u30a4\u30c8:<\/p>\n<ul>\n<li><a href=\"http:\/\/www.wildcardconsulting.dk\/useful-information\/molecular-neural-network-models-with-rdkit-and-keras-in-python\/\">Wash that gold- Modelling solubility with Molecular fingerprints by Esben Jannik Bjerrum<\/a><\/li>\n<li><a href=\"http:\/\/cheminformist.itmol.com\/TEST\/?p=1583\">Chainer: \u30af\u30e9\u30b9\u5206\u985e\u306b\u3088\u308b\u6eb6\u89e3\u5ea6\u306e\u4e88\u6e2c &#8211; Cheminformist3<\/a><\/li>\n<\/ul>\n<p>\u4eba\u5de5\u77e5\u80fd\u304c\u8a71\u984c\u306a\u306e\u3067\uff0c\u898b\u69d8\u898b\u771f\u4f3c\u3067\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u3092\u4f7f\u3063\u305f\u5316\u5408\u7269\u306e\u7269\u6027\u4e88\u6e2c\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\u3092\u3084\u3063\u3066\u307f\u307e\u3057\u305f\uff0e\u6a5f\u68b0\u5b66\u7fd2\u306e\u30e9\u30a4\u30d6\u30e9\u30ea\u306fGoogle\u304c\u958b\u767a\u30fb\u4f7f\u7528\u3057\u3066\u3044\u308b<a href=\"https:\/\/www.tensorflow.org\/\">TensorFlow<\/a>\u3067\u3059\uff0eKeras\u306fTensorFlow\u306e\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306e\u6a5f\u80fd\u3092\u4f7f\u3044\u6613\u304f\u3057\u3066\u304f\u308c\u308b\u30d1\u30c3\u30b1\u30fc\u30b8\u3067\u3059\uff08Theano\u3068\u3044\u3046\u30e9\u30a4\u30d6\u30e9\u30ea\u3082\u4f7f\u3048\u308b\uff09\uff0e<\/p>\n<h2>\u6e96\u50991<\/h2>\n<p>pyenv + pyenv-virtualenv\u3067\u4eee\u60f3\u74b0\u5883\u3092\u69cb\u7bc9\u3057\u307e\u3059 (\u53c2\u8003\u30b5\u30a4\u30c8: <a href=\"http:\/\/qiita.com\/shizuma\/items\/027167c6257f1c9d2a6f\">Python\u306e\u74b0\u5883\u69cb\u7bc9 on Mac ( pyenv, virtualenv, anaconda, ipython notebook )<\/a>)\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">$ brew install gcc open-babel\n$ brew install pyenv\n$ brew install pyenv-virtualenv\n<\/code><\/pre>\n<p>~\/.bash_profile\u306b\u4ee5\u4e0b\u306e\u5185\u5bb9\u3092\u8ffd\u52a0\uff0e<\/p>\n<pre class=\"brush: bash; title: ; notranslate\" title=\"\">\nPYENV_ROOT=~\/.pyenv\nexport PATH=$PATH:$PYENV_ROOT\/bin\neval &quot;$(pyenv init -)&quot;\neval &quot;$(pyenv virtualenv-init -)&quot;\n<\/pre>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">$ source ~\/.bash_profile\n<\/code><\/pre>\n<p>\u6700\u65b0\u306e\u30d1\u30c3\u30b1\u30fc\u30b8\u30de\u30cd\u30fc\u30b8\u30e3\u30fc\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\uff0eNumpy\u306a\u3069\u79d1\u5b66\u8a08\u7b97\u306b\u5fc5\u8981\u306a\u30d1\u30c3\u30b1\u30fc\u30b8\u306f\u5927\u62b5\u30c7\u30d5\u30a9\u30eb\u30c8\u3067\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u307e\u3059\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">$ pyenv install --list\n$ pyenv install anaconda3-4.4.0\n$ pyenv global anaconda3-4.4.0\n(anaconda3-4.4.0)$ conda update conda\n<\/code><\/pre>\n<p>\u5206\u5b50\u8a18\u8ff0\u5b50\u3092\u6271\u3046\u305f\u3081\u306eRDKit\u3068<a href=\"https:\/\/www.jstage.jst.go.jp\/article\/ciqs\/2016\/0\/2016_Y4\/_article\/-char\/ja\/\">mordred<\/a>\u3092\u5c0e\u5165\u3057\u307e\u3059 (\u53c2\u8003\u30b5\u30a4\u30c8: <a href=\"https:\/\/github.com\/mordred-descriptor\">mordred-descriptor<\/a>)\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ conda install -c rdkit -c mordred-descriptor mordred\n<\/code><\/pre>\n<p>rdkit\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306f2017.03.3\uff0cmordred\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306f0.4.0.post1.dev1\u3067\u3057\u305f\uff0e<br \/>\n<!--\n\u9069\u5f53\u306a\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u79fb\u52d5\u3057\u305f\u5f8c\uff0c\u4f5c\u696d\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3092\u4f5c\u6210\uff0e\u4eee\u60f3\u74b0\u5883\u306e\u69cb\u7bc9\u30fb\u8d77\u52d5\uff0e\n\n\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">$ mkdir PythonTest\n$ virtualenv --no-site-packages PythonTest\n$ cd PythonTest\n$ source bin\/activate\n<\/code><\/pre>\n\n\n--><\/p>\n<h2>\u6e96\u50992<\/h2>\n<p>\u6a5f\u68b0\u5b66\u7fd2\u306b\u5fc5\u8981\u306a\u30d1\u30c3\u30b1\u30fc\u30b8\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ pip install keras\n<\/code><\/pre>\n<p>TensorFlow\u306f<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ pip install tensorflow<\/code><\/pre>\n<p>\u3067\u30b3\u30f3\u30d1\u30a4\u30eb\u6e08\u306e\u3082\u306e\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3067\u304d\u307e\u3059\uff0e\u3053\u308c\u306fCPU\u62e1\u5f35\u547d\u4ee4\uff08AVX2\u306a\u3069\uff09\u3092\u4f7f\u7528\u3067\u304d\u306a\u3044\u306e\u3067\uff0c\u30bd\u30fc\u30b9\u304b\u3089\u30b3\u30f3\u30d1\u30a4\u30eb\u3057\u3066\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u65b9\u6cd5\u3092\u4e0b\u306b\u8a18\u9332\u3057\u3066\u304a\u304d\u307e\u3059\uff08\u53c2\u8003\u30b5\u30a4\u30c8: <a href=\"https:\/\/www.tensorflow.org\/install\/install_sources#build_the_pip_package\">Installing TensorFlow from Sources &#8211; TensorFlow<\/a>\uff09\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ brew install bazel\n(anaconda3-4.4.0)$ git clone https:\/\/github.com\/tensorflow\/tensorflow \n(anaconda3-4.4.0)$ cd tensorflow\n(anaconda3-4.4.0)$ git checkout r1.2\n(anaconda3-4.4.0)$ .\/configure\n#Return\u3092\u9023\u6253\n(anaconda3-4.4.0)$ bazel build --config=opt \/\/tensorflow\/tools\/pip_package:build_pip_package\n#\u6642\u9593\u304c\u304b\u304b\u308b (\u7d0425\u5206)\n(anaconda3-4.4.0)$ bazel-bin\/tensorflow\/tools\/pip_package\/build_pip_package \/tmp\/tensorflow_pkg\n(anaconda3-4.4.0)$ pip install \/tmp\/tensorflow_pkg\/tensorflow-1.3.0rc0-cp36-cp36m-macosx_10_7_x86_64.whl\n<\/code><\/pre>\n<p>TensorFlow\u306e\u52d5\u4f5c\u78ba\u8a8d\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ ipython\nimport tensorflow as tf\nhello = tf.constant(&#039;Hello, TensorFlow!&#039;)\nsess = tf.Session()\nprint(sess.run(hello))\n<\/code><\/pre>\n<p>&#8220;b&#8217;Hello, TensorFlow!'&#8221;\u3068\u8868\u793a\u3055\u308c\u308c\u3070OK\uff0e<\/p>\n<p>jupyter-notebook\u3092\u8d77\u52d5\u3057\u307e\u3059 (\u30d6\u30e9\u30a6\u30b6\u304c\u958b\u304f)\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ jupyter-notebook\n<\/code><\/pre>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-medium wp-image-1167\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-300x82.png\" alt=\"\" width=\"600\" height=\"164\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-300x82.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-768x209.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-1024x279.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-360x98.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png 1652w\" sizes=\"auto, (max-width: 600px) 100vw, 600px\" \/><br \/>\nNew\u25bc \u2192 Notebook: Python 3 \u3067\u65b0\u3057\u3044\u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3092\u4f5c\u6210\u3057\u307e\u3059\uff0eShift + return\u3067\u30b3\u30de\u30f3\u30c9\u304c\u5b9f\u884c\u3067\u304d\u307e\u3059\uff0e\u7d42\u4e86\u306fFile \u2192 Close and halt\u3067\u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3092\u9589\u3058\u305f\u5f8c\uff0cterminal\u3067control + C\u3067\u30b5\u30fc\u30d0\u30fc\u3092shutdown\u3057\u307e\u3059\uff0e<\/p>\n<h2>Neural network model with mordred and Keras<\/h2>\n<p><a href=\"http:\/\/www.wildcardconsulting.dk\/useful-information\/molecular-neural-network-models-with-rdkit-and-keras-in-python\/\">Molecular neural network models with RDKit and Keras in Python<\/a>\u3092\u53c2\u8003\u306b\u5316\u5408\u7269\u306e\u6eb6\u89e3\u5ea6\u4e88\u6e2c\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3057\u3066\u307f\u307e\u3059\uff0e<a href=\"https:\/\/github.com\/rdkit\/rdkit-orig\/tree\/master\/Docs\/Book\/data\">RDKit\u306e\u30ec\u30dd\u30b8\u30c8\u30ea<\/a>\u304b\u3089\uff0c\u5206\u5b50\u306e\u69cb\u9020\u3068\u6eb6\u89e3\u5ea6\u306e\u30c7\u30fc\u30bf\u304c\u542b\u307e\u308c\u308bsolubility.train.sdf\uff081025\u5206\u5b50\uff09\u3068solubility.test.sdf\uff08257\u5206\u5b50\uff09\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3057\u307e\u3059\uff0e\u5f8c\u3005\u9762\u5012\u306a\u306e\u3067\uff0c\u3053\u306e2\u3064\u306e\u30d5\u30a1\u30a4\u30eb\u30921\u3064\u306e\u30d5\u30a1\u30a4\u30eb\u306b\u3057\u307e\u3059\uff0e<\/p>\n<pre class=\"command-line\"><code class=\"language-bash\" data-line=\"\">(anaconda3-4.4.0)$ obabel solubility.train.sdf solubility.test.sdf -O solubility.sdf\n<\/code><\/pre>\n<p>\u4ee5\u4e0b\u306e\u4f5c\u696d\u306fjupyter-notebook\u3067\u884c\u3044\u307e\u3059\uff0e\u65b0\u898f\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u305f\u3089\uff0c\u4ee5\u4e0b\u306e\u5185\u5bb9\u3092\u30b3\u30d4\u30da\u3057\u3066\uff0cshift-return\u3067\u5b9f\u884c\u3067\u304d\u307e\u3059\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#\u30e2\u30b8\u30e5\u30fc\u30eb\u306e\u8aad\u307f\u8fbc\u307f\nfrom rdkit import Chem\nfrom rdkit.Chem.Draw import IPythonConsole\nfrom mordred import descriptors, Calculator\nimport numpy as np\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn import model_selection\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Activation\nfrom keras.optimizers import SGD\n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\ncalc = Calculator(descriptors, ignore_3D = True)\n#\u8aad\u307f\u8fbc\u3080\u30d5\u30a1\u30a4\u30eb\nsdf = &#x5B; mol for mol in Chem.SDMolSupplier('solubility.sdf')]\n#mordred\u3092\u4f7f\u3063\u3066sdf\u30d5\u30a1\u30a4\u30eb\u4e2d\u306e\u5206\u5b50\u306e\u5316\u5b66\u8a18\u8ff0\u5b50 (\u539f\u5b50\u6570\u3084\u5206\u5b50\u91cf\u306a\u3069\u5206\u5b50\u3092\u8868\u73fe\u3059\u308b\u6570\u5024) \u3092\u8a08\u7b97\uff0e\nX = calc.pandas(sdf).astype('float').dropna(axis = 1)\n<\/pre>\n<p>1282\u5206\u5b50\u3092\u51e6\u7406\u3059\u308b\u306e\u306b33\u79d2\u304b\u304b\u308a\u307e\u3057\u305f\uff0emordred\u304c\u8a08\u7b97\u3067\u304d\u308b\u8a18\u8ff0\u5b50\u306f2\u6b21\u5143\u8a18\u8ff0\u5b50\u304c1610\u7a2e\u985e, 3\u6b21\u5143\u8a18\u8ff0\u5b50\u304c214\u7a2e\u985e\u3067\u3059\uff0e\u4eca\u56de\u306f2\u6b21\u5143\u306e\u307f\u8a08\u7b97\u3057\uff0c\u5024\u304c\u8a08\u7b97\u3067\u304d\u306a\u304b\u3063\u305f\u8a18\u8ff0\u5b50\u3092dropna(axis = 1) \u3067\u524a\u3063\u3066\u3044\u308b\u306e\u3067\uff0c\u5f97\u3089\u308c\u305f\u8a18\u8ff0\u5b50\u306f1114\u7a2e\u985e\u3067\u3057\u305f\uff0e<\/p>\n<p>\u8a66\u3057\u306b<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nX\n<\/pre>\n<p>\u3068\u6253\u3063\u3066\uff0cshift + return\u3059\u308b\u3068\u4e2d\u8eab\u304c\u8868\u793a\u3055\u308c\u307e\u3059\uff0e\u5206\u5b50\u306e\u6570\u304c1282\uff0c\u5909\u6570 (\u8a18\u8ff0\u5b50) \u306e\u6570\u304c1114\u3067\u3059<br \/>\n<img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-large wp-image-1195\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee-1024x326.png\" alt=\"\" width=\"640\" height=\"204\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee-1024x326.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee-300x96.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee-768x245.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee-360x115.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/7b99f59e642fb0828b0fee79f2540fee.png 1388w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#Numpy\u5f62\u5f0f\u306e\u914d\u5217\u306b\u5909\u63db\nX = np.array(X, dtype = np.float32)\n#\u5404\u8a18\u8ff0\u5b50\u306b\u3064\u3044\u3066\u5e73\u57470, \u5206\u65631\u306b\u5909\u63db\nst = StandardScaler()\nX= st.fit_transform(X)\n#\u5f8c\u3067\u518d\u5229\u7528\u3059\u308b\u305f\u3081\u306b\u30d5\u30a1\u30a4\u30eb\u306b\u4fdd\u5b58\nnp.save(&quot;X_2d.npy&quot;, X)\n<\/pre>\n<p>\u6eb6\u89e3\u5ea6\u3092sdf\u30d5\u30a1\u30a4\u30eb\u304b\u3089\u8aad\u307f\u51fa\u3059 (\u53c2\u8003: <a href=\"https:\/\/iwatobipen.wordpress.com\/2016\/12\/13\/build-regression-model-in-keras\/\">Build regression model in Keras<\/a>)\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#\u6eb6\u89e3\u5ea6\u3092\u8aad\u307f\u51fa\u3059\u95a2\u6570\u3092\u5b9a\u7fa9\ndef getResponse( mols, prop= &quot;SOL&quot; ):\nY = &#x5B;]\nfor mol in mols:\nact = mol.GetProp( prop )\nY.append( act )\nreturn Y\n#\u6eb6\u89e3\u5ea6\u3092sdf\u30d5\u30a1\u30a4\u30eb\u304b\u3089\u8aad\u307f\u8fbc\u3080\nY = getResponse(sdf)\n#Numpy\u5f62\u5f0f\u306e\u914d\u5217\u306b\u5909\u63db\nY = np.array(Y, dtype = np.float32)\n#\u5f8c\u3067\u518d\u5229\u7528\u3059\u308b\u305f\u3081\u306b\u30d5\u30a1\u30a4\u30eb\u306b\u4fdd\u5b58\nnp.save(&quot;Y_2d.npy&quot;, Y)\n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#\u8a13\u7df4\u30bb\u30c3\u30c8\u3068\u30c6\u30b9\u30c8\u30bb\u30c3\u30c8\u306b\u518d\u5206\u5272\uff08\u30e9\u30f3\u30c0\u30e0\uff09\uff0e\nX_train, X_test, y_train, y_test = model_selection.train_test_split(X,\nY, test_size=0.25, random_state=42)\nnp.save(&quot;X_train.npy&quot;, X_train)\nnp.save(&quot;X_test.npy&quot;, X_test)\nnp.save(&quot;y_train.npy&quot;, y_train)\nnp.save(&quot;y_test.npy&quot;, y_test)\n<\/pre>\n<p>\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u5c64\u3092\u4e0a\u304b\u3089\u9806\u756a\u306b\u8a18\u8ff0\u3057\u307e\u3059\uff0e\u4e0b\u306e\u5834\u5408\u306f\u5358\u5c64\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306b\u306a\u308a\u307e\u3059\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nmodel = Sequential()\n#\u5165\u529b\u5c64\uff0eDense\u306f\u5168\u7d50\u5408\u5c64\u306e\u610f\u5473\uff0e\u6b21\u306e\u5c64\u306b\u6e21\u3055\u308c\u308b\u6b21\u5143\u306f50\uff0e\u5165\u529b\u30c7\u30fc\u30bf\u306e\u6b21\u5143\uff08input_dim\uff09\u306f1114\uff0e\nmodel.add(Dense(units = 50, input_dim = X.shape&#x5B;1]))\nmodel.add(Activation(&quot;sigmoid&quot;))\n#\u51fa\u529b\u5c64\uff0e\u6b21\u51431\uff0c\u3064\u307e\u308a\u4e00\u3064\u306e\u5024\u3092\u51fa\u529b\u3059\u308b\uff0e\nmodel.add(Dense(units = 1))\n\nmodel.summary()\n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#SGD\u306fStochastic Gradient Descent (\u78ba\u7387\u7684\u52fe\u914d\u964d\u4e0b\u6cd5)\uff0e\u5c40\u6240\u7684\u6700\u5c0f\u5024\u306b\u3068\u3069\u307e\u3089\u306a\u3044\u3088\u3046\u306b\u3059\u308b\u65b9\u6cd5\u3089\u3057\u3044\uff0enesterov\u306fNesterov\u306e\u52a0\u901f\u52fe\u914d\u964d\u4e0b\u6cd5\uff0e\nmodel.compile(loss = 'mean_squared_error',\noptimizer = SGD(lr = 0.01, momentum = 0.9, nesterov = True),\nmetrics=&#x5B;'accuracy'])\nhistory = model.fit(X_train, y_train, epochs = 100, batch_size = 32,\nvalidation_data = (X_test, y_test))\nscore = model.evaluate(X_test, y_test, verbose = 0)\nprint('Test loss:', score&#x5B;0])\nprint('Test accuracy:', score&#x5B;1])\ny_pred = model.predict(X_test)\nrms = (np.mean((y_test - y_pred) ** 2)) ** 0.5\n#s = np.std(y_test - y_pred)\nprint(&quot;Neural Network RMS&quot;, rms)\n<\/pre>\n<p>\u8a08\u7b97\u306f5\u79d2\u5f31\u3067\u7d42\u4e86\u3057\u307e\u3057\u305f\uff0e<\/p>\n<pre class=\"prettyprint lang-python\">Epoch 499&#47;500\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.1511 - acc: 0.0166 - val_loss: 0.7379 - val_acc: 0.0125\nEpoch 500&#47;500\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.1454 - acc: 0.0166 - val_loss: 0.7537 - val_acc: 0.0125   \nTest loss: 0.428150146735\nTest accuracy: 0.0155763239875\nNeural Network RMS 2.81453055091\n<\/pre>\n<p>\u7d50\u679c\u306e\u53ef\u8996\u5316<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nimport matplotlib.pyplot as plt\nplt.figure()\nplt.scatter(y_train, model.predict(X_train), label = 'Train', c = 'blue')\nplt.title('Neural Network Predictor')\nplt.xlabel('Measured Solubility')\nplt.ylabel('Predicted Solubility')\nplt.scatter(y_test, model.predict(X_test), c = 'lightgreen', label = 'Test', alpha = 0.8)\nplt.legend(loc = 4)\nplt.show()\n<\/pre>\n<p><img decoding=\"async\" class=\"aligncenter size-large wp-image-1230\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4-1024x768.png\" alt=\"\" width=\"480\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4-1024x768.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4-300x225.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4-768x576.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4-360x270.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-4.png 1280w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/p>\n<p>\u8a13\u7df4\u30bb\u30c3\u30c8\u306f\u3074\u3063\u305f\u308a\u56de\u5e30\u3055\u308c\u3066\u3044\u307e\u3059\u304c\u30c6\u30b9\u30c8\u30bb\u30c3\u30c8\u306f\u3070\u3089\u3051\u3066\u3044\u308b\u306e\u3067\uff0c\u904e\u5b66\u7fd2\u306b\u9665\u3044\u3063\u3066\u3044\u307e\u3059\uff0e\u3053\u306e\u7a0b\u5ea6\u306e\u8a08\u7b97\u3067100 epoch\u306f\u56de\u3057\u904e\u304e\u306e\u3088\u3046\u3067\u3059\uff0e\u8a13\u7df4\u30bb\u30c3\u30c8\u306eloss\u5024\u3068\u30c6\u30b9\u30c8\u30bb\u30c3\u30c8\u306eloss\u5024\u3092\u30d7\u30ed\u30c3\u30c8\u3057\u3066\u307f\u307e\u3059\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nimport matplotlib.pyplot as plt\nloss = history.history&#x5B;'loss']\nval_loss = history.history&#x5B;'val_loss']\nepochs = len(loss)\nplt.plot(range(epochs), loss, marker = '.', label = 'loss')\nplt.plot(range(epochs), val_loss, marker = '.', label = 'val_loss')\nplt.legend(loc = 'best')\nplt.grid()\nplt.xlabel('epoch')\nplt.ylabel('loss')\nplt.show()\n<\/pre>\n<p><img decoding=\"async\" class=\"aligncenter size-large wp-image-1229\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3-1024x768.png\" alt=\"\" width=\"480\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3-1024x768.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3-300x225.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3-768x576.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3-360x270.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-3.png 1280w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><br \/>\n\u9014\u4e2d\u304b\u3089\u30c6\u30b9\u30c8\u30bb\u30c3\u30c8\u306eloss\u304c\u6f38\u5897\u3057\u3066\u3044\u307e\u3059\uff0e12epoch\u3050\u3089\u3044\u3067\u6253\u3061\u5207\u3063\u305f\u65b9\u304c\u3088\u3055\u3052\u3067\u3059\uff0e<\/p>\n<p>\u904e\u5b66\u7fd2\u3092\u9632\u6b62\u3059\u308b\u305f\u3081\uff0cval_loss\u306b\u5909\u5316\u304c\u306a\u304f\u306a\u3063\u305f\u6bb5\u968e\u3067\u8a13\u7df4\u3092\u6253\u3061\u5207\u308bEarlyStopping\u304c\u6709\u52b9\u3067\u3059\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nmodel.compile(loss = 'mean_squared_error',\noptimizer = SGD(lr = 0.01, momentum = 0.9, nesterov = True),\nmetrics=&#x5B;'accuracy'])\nfrom keras.callbacks import EarlyStopping\nhistory = model.fit(X_train, y_train, epochs = 100, batch_size = 32,\nvalidation_data=(X_test, y_test), callbacks = &#x5B;EarlyStopping()])\nscore = model.evaluate(X_test, y_test, verbose = 0)\nprint('Test loss:', score&#x5B;0])\nprint('Test accuracy:', score&#x5B;1])\ny_pred = model.predict(X_test)\nrms = (np.mean((y_test - y_pred) ** 2)) ** 0.5\n#s = np.std(y_test - y_pred)\nprint(&quot;Neural Network RMS&quot;, rms)\n<\/pre>\n<pre class=\"prettyprint lang-python\">Epoch 6&#47;500\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.2442 - acc: 0.0104 - val_loss: 0.3072 - val_acc: 0.0156\nEpoch 7&#47;500\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.2215 - acc: 0.0083 - val_loss: 0.9356 - val_acc: 0.0062\nTest loss: 0.935621553887\nTest accuracy: 0.00623052959502\nNeural Network RMS 2.86083230159\n<\/pre>\n<p>\u3059\u3050\u306b\u8a08\u7b97\u304c\u6253\u3061\u5207\u3089\u308c\u307e\u3057\u305f\uff0e\u904e\u5b66\u7fd2\u3082\u9632\u3050\u3053\u3068\u304c\u3067\u304d\u3066\u3044\u307e\u3057\u305f\uff0e<br \/>\n<img decoding=\"async\" class=\"aligncenter size-large wp-image-1204\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1-1024x768.png\" alt=\"\" width=\"480\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1-1024x768.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1-300x225.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1-768x576.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1-360x270.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-1.png 1280w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/p>\n<h3>KerasRegressor<\/h3>\n<p><a href=\"https:\/\/iwatobipen.wordpress.com\/2016\/12\/13\/build-regression-model-in-keras\/\">Build regression in Keras<\/a>\u3092\u53c2\u8003\u306bKerasRegressor\u3092\u4f7f\u3046\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nimport numpy as np\nfrom sklearn import model_selection\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.metrics import r2_score\nfrom keras.models import Sequential\nfrom keras.layers import Activation, Dense\nfrom keras.wrappers.scikit_learn import KerasRegressor\nfrom keras.optimizers import SGD\n#\u4fdd\u5b58\u6e08\u307f\u30d5\u30a1\u30a4\u30eb\u3092\u8aad\u307f\u8fbc\u3080\nX_train = np.load(&quot;X_train.npy&quot;)\ny_train = np.load(&quot;y_train.npy&quot;)\nX_test = np.load(&quot;X_test.npy&quot;)\ny_test = np.load(&quot;y_test.npy&quot;)\n\ndef base_model():\nmodel = Sequential()\nmodel.add( Dense( units = 50, input_dim = X_train.shape&#x5B;1] ) )\nmodel.add( Activation( &quot;sigmoid&quot; ) )\nmodel.add( Dense( units = 1 ) )\nmodel.compile( loss = 'mean_squared_error',  optimizer = SGD(lr = 0.01, momentum = 0.9, nesterov = True),\nmetrics=&#x5B;'accuracy'] )\nreturn model\n\nestimator = KerasRegressor( build_fn = base_model,\nepochs = 10,\nbatch_size = 32,\nvalidation_data=(X_test, y_test)\n)\n\nestimator.fit( X_train, y_train )\n<\/pre>\n<p>\u51fa\u529b<\/p>\n<pre class=\"prettyprint lang-python\">Epoch 8&#47;10\nEpoch 8&#47;10\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.2449 - acc: 0.0114 - val_loss: 0.3433 - val_acc: 0.0187\nEpoch 9&#47;10\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.2063 - acc: 0.0135 - val_loss: 0.4258 - val_acc: 0.0125\nEpoch 10&#47;10\n961&#47;961 &#91;==============================&#93; - 0s - loss: 0.2354 - acc: 0.0114 - val_loss: 0.2858 - val_acc: 0.0187   \n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\ny_pred = estimator.predict(X_test)\nrms = (np.mean((y_test - y_pred) ** 2)) ** 0.5\n#s = np.std(y_test - y_pred)\nprint(&quot;Neural Network RMS&quot;, rms)\n<\/pre>\n<pre class=\"prettyprint lang-python\">Neural Network RMS 0.53456464492\n<\/pre>\n<p>\u53ef\u8996\u5316<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nimport matplotlib.pyplot as plt\nplt.figure()\nplt.scatter(y_train, estimator.predict(X_train), label = 'Train', c = 'blue')\nplt.title('Neural Network Predictor')\nplt.xlabel('Measured Solubility')\nplt.ylabel('Predicted Solubility')\nplt.scatter(y_test, estimator.predict(X_test), c = 'lightgreen', label = 'Test', alpha = 0.8)\nplt.legend(loc = 4)\nplt.show()\n<\/pre>\n<p><img decoding=\"async\" class=\"aligncenter size-large wp-image-1224\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2-1024x768.png\" alt=\"\" width=\"480\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2-1024x768.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2-300x225.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2-768x576.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2-360x270.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_2.png 1280w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><br \/>\n(\u4e86)<\/p>\n<h2>&lt;&lt;\u4ed8\u9332&gt;&gt; \u90e8\u5206\u6700\u5c0f\u4e8c\u4e57\u56de\u5e30 (PLSR)<\/h2>\n<p>\u53c2\u8003\u30b5\u30a4\u30c8: <a href=\"http:\/\/cheminformist.itmol.com\/TEST\/wp-content\/uploads\/2015\/04\/PLS2.html\">scikit-learn: Partial Least Squares<\/a><br \/>\n\u30b1\u30e2\u30e1\u30c8\u30ea\u30c3\u30af\u30b9 (\u5316\u5b66\u8a08\u91cf\u5b66) \u5206\u91ce\u3067\u4f7f\u308f\u308c\u308b\u90e8\u5206\u6700\u5c0f\u4e8c\u4e57\u56de\u5e30 (Partial least squares regression) \u3067\u306f\u3069\u3046\u3067\u3057\u3087\u3046\u304b\uff0e<\/p>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\nimport numpy as np\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn import model_selection\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.metrics import r2_score\nfrom sklearn.cross_decomposition import PLSRegression\nimport sklearn\nprint(&quot;sklearn ver.&quot;, sklearn.__version__)\nprint(&quot;numpy ver.&quot;, np.__version__)\n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#\u4fdd\u5b58\u6e08\u307f\u30d5\u30a1\u30a4\u30eb\u3092\u8aad\u307f\u8fbc\u3080\nX = np.load(&quot;X_2d.npy&quot;)\nY = np.load(&quot;Y_2d.npy&quot;)\n#\u8a13\u7df4\u30bb\u30c3\u30c8\u3068\u30c6\u30b9\u30c8\u30bb\u30c3\u30c8\u306b\u30e9\u30f3\u30c0\u30e0\u3067\u5206\u5272\nX_train, X_test, y_train, y_test = model_selection.train_test_split(X,\nY, test_size = 0.25, random_state = 42)\n<\/pre>\n<pre class=\"brush: python; title: ; notranslate\" title=\"\">\n#\u6eb6\u89e3\u5ea6\u306e\u5206\u6563\u3092\u3088\u304f\u8aac\u660e\u3059\u308b\u56e0\u5b50\u3092\u8a08\u7b97\u3057 (\u4e3b\u6210\u5206\u5206\u6790\u306f\u30c7\u30fc\u30bf\u3092\u3088\u304f\u8aac\u660e\u3059\u308b\u56e0\u5b50\u3092\u8a08\u7b97\u3059\u308b)\uff0c15\u756a\u76ee\u307e\u3067\u306e\u56e0\u5b50\u3092\u4f7f\u3063\u3066\u56de\u5e30\u5206\u6790\u3092\u884c\u3046\uff0e\npls2 = PLSRegression(n_components = 15, scale = True)\npls2.fit(X_train, y_train)\npred_train = pls2.predict(X_train)\npred_test = pls2.predict(X_test)\nrms = (np.mean((y_test - pred_test)**2))**0.5\n#s = np.std(y_test - y_pred)\nprint(&quot;PLS regression RMS&quot;, rms)\n#\u53ef\u8996\u5316\nimport pylab as plt\nplt.figure()\nplt.scatter(y_train, pred_train, label = 'Train', c = 'blue')\nplt.title('PLSR Predictor')\nplt.xlabel('Measured Solubility')\nplt.ylabel('Predicted Solubility')\nplt.scatter(y_test, pred_test, c = 'lightgreen', label = 'Test', alpha = 0.8)\nplt.legend(loc = 4)\nplt.show()\n<\/pre>\n<p>PLS regression RMS 2.82767316393\uff0c\u3067\u3057\u305f\uff0e\u304d\u308c\u3044\u306b\u56de\u5e30\u3067\u304d\u3066\u3044\u307e\u3057\u305f\uff0e\u8a08\u7b97\u306f\u4e00\u77ac\u3067\u7d42\u308f\u308a\u307e\u3057\u305f\uff0e<br \/>\n<img decoding=\"async\" class=\"aligncenter size-large wp-image-1205\" src=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2-1024x768.png\" alt=\"\" width=\"480\" srcset=\"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2-1024x768.png 1024w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2-300x225.png 300w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2-768x576.png 768w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2-360x270.png 360w, https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/Figure_1-2.png 1280w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u74b0\u5883: macOS Sierra 10.12.5, CPU: &#8230;<\/p>\n","protected":false},"author":1,"featured_media":1167,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_uag_custom_page_level_css":"","_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_uf_show_specific_survey":0,"_uf_disable_surveys":false,"_locale":"","_original_post":"","footnotes":""},"categories":[5],"tags":[],"class_list":["post-1160","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-5","ja"],"aioseo_notices":[],"uagb_featured_image_src":{"full":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",1652,450,false],"thumbnail":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-150x150.png",150,150,true],"medium":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-300x82.png",300,82,true],"medium_large":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-768x209.png",768,209,true],"large":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781-1024x279.png",800,218,true],"1536x1536":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",1536,418,false],"2048x2048":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",1652,450,false],"onepress-blog-small":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",300,82,false],"onepress-small":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",480,131,false],"onepress-medium":["https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp\/wp-content\/uploads\/2017\/07\/d2e1df578a84f818e57b819e7ecef781.png",640,174,false]},"uagb_author_info":{"display_name":"RCY","author_link":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/author\/charlesy\/"},"uagb_comment_info":2,"uagb_excerpt":"\u74b0\u5883: macOS Sierra 10.12.5, CPU: ...","_links":{"self":[{"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/posts\/1160","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/comments?post=1160"}],"version-history":[{"count":60,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/posts\/1160\/revisions"}],"predecessor-version":[{"id":1645,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/posts\/1160\/revisions\/1645"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/media\/1167"}],"wp:attachment":[{"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/media?parent=1160"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/categories?post=1160"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ag.kagawa-u.ac.jp\/charlesy\/wp-json\/wp\/v2\/tags?post=1160"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}