{
"cells": [
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import tensorflow as tf\n",
"\n",
"# Basic math operations:\n",
"tf.add() \t sum\n",
"tf.subtract() \tsubstraction\n",
"tf.multiply() \tmultiplication\n",
"tf.div() \t division\n",
"tf.mod() \t module\n",
"tf.abs() \t absolute value\n",
"tf.negative() \tnegative value\n",
"tf.sign() \t return sign\n",
"tf.reciprocal() reciprocal\n",
"tf.square() \tsquare\n",
"tf.round() \t nearest intiger\n",
"tf.sqrt() \t square root\n",
"tf.pow() \t power\n",
"tf.exp() \t exponent\n",
"tf.log() \t logarithm\n",
"tf.maximum() \tmaximum\n",
"tf.minimum() \tminimum\n",
"tf.cos() \t cosine\n",
"tf.sin() \t sine\n",
"\n",
"# Basic operations on tensors:\n",
"tf.string_to_number() \tconverts string to numeric type\n",
"tf.cast() \t\t\t\tcasts to new type\n",
"tf.shape() \t\t\t\treturns shape of tensor\n",
"tf.reshape() \t\t\treshapes tensor\n",
"tf.diag() \t\t\t\tcreates tensor with given diagonal values\n",
"tf.zeros() \t\t\t\tcreates tensor with all elements set to zero\n",
"tf.fill() \t\t\t\tcreates tensor with all elements set given value\n",
"tf.concat() \t\t\tconcatenates tensors\n",
"tf.slice() \t\t\t\textracts slice from tensor\n",
"tf.transpose() \t\t\ttranspose the argument\n",
"tf.matmul() \t\t\tmatrices multiplication\n",
"tf.matrix_determinant() determinant of matrices\n",
"tf.matrix_inverse() \tcomputes inverse of matrices\n",
"\n",
"# Control Flow:\n",
"tf.while_loop() \trepeat body while condition true\n",
"tf.case() \t\t\tcase operator\n",
"tf.count_up_to() \tincriments ref untill limit\n",
"tf.tuple() \t\t\tgroups tensors together\n",
"\n",
"# Logical/Comparison Operators:\n",
"tf.equal() \t\t\treturns truth value element-wise\n",
"tf.not_equal() \t\treturns truth value of X!=Y\n",
"tf.less() \t\t\treturns truth value of X<Y\n",
"tf.less_equal() \treturns truth value of X<=Y\n",
"tf.greater() \t\treturns truth value of X>Y\n",
"tf.greater_equal() \treturns truth value of X>=Y\n",
"tf.is_nan() \t\treturns which elements are NaN\n",
"tf.logical_and() \treturns truth value of ‘AND’ for given tensors\n",
"tf.logical_or() \treturns truth value of ‘OR’ for given tensors\n",
"tf.logical_not() \treturns truth value of ‘NOT’ for given tensors\n",
"tf.logical_xor() \treturns truth value of ‘XOR’ for given tensors\n",
"\n",
"# Working with Images:\n",
"tf.image.decode_image() \t\t\t\t\tconverts image to tensor type uint8\n",
"tf.image.resize_images() \t\t\t\t\tresize images\n",
"tf.image.resize_image_with_crop_or_pad() \tresize image by cropping or padding\n",
"tf.image.flip_up_down() \t\t\t\t\tflip image horizontally\n",
"tf.image.rot90() \t\t\t\t\t\t\trotate image 90 degrees counter-clockwise\n",
"tf.image.rgb_to_grayscale() \t\t\t\tconverts image from RGB to grayscale\n",
"tf.image.per_image_standardization() \t\tscales image to zero mean and unit norm\n",
"\n",
"# Neural Networks:\n",
"tf.nn.relu() \t\trectified linear activation function\n",
"tf.nn.softmax() \tsoftmax activation function\n",
"tf.nn.sigmoid() \tsigmoid activation function\n",
"tf.nn.tanh() \t\thyperbolic tangent activation function\n",
"tf.nn.dropout \t\tdropout\n",
"tf.nn.bias_add \t\tadds bias to value\n",
"tf.nn.all_candidate_sampler() \tset of all classes\n",
"tf.nn.weighted_moments() \t\treturns mean and variance\n",
"tf.nn.softmax_cross_entropy_with_logits() \tsoftmax cross entropy\n",
"tf.nn.sigmoid_cross_entropy_with_logits() \tsigmoid cross entropy\n",
"tf.nn.l2_normalize() \tnormalization using L2 Norm\n",
"tf.nn.l2_loss() \t\tL2 loss\n",
"tf.nn.dynamic_rnn() \tRNN specified by given cell\n",
"tf.nn.conv2d() \t\t\t2D convolutions given 4D input\n",
"tf.nn.conv1d() \t\t\t1D convolution given 3D input\n",
"tf.nn.batch_normalization() \tbatch normalization\n",
"tf.nn.xw_plus_b() \t\t\t\tcomputes matmul(x,weights)+biases\n",
"\n",
"# High level Machine Learning:\n",
"tf.contrib.keras \tKeras API as high level API for TensorFlow\n",
"tf.contrib.layers.one_hot_column() \t\tone hot encoding\n",
"tf.contrib.learn.LogisticRegressor() \tlogistic regression\n",
"tf.contrib.learn.DNNClassifier() \t\tDNN classifier\n",
"tf.contrib.learn.DynamicRnnEstimator() \tRnn Estimator\n",
"tf.contrib.learn.KMeansClustering() \tK-Means Clusstering\n",
"tf.contrib.learn.LinearClassifier() \tlinear classifier\n",
"tf.contrib.learn.LinearRegressor() \t\tlinear regressor\n",
"tf.contrib.learn.extract_pandas_data() \textract data from Pandas dataframe\n",
"tf.contrib.metrics.accuracy() \t\t\taccuracy\n",
"tf.contrib.metrics.auc_using_histogram() \tAUC\n",
"tf.contrib.metrics.confusion_matrix() \tconfusion matrix\n",
"tf.contrib.metrics.streaming_mean_absolute_error() \tmean absolute error\n",
"tf.contrib.rnn.BasicLSTMCell() \t\t\tbasic lstm cell\n",
"tf.contrib.rnn.BasicRNNCell() \t\t\tbasic rnn cell\n",
"\n",
"# Placeholders and Variables:\n",
"tf.placeholder() \tdefines placeholder\n",
"tf.Variable(tf.random_normal([3, 4], stddev=0.1) \tdefines variable\n",
"tf.Variable(tf.zeros([50]), name=’x’) \t\t\t\tdefines variable\n",
"tf.global_variables_initializer() \tinitialize global variables\n",
"tf.local_variables_initializer() \tinitialize local variables\n",
"## pin variable to CPU\n",
"with tf.device(“/cpu:0”):\n",
" v = tf.Variable()\n",
"## pin variable to GPU\n",
"with tf.device(“/gpu:0”): \n",
" v = tf.Variable()\n",
"## run session\n",
"sess = tf.Session()\n",
"sess.run()\n",
"sess.close()\n",
"## run session(2)\n",
"with tf.Session() as session:\n",
" session.run()\n",
"\n",
"# Saving and restoring variables.\n",
"saver=tf.train.Saver() \t\n",
"saver.save(sess,’file_name’) \t\n",
"saver.restore(sess,’file_name’) \t\n",
"\n",
"# Working with Data:\n",
"tf.decode_csv() \tconverts csv to tensors\n",
"tf.read_file() \t\treads file\n",
"tf.write_file() \twrites to file\n",
"tf.train.batch() \tcreates batches of tensors\n"
]
}
],
"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": "3.5.2"
},
"toc": {
"colors": {
"hover_highlight": "#DAA520",
"navigate_num": "#000000",
"navigate_text": "#333333",
"running_highlight": "#FF0000",
"selected_highlight": "#FFD700",
"sidebar_border": "#EEEEEE",
"wrapper_background": "#FFFFFF"
},
"moveMenuLeft": true,
"nav_menu": {
"height": "12px",
"width": "252px"
},
"navigate_menu": true,
"number_sections": true,
"sideBar": true,
"threshold": 4,
"toc_cell": false,
"toc_section_display": "block",
"toc_window_display": false,
"widenNotebook": false
}
},
"nbformat": 4,
"nbformat_minor": 2
}