{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"toc_visible":true},"kernelspec":{"name":"python3","display_name":"Python 3"},"accelerator":"GPU"},"cells":[{"cell_type":"markdown","metadata":{"id":"LpnEVMCYYlnD"},"source":["# CIFAR-100 και Tranfer Learning"]},{"cell_type":"code","metadata":{"id":"STXQMBuN3nZ6","executionInfo":{"status":"ok","timestamp":1671004792149,"user_tz":-120,"elapsed":468,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["from __future__ import absolute_import, division, print_function, unicode_literals # legacy compatibility\n","\n","import tensorflow as tf\n","from tensorflow.keras import datasets, layers, models\n","from tensorflow.keras.preprocessing.image import ImageDataGenerator\n","from tensorflow.keras.utils import to_categorical\n","\n","import numpy as np\n","import pandas as pd\n","import matplotlib.pyplot as plt"],"execution_count":2,"outputs":[]},{"cell_type":"markdown","source":["## To CIFAR-100\n","\n","![](https://datarepository.wolframcloud.com/resources/images/69f/69f1e629-81e6-4eaa-998f-f6734fcd2cb3-io-4-o.en.gif)\n","\n","Τα [CIFAR-10 και CIFAR-100](https://www.cs.toronto.edu/~kriz/cifar.html) είναι μαζί με το MNIST τα διασημότερα dataset στην όραση υπολογιστών\n","\n","Αποτελούν υποσύνολα του συνόλου δεδομένων 80 million tiny images. Συλλέχθηκαν από τους Alex Krizhevsky (του AlexNet), Vinod Nair και Geoffrey Hinton.\n","\n","Το CIFAR-10 έχει 10 κατηγορίες εικόνων και το CIFAR-100 100.\n","\n","CIFAR σημαίνει Canadian Institute for Advanced Research.\n","\n","Το 2004, ο Geoffrey Hinton άρχισε να ηγείται του προγράμματος Νευρωνικός Υπολογισμός και Προσαρμοστική Αντίληψη του CIFAR. Στα μέλη του περιλαμβάνονται οι Yoshua Bengio και Yann LeCun, μεταξύ άλλων νευροεπιστημόνων, επιστημόνων υπολογιστών, βιολόγων, ηλεκτρολόγων μηχανικών, φυσικών και ψυχολόγων. \n","\n","Σήμερα, οι τρεις τους αναγνωρίζονται ευρέως ως οι πρωτοπόροι της βαθιάς μάθησης. \n","\n","Το 2019, η Association for Computing Machinery (ACM), ονόμασε τους Hinton, Bengio και LeCun ως αποδέκτες του βραβείου ACM A.M. Turing Award 2018 για τις εννοιολογικές και μηχανικές ανακαλύψεις που έκαναν τα βαθιά νευρωνικά δίκτυα ένα κρίσιμο συστατικό της πληροφορικής.\n","\n","Oυσιαστικά, το Turing Award απονεμήθηκε στο CIFAR, το οποίο μετράει επίσης 20 νομπελίστες στις τάξεις του. \n","\n","\n","To CIFAR-100 δεν είναι εύκολο dataset. Έχει πολλές κατηγορίες, κάποιες πολύ κοντινές, και η ανάλυση είναι χαμηλή, 32x32 pixels."],"metadata":{"id":"M40EutWIR0bY"}},{"cell_type":"markdown","metadata":{"id":"DfEMjsB4Yurm"},"source":["## Εισαγωγή και επισκόπηση του συνόλου δεδομένων"]},{"cell_type":"code","metadata":{"id":"OCW71UaGzz0Q","colab":{"base_uri":"https://localhost:8080/"},"outputId":"d4e2e97b-eee5-4526-8ef1-e016819c6f81","executionInfo":{"status":"ok","timestamp":1671004906214,"user_tz":-120,"elapsed":17223,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# load the entire dataset\n","(x_train_ds, y_train_ds), (x_test_ds, y_test_ds) = tf.keras.datasets.cifar100.load_data(label_mode='fine')"],"execution_count":3,"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading data from https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz\n","169001437/169001437 [==============================] - 14s 0us/step\n"]}]},{"cell_type":"code","metadata":{"id":"kGKYHffEE1do","colab":{"base_uri":"https://localhost:8080/"},"outputId":"c1dc6e79-a2f9-497f-9a6c-43aeb0fd19df","executionInfo":{"status":"ok","timestamp":1671004908533,"user_tz":-120,"elapsed":5,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["print(x_train_ds.shape)"],"execution_count":4,"outputs":[{"output_type":"stream","name":"stdout","text":["(50000, 32, 32, 3)\n"]}]},{"cell_type":"code","metadata":{"id":"PgIN2h_KuCp_","executionInfo":{"status":"ok","timestamp":1671004920746,"user_tz":-120,"elapsed":615,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# διαβάζουμρ τα ονόματα των κλάσεων από ένα text file που τα έχουμε αποθηκεύσει \n","\n","CIFAR100_LABELS_LIST = pd.read_csv('https://pastebin.com/raw/qgDaNggt', sep=',', header=None).astype(str).values.tolist()[0]"],"execution_count":5,"outputs":[]},{"cell_type":"code","metadata":{"id":"_B4-tvVOQq3j","colab":{"base_uri":"https://localhost:8080/"},"outputId":"3c04b238-506e-4b02-d8a2-ddf906448dfa","executionInfo":{"status":"ok","timestamp":1671004922891,"user_tz":-120,"elapsed":7,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# print our classes\n","print(CIFAR100_LABELS_LIST)"],"execution_count":6,"outputs":[{"output_type":"stream","name":"stdout","text":["['apple', ' aquarium_fish', ' baby', ' bear', ' beaver', ' bed', ' bee', ' beetle', ' bicycle', ' bottle', ' bowl', ' boy', ' bridge', ' bus', ' butterfly', ' camel', ' can', ' castle', ' caterpillar', ' cattle', ' chair', ' chimpanzee', ' clock', ' cloud', ' cockroach', ' couch', ' crab', ' crocodile', ' cup', ' dinosaur', ' dolphin', ' elephant', ' flatfish', ' forest', ' fox', ' girl', ' hamster', ' house', ' kangaroo', ' keyboard', ' lamp', ' lawn_mower', ' leopard', ' lion', ' lizard', ' lobster', ' man', ' maple_tree', ' motorcycle', ' mountain', ' mouse', ' mushroom', ' oak_tree', ' orange', ' orchid', ' otter', ' palm_tree', ' pear', ' pickup_truck', ' pine_tree', ' plain', ' plate', ' poppy', ' porcupine', ' possum', ' rabbit', ' raccoon', ' ray', ' road', ' rocket', ' rose', ' sea', ' seal', ' shark', ' shrew', ' skunk', ' skyscraper', ' snail', ' snake', ' spider', ' squirrel', ' streetcar', ' sunflower', ' sweet_pepper', ' table', ' tank', ' telephone', ' television', ' tiger', ' tractor', ' train', ' trout', ' tulip', ' turtle', ' wardrobe', ' whale', ' willow_tree', ' wolf', ' woman', ' worm']\n"]}]},{"cell_type":"code","metadata":{"id":"QpGXgTs_5ZCk","colab":{"base_uri":"https://localhost:8080/","height":459},"outputId":"e20e7b8a-f588-442c-91e7-70f972a7c508","executionInfo":{"status":"ok","timestamp":1671004928132,"user_tz":-120,"elapsed":2568,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# get (train) dataset dimensions\n","data_size, img_rows, img_cols, img_channels = x_train_ds.shape\n","\n","# set validation set percentage (wrt the training set size)\n","validation_percentage = 0.15\n","val_size = round(validation_percentage * data_size)\n","\n","# Reserve val_size samples for validation and normalize all values\n","x_val = x_train_ds[-val_size:]/255\n","y_val = y_train_ds[-val_size:]\n","x_train = x_train_ds[:-val_size]/255\n","y_train = y_train_ds[:-val_size]\n","x_test = x_test_ds/255\n","y_test = y_test_ds\n","\n","print(len(x_val))\n","\n","# summarize loaded dataset\n","print('Train: X=%s, y=%s' % (x_train.shape, y_train.shape))\n","print('Validation: X=%s, y=%s' % (x_val.shape, y_val.shape))\n","print('Test: X=%s, y=%s' % (x_test.shape, y_test.shape))\n","\n","# get class label from class index\n","def class_label_from_index(fine_category):\n","  return(CIFAR100_LABELS_LIST[fine_category.item(0)])\n","\n","# plot first few images\n","plt.figure(figsize=(6, 6))\n","for i in range(9):\n","\t# define subplot\n","  plt.subplot(330 + 1 + i).set_title(class_label_from_index(y_train[i]))\n","\t# plot raw pixel data\n","  plt.imshow(x_train[i], cmap=plt.get_cmap('gray'))\n","  #show the figure\n","plt.show()"],"execution_count":7,"outputs":[{"output_type":"stream","name":"stdout","text":["7500\n","Train: X=(42500, 32, 32, 3), y=(42500, 1)\n","Validation: X=(7500, 32, 32, 3), y=(7500, 1)\n","Test: X=(10000, 32, 32, 3), y=(10000, 1)\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 432x432 with 9 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"0cniJE8eZQAA"},"source":["## Συναρτήσεις εκπαίδευσης"]},{"cell_type":"markdown","metadata":{"id":"G3xB9x5JZjSN"},"source":["Θα χρησιμοποιήσουμε την ιδιότητα prefetch του TF2 για καλύτερες επιδόσεις στην εκπαίδευση: "]},{"cell_type":"code","metadata":{"id":"KPFYqOmIa5Fr","executionInfo":{"status":"ok","timestamp":1671004956810,"user_tz":-120,"elapsed":5143,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# we user prefetch https://www.tensorflow.org/api_docs/python/tf/data/Dataset#prefetch \n","# see also AUTOTUNE\n","# the dataset is now \"infinite\"\n","\n","BATCH_SIZE = 128\n","AUTOTUNE = tf.data.experimental.AUTOTUNE # https://www.tensorflow.org/guide/data_performance\n","\n","def _input_fn(x,y, BATCH_SIZE):\n","  ds = tf.data.Dataset.from_tensor_slices((x,y))\n","  ds = ds.shuffle(buffer_size=data_size)\n","  ds = ds.repeat()\n","  ds = ds.batch(BATCH_SIZE)\n","  ds = ds.prefetch(buffer_size=AUTOTUNE)\n","  return ds\n","\n","train_ds =_input_fn(x_train,y_train, BATCH_SIZE) #PrefetchDataset object\n","validation_ds =_input_fn(x_val,y_val, BATCH_SIZE) #PrefetchDataset object\n","test_ds =_input_fn(x_test,y_test, BATCH_SIZE) #PrefetchDataset object\n","\n","# steps_per_epoch and validation_steps for training and validation: https://www.tensorflow.org/guide/keras/train_and_evaluate\n","\n","def train_model(model, epochs = 10, steps_per_epoch = 2, validation_steps = 1):\n","  history = model.fit(train_ds, epochs=epochs, steps_per_epoch=steps_per_epoch, validation_data=validation_ds, validation_steps=validation_steps)\n","  return(history)"],"execution_count":8,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"onYIggE9Z2f4"},"source":["## Γραφικές παραστάσεις εκπαίδευσης και απόδοση στο σύνολο ελέγχου"]},{"cell_type":"code","metadata":{"id":"vdWPm3zqbRo1","executionInfo":{"status":"ok","timestamp":1671004958313,"user_tz":-120,"elapsed":3,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# plot diagnostic learning curves\n","def summarize_diagnostics(history):\n","\tplt.figure(figsize=(8, 8))\n","\tplt.suptitle('Training Curves')\n","\t# plot loss\n","\tplt.subplot(211)\n","\tplt.title('Cross Entropy Loss')\n","\tplt.plot(history.history['loss'], color='blue', label='train')\n","\tplt.plot(history.history['val_loss'], color='orange', label='val')\n","\tplt.legend(loc='upper right')\n","\t# plot accuracy\n","\tplt.subplot(212)\n","\tplt.title('Classification Accuracy')\n","\tplt.plot(history.history['accuracy'], color='blue', label='train')\n","\tplt.plot(history.history['val_accuracy'], color='orange', label='val')\n","\tplt.legend(loc='lower right')\n","\treturn plt\n"," \n","# print test set evaluation metrics\n","def model_evaluation(model, evaluation_steps):\n","\tprint('\\nTest set evaluation metrics')\n","\tloss0,accuracy0 = model.evaluate(test_ds, steps = evaluation_steps)\n","\tprint(\"loss: {:.2f}\".format(loss0))\n","\tprint(\"accuracy: {:.2f}\".format(accuracy0))\n","\n","def model_report(model, history, evaluation_steps = 10):\n","\tplt = summarize_diagnostics(history)\n","\tplt.show()\n","\tmodel_evaluation(model, evaluation_steps)"],"execution_count":9,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"Waa4c40Yay3k"},"source":["## Μοντέλα δικτύων"]},{"cell_type":"markdown","metadata":{"id":"cFTzQtMOa3Rv"},"source":["### Ένα μικρό συνελικτικό δίκτυο \"from scratch\""]},{"cell_type":"code","metadata":{"id":"LtcgTkHojt0G","executionInfo":{"status":"ok","timestamp":1671004963560,"user_tz":-120,"elapsed":3,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# a simple CNN https://www.tensorflow.org/tutorials/images/cnn\n","\n","def init_simple_model(summary):\n","  model = models.Sequential()\n","  model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32,32,3)))\n","  model.add(layers.MaxPooling2D((2, 2)))\n","  model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n","  model.add(layers.MaxPooling2D((2, 2)))\n","  model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n","  model.add(layers.Flatten())\n","  model.add(layers.Dense(64, activation='relu'))\n","  model.add(layers.Dense(100, activation='softmax'))\n","  \n","  model.compile(optimizer=tf.optimizers.Adam(learning_rate=0.0001), loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=[\"accuracy\"])\n","  if summary: \n","    model.summary()\n","  return model"],"execution_count":10,"outputs":[]},{"cell_type":"code","metadata":{"id":"dSbtouO9lGvj","colab":{"base_uri":"https://localhost:8080/"},"outputId":"e8573e93-f087-4613-ad70-d80da162fe42","executionInfo":{"status":"ok","timestamp":1671005068592,"user_tz":-120,"elapsed":42856,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["SIMPLE_MODEL = init_simple_model(summary = True)\n","SIMPLE_MODEL_history = train_model(SIMPLE_MODEL, 50, 30, 5)"],"execution_count":11,"outputs":[{"output_type":"stream","name":"stdout","text":["Model: \"sequential\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," conv2d (Conv2D)             (None, 30, 30, 32)        896       \n","                                                                 \n"," max_pooling2d (MaxPooling2D  (None, 15, 15, 32)       0         \n"," )                                                               \n","                                                                 \n"," conv2d_1 (Conv2D)           (None, 13, 13, 64)        18496     \n","                                                                 \n"," max_pooling2d_1 (MaxPooling  (None, 6, 6, 64)         0         \n"," 2D)                                                             \n","                                                                 \n"," conv2d_2 (Conv2D)           (None, 4, 4, 64)          36928     \n","                                                                 \n"," flatten (Flatten)           (None, 1024)              0         \n","                                                                 \n"," dense (Dense)               (None, 64)                65600     \n","                                                                 \n"," dense_1 (Dense)             (None, 100)               6500      \n","                                                                 \n","=================================================================\n","Total params: 128,420\n","Trainable params: 128,420\n","Non-trainable params: 0\n","_________________________________________________________________\n","Epoch 1/50\n","30/30 [==============================] - 9s 16ms/step - loss: 4.6040 - accuracy: 0.0109 - val_loss: 4.6045 - val_accuracy: 0.0172\n","Epoch 2/50\n","30/30 [==============================] - 0s 10ms/step - loss: 4.6026 - accuracy: 0.0107 - val_loss: 4.5965 - val_accuracy: 0.0141\n","Epoch 3/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.5977 - accuracy: 0.0156 - val_loss: 4.5958 - val_accuracy: 0.0125\n","Epoch 4/50\n","30/30 [==============================] - 0s 10ms/step - loss: 4.5892 - accuracy: 0.0203 - val_loss: 4.5869 - val_accuracy: 0.0156\n","Epoch 5/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.5793 - accuracy: 0.0161 - val_loss: 4.5750 - val_accuracy: 0.0156\n","Epoch 6/50\n","30/30 [==============================] - 0s 7ms/step - loss: 4.5597 - accuracy: 0.0201 - val_loss: 4.5606 - val_accuracy: 0.0203\n","Epoch 7/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.5262 - accuracy: 0.0258 - val_loss: 4.5338 - val_accuracy: 0.0203\n","Epoch 8/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.4898 - accuracy: 0.0253 - val_loss: 4.4744 - val_accuracy: 0.0188\n","Epoch 9/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.4764 - accuracy: 0.0232 - val_loss: 4.4875 - val_accuracy: 0.0172\n","Epoch 10/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.4354 - accuracy: 0.0297 - val_loss: 4.4201 - val_accuracy: 0.0328\n","Epoch 11/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.4150 - accuracy: 0.0276 - val_loss: 4.3638 - val_accuracy: 0.0422\n","Epoch 12/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.3732 - accuracy: 0.0378 - val_loss: 4.3530 - val_accuracy: 0.0328\n","Epoch 13/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.3227 - accuracy: 0.0440 - val_loss: 4.3296 - val_accuracy: 0.0453\n","Epoch 14/50\n","30/30 [==============================] - 0s 7ms/step - loss: 4.3011 - accuracy: 0.0544 - val_loss: 4.2790 - val_accuracy: 0.0594\n","Epoch 15/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.2811 - accuracy: 0.0461 - val_loss: 4.2956 - val_accuracy: 0.0484\n","Epoch 16/50\n","30/30 [==============================] - 0s 9ms/step - loss: 4.2068 - accuracy: 0.0617 - val_loss: 4.2259 - val_accuracy: 0.0594\n","Epoch 17/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.2053 - accuracy: 0.0591 - val_loss: 4.1944 - val_accuracy: 0.0625\n","Epoch 18/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.1777 - accuracy: 0.0698 - val_loss: 4.1646 - val_accuracy: 0.0688\n","Epoch 19/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.1536 - accuracy: 0.0688 - val_loss: 4.1479 - val_accuracy: 0.0641\n","Epoch 20/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.1566 - accuracy: 0.0688 - val_loss: 4.1805 - val_accuracy: 0.0781\n","Epoch 21/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.1033 - accuracy: 0.0802 - val_loss: 4.1246 - val_accuracy: 0.0750\n","Epoch 22/50\n","30/30 [==============================] - 0s 7ms/step - loss: 4.1058 - accuracy: 0.0833 - val_loss: 4.0961 - val_accuracy: 0.0844\n","Epoch 23/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.1057 - accuracy: 0.0773 - val_loss: 4.1447 - val_accuracy: 0.0812\n","Epoch 24/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0758 - accuracy: 0.0812 - val_loss: 4.0685 - val_accuracy: 0.0766\n","Epoch 25/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0446 - accuracy: 0.0951 - val_loss: 4.0530 - val_accuracy: 0.1047\n","Epoch 26/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0492 - accuracy: 0.0930 - val_loss: 4.0672 - val_accuracy: 0.0781\n","Epoch 27/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0408 - accuracy: 0.0906 - val_loss: 4.1191 - val_accuracy: 0.0734\n","Epoch 28/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9908 - accuracy: 0.1031 - val_loss: 4.0505 - val_accuracy: 0.0906\n","Epoch 29/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0043 - accuracy: 0.0987 - val_loss: 4.0430 - val_accuracy: 0.0797\n","Epoch 30/50\n","30/30 [==============================] - 0s 8ms/step - loss: 4.0033 - accuracy: 0.0964 - val_loss: 3.9554 - val_accuracy: 0.1047\n","Epoch 31/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9742 - accuracy: 0.1081 - val_loss: 4.0049 - val_accuracy: 0.1125\n","Epoch 32/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9675 - accuracy: 0.0987 - val_loss: 4.0033 - val_accuracy: 0.0891\n","Epoch 33/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9868 - accuracy: 0.1010 - val_loss: 3.9739 - val_accuracy: 0.1141\n","Epoch 34/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9273 - accuracy: 0.1094 - val_loss: 3.9180 - val_accuracy: 0.1141\n","Epoch 35/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9448 - accuracy: 0.1055 - val_loss: 4.0075 - val_accuracy: 0.1063\n","Epoch 36/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9008 - accuracy: 0.1133 - val_loss: 3.9301 - val_accuracy: 0.1000\n","Epoch 37/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8969 - accuracy: 0.1138 - val_loss: 3.9767 - val_accuracy: 0.1016\n","Epoch 38/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.9266 - accuracy: 0.1141 - val_loss: 4.0029 - val_accuracy: 0.0969\n","Epoch 39/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8800 - accuracy: 0.1161 - val_loss: 3.9165 - val_accuracy: 0.1141\n","Epoch 40/50\n","30/30 [==============================] - 0s 9ms/step - loss: 3.8801 - accuracy: 0.1159 - val_loss: 3.8286 - val_accuracy: 0.1312\n","Epoch 41/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8655 - accuracy: 0.1255 - val_loss: 3.9459 - val_accuracy: 0.1172\n","Epoch 42/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8472 - accuracy: 0.1273 - val_loss: 3.9387 - val_accuracy: 0.0969\n","Epoch 43/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8921 - accuracy: 0.1169 - val_loss: 3.8542 - val_accuracy: 0.1281\n","Epoch 44/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8983 - accuracy: 0.1172 - val_loss: 3.8667 - val_accuracy: 0.1063\n","Epoch 45/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8460 - accuracy: 0.1289 - val_loss: 3.8271 - val_accuracy: 0.1234\n","Epoch 46/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8294 - accuracy: 0.1221 - val_loss: 3.7809 - val_accuracy: 0.1297\n","Epoch 47/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8391 - accuracy: 0.1266 - val_loss: 3.9372 - val_accuracy: 0.0922\n","Epoch 48/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.7936 - accuracy: 0.1289 - val_loss: 3.8227 - val_accuracy: 0.1437\n","Epoch 49/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8124 - accuracy: 0.1302 - val_loss: 3.8468 - val_accuracy: 0.1281\n","Epoch 50/50\n","30/30 [==============================] - 0s 8ms/step - loss: 3.8220 - accuracy: 0.1292 - val_loss: 3.8324 - val_accuracy: 0.1141\n"]}]},{"cell_type":"markdown","metadata":{"id":"12z4zsoVbPWR"},"source":["### Μεταφορά μάθησης: VGG16\n","Θα χρησιμοποιήσουμε ένα [VGG16](https://www.tensorflow.org/api_docs/python/tf/keras/applications/VGG16) προεκπαιδευμένο στο ImageNet, χωρίς το classification head."]},{"cell_type":"code","metadata":{"id":"6QJueWvUXUTJ","executionInfo":{"status":"ok","timestamp":1671005726048,"user_tz":-120,"elapsed":457,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["# transfer learning: VGG16 trained on ImageNet without the top layer\n","\n","def init_VGG16_model(summary):\n","  vgg_model=tf.keras.applications.VGG16(input_shape=(32,32,3), include_top=False, weights='imagenet')\n","  \n","  VGG16_MODEL=vgg_model.layers[0](vgg_model)\n","\n","  # unfreeze conv layers\n","  VGG16_MODEL.trainable=True\n","  \n","  dropout_layer = tf.keras.layers.Dropout(rate = 0.5)\n","  global_average_layer = tf.keras.layers.GlobalAveragePooling2D()\n","\n","  # add top layer for CIFAR100 classification\n","  prediction_layer = tf.keras.layers.Dense(len(CIFAR100_LABELS_LIST),activation='softmax')\n","  model = tf.keras.Sequential([VGG16_MODEL, dropout_layer, global_average_layer, prediction_layer])\n","  model.compile(optimizer=tf.optimizers.Adam(learning_rate=0.00005), loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=[\"accuracy\"])\n","  if summary: \n","    model.summary()\n","  return model"],"execution_count":14,"outputs":[]},{"cell_type":"code","metadata":{"id":"2bZChKpdh0Cn","colab":{"base_uri":"https://localhost:8080/"},"outputId":"ccd5968c-8fcf-42a6-968d-d9b642d30863","executionInfo":{"status":"ok","timestamp":1671005890743,"user_tz":-120,"elapsed":147219,"user":{"displayName":"Giorgos Siolas","userId":"10127542075805046236"}}},"source":["VGG16_MODEL = init_VGG16_model(True)\n","VGG16_MODEL_history = train_model(VGG16_MODEL, 50, 30, 5)\n","\n","#model_report(VGG16_MODEL, VGG16_MODEL_history, 30)"],"execution_count":15,"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n","58889256/58889256 [==============================] - 3s 0us/step\n","Model: \"sequential_1\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," vgg16 (Functional)          (None, 1, 1, 512)         14714688  \n","                                                                 \n"," dropout (Dropout)           (None, 1, 1, 512)         0         \n","                                                                 \n"," global_average_pooling2d (G  (None, 512)              0         \n"," lobalAveragePooling2D)                                          \n","                                                                 \n"," dense_2 (Dense)             (None, 100)               51300     \n","                                                                 \n","=================================================================\n","Total params: 14,765,988\n","Trainable params: 14,765,988\n","Non-trainable params: 0\n","_________________________________________________________________\n","Epoch 1/50\n","30/30 [==============================] - 4s 65ms/step - loss: 4.7653 - accuracy: 0.0120 - val_loss: 4.5828 - val_accuracy: 0.0312\n","Epoch 2/50\n","30/30 [==============================] - 2s 59ms/step - loss: 4.5938 - accuracy: 0.0128 - val_loss: 4.5713 - val_accuracy: 0.0156\n","Epoch 3/50\n","30/30 [==============================] - 2s 59ms/step - loss: 4.5901 - accuracy: 0.0141 - val_loss: 4.5555 - val_accuracy: 0.0156\n","Epoch 4/50\n","30/30 [==============================] - 2s 59ms/step - loss: 4.5692 - accuracy: 0.0151 - val_loss: 4.5264 - val_accuracy: 0.0250\n","Epoch 5/50\n","30/30 [==============================] - 2s 59ms/step - loss: 4.5299 - accuracy: 0.0240 - val_loss: 4.4326 - val_accuracy: 0.0531\n","Epoch 6/50\n","30/30 [==============================] - 2s 60ms/step - loss: 4.4970 - accuracy: 0.0299 - val_loss: 4.3707 - val_accuracy: 0.0875\n","Epoch 7/50\n","30/30 [==============================] - 2s 60ms/step - loss: 4.3968 - accuracy: 0.0484 - val_loss: 4.2963 - val_accuracy: 0.0828\n","Epoch 8/50\n","30/30 [==============================] - 2s 63ms/step - loss: 4.3287 - accuracy: 0.0625 - val_loss: 4.1478 - val_accuracy: 0.0906\n","Epoch 9/50\n","30/30 [==============================] - 2s 61ms/step - loss: 4.2562 - accuracy: 0.0802 - val_loss: 4.0623 - val_accuracy: 0.1000\n","Epoch 10/50\n","30/30 [==============================] - 2s 61ms/step - loss: 4.1256 - accuracy: 0.0956 - val_loss: 3.9631 - val_accuracy: 0.1297\n","Epoch 11/50\n","30/30 [==============================] - 2s 60ms/step - loss: 3.9432 - accuracy: 0.1260 - val_loss: 3.8004 - val_accuracy: 0.1703\n","Epoch 12/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.8514 - accuracy: 0.1471 - val_loss: 3.5655 - val_accuracy: 0.2109\n","Epoch 13/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.7256 - accuracy: 0.1651 - val_loss: 3.3720 - val_accuracy: 0.2516\n","Epoch 14/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.6062 - accuracy: 0.1927 - val_loss: 3.3615 - val_accuracy: 0.2234\n","Epoch 15/50\n","30/30 [==============================] - 2s 60ms/step - loss: 3.5270 - accuracy: 0.1940 - val_loss: 3.1731 - val_accuracy: 0.2609\n","Epoch 16/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.3376 - accuracy: 0.2289 - val_loss: 3.0568 - val_accuracy: 0.2969\n","Epoch 17/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.2549 - accuracy: 0.2352 - val_loss: 2.9977 - val_accuracy: 0.2969\n","Epoch 18/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.2195 - accuracy: 0.2521 - val_loss: 2.9309 - val_accuracy: 0.2875\n","Epoch 19/50\n","30/30 [==============================] - 2s 61ms/step - loss: 3.0773 - accuracy: 0.2721 - val_loss: 2.8319 - val_accuracy: 0.3266\n","Epoch 20/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.9823 - accuracy: 0.3008 - val_loss: 2.7604 - val_accuracy: 0.3469\n","Epoch 21/50\n","30/30 [==============================] - 2s 62ms/step - loss: 2.9090 - accuracy: 0.3016 - val_loss: 2.7093 - val_accuracy: 0.3609\n","Epoch 22/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.8642 - accuracy: 0.3143 - val_loss: 2.5988 - val_accuracy: 0.3484\n","Epoch 23/50\n","30/30 [==============================] - 2s 62ms/step - loss: 2.6695 - accuracy: 0.3432 - val_loss: 2.4390 - val_accuracy: 0.3688\n","Epoch 24/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.6098 - accuracy: 0.3589 - val_loss: 2.4339 - val_accuracy: 0.3594\n","Epoch 25/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.6127 - accuracy: 0.3508 - val_loss: 2.3084 - val_accuracy: 0.4109\n","Epoch 26/50\n","30/30 [==============================] - 2s 62ms/step - loss: 2.4964 - accuracy: 0.3708 - val_loss: 2.3636 - val_accuracy: 0.3734\n","Epoch 27/50\n","30/30 [==============================] - 2s 62ms/step - loss: 2.4971 - accuracy: 0.3815 - val_loss: 2.3976 - val_accuracy: 0.4031\n","Epoch 28/50\n","30/30 [==============================] - 2s 60ms/step - loss: 2.4456 - accuracy: 0.3935 - val_loss: 2.2308 - val_accuracy: 0.4516\n","Epoch 29/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.4140 - accuracy: 0.3917 - val_loss: 2.2595 - val_accuracy: 0.4344\n","Epoch 30/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.3683 - accuracy: 0.4031 - val_loss: 2.1995 - val_accuracy: 0.4453\n","Epoch 31/50\n","30/30 [==============================] - 2s 62ms/step - loss: 2.3262 - accuracy: 0.4102 - val_loss: 2.1314 - val_accuracy: 0.4516\n","Epoch 32/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.2991 - accuracy: 0.4120 - val_loss: 2.0713 - val_accuracy: 0.4688\n","Epoch 33/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.2715 - accuracy: 0.4122 - val_loss: 2.0959 - val_accuracy: 0.4609\n","Epoch 34/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.1794 - accuracy: 0.4479 - val_loss: 2.0017 - val_accuracy: 0.4688\n","Epoch 35/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.0622 - accuracy: 0.4641 - val_loss: 2.0641 - val_accuracy: 0.4656\n","Epoch 36/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.0812 - accuracy: 0.4656 - val_loss: 2.0374 - val_accuracy: 0.4578\n","Epoch 37/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.0821 - accuracy: 0.4656 - val_loss: 1.9930 - val_accuracy: 0.4609\n","Epoch 38/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.0199 - accuracy: 0.4794 - val_loss: 2.0539 - val_accuracy: 0.4750\n","Epoch 39/50\n","30/30 [==============================] - 2s 60ms/step - loss: 1.9943 - accuracy: 0.4857 - val_loss: 1.9744 - val_accuracy: 0.4781\n","Epoch 40/50\n","30/30 [==============================] - 2s 61ms/step - loss: 2.0240 - accuracy: 0.4734 - val_loss: 1.9476 - val_accuracy: 0.4781\n","Epoch 41/50\n","30/30 [==============================] - 2s 62ms/step - loss: 1.9945 - accuracy: 0.4826 - val_loss: 1.9304 - val_accuracy: 0.4906\n","Epoch 42/50\n","30/30 [==============================] - 2s 61ms/step - loss: 1.9549 - accuracy: 0.4977 - val_loss: 1.9342 - val_accuracy: 0.4875\n","Epoch 43/50\n","30/30 [==============================] - 2s 61ms/step - loss: 1.9270 - accuracy: 0.5036 - val_loss: 1.8346 - val_accuracy: 0.5281\n","Epoch 44/50\n","30/30 [==============================] - 2s 62ms/step - loss: 1.9759 - accuracy: 0.4859 - val_loss: 1.9668 - val_accuracy: 0.4766\n","Epoch 45/50\n","30/30 [==============================] - 2s 61ms/step - loss: 1.7994 - accuracy: 0.5273 - val_loss: 1.9548 - val_accuracy: 0.4734\n","Epoch 46/50\n","30/30 [==============================] - 2s 61ms/step - loss: 1.8003 - accuracy: 0.5143 - val_loss: 1.7756 - val_accuracy: 0.5266\n","Epoch 47/50\n","30/30 [==============================] - 2s 60ms/step - loss: 1.7769 - accuracy: 0.5320 - val_loss: 1.9514 - val_accuracy: 0.4922\n","Epoch 48/50\n","30/30 [==============================] - 2s 60ms/step - loss: 1.7765 - accuracy: 0.5271 - val_loss: 1.8205 - val_accuracy: 0.5063\n","Epoch 49/50\n","30/30 [==============================] - 2s 60ms/step - loss: 1.6827 - accuracy: 0.5594 - val_loss: 1.7065 - val_accuracy: 0.5656\n","Epoch 50/50\n","30/30 [==============================] - 2s 60ms/step - loss: 1.6864 - accuracy: 0.5516 - val_loss: 1.8616 - val_accuracy: 0.5203\n"]}]},{"cell_type":"markdown","metadata":{"id":"OadPASXYOllM"},"source":["### Μεταφορά μάθησης: ResNet152V2\n","Θα χρησιμοποιήσουμε ένα [ResNet152V2](https://www.tensorflow.org/api_docs/python/tf/keras/applications/ResNet152V2) προεκπαιδευμένο στο ImageNet, χωρίς το classification head."]},{"cell_type":"code","metadata":{"id":"-6WG_6YJM4T7"},"source":["# transfer learning: ResNet152V2 trained on ImageNet without the top layer\n","\n","\n","def init_ResNet152V2_model(summary):\n","  ResNet152V2_model=tf.keras.applications.ResNet152V2(input_shape=(32,32,3), include_top=False, weights='imagenet')\n","  \n","  ResNet152V2_MODEL=ResNet152V2_model.layers[0](ResNet152V2_model)\n","\n","  # unfreeze conv layers\n","  ResNet152V2_MODEL.trainable=True\n","  \n","#  dropout_layer = tf.keras.layers.Dropout(rate = 0.5)\n","#  global_average_layer = tf.keras.layers.GlobalAveragePooling2D()\n","\n","  # add top layer for CIFAR100 classification\n","  prediction_layer = tf.keras.layers.Dense(len(CIFAR100_LABELS_LIST),activation='softmax')\n","  model = tf.keras.Sequential([ResNet152V2_MODEL, prediction_layer])\n","  #model = tf.keras.Sequential([ResNet152V2_MODEL, dropout_layer, global_average_layer, prediction_layer])\n","  model.compile(optimizer=tf.optimizers.Adam(learning_rate=0.00005), loss=tf.keras.losses.sparse_categorical_crossentropy, metrics=[\"accuracy\"])\n","  if summary: \n","    model.summary()\n","  return model"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"C3TznZ7ENl30"},"source":["ResNet152V2_MODEL = init_ResNet152V2_model(True)\n","ResNet152V2_MODEL_history = train_model(ResNet152V2_MODEL,  50, 30, 5)"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"HUfob-kNfaSD"},"source":["# Βελτίωση της επίδοσης με πειράματα\n","\n","Από τα τρια προηγούμενα νευρωνικά, με τον μικρό αριθμό εποχών που θέσαμε, μοιάζει ότι δεν ισχύει το \"bigger is better\" καθώς το VGG16 φαίνεται να είναι το καταλληλότερο για το task.\n","\n","- Μπορείτε να δώσετε κάποιες ερμηνείες για την επίδοση των τριών μοντέλων;\n","\n","- Μπορείτε να βρείτε μέθοδο (μοντέλο και εκπαιδευτική διαδικασία) για την βελτίωση των επιδόσεων στο CIFAR-100;\n","\n","Α word of wisdom για το θέμα αυτό είναι: το κατάλληλο μοντέλο εξαρτάται από τρία πράγματα: *πόσο χρόνο έχουμε, τί υπολογιστικούς πόρους διαθέτουμε, ποιες επιδόσεις θεωρούμε καλές.*\n","\n","Στη συνέχεια εξετάζουμε κάποιες από τις δυνατότητες που έχουμε για να βελτιώσουμε τις επιδόσεις των νευρωνικών μας με πειράματα."]},{"cell_type":"markdown","metadata":{"id":"FiVKTH5rmhWn"},"source":["## Δοκιμές διαφορετικών μοντέλων\n","\n","Μπορείτε είτε να δοκιμάσετε μοντέλα \"from scratch\", όπου ορίζετε την αρχιτεκτονική του δικτύου όπως θέλετε, είτε να χρησιμοποιήσετε μεταφορά μάθησης.\n"]},{"cell_type":"markdown","metadata":{"id":"pAbqI8hapbX9"},"source":["\n","### Μοντέλα \"from scratch\"\n","\n","Μπορείτε να τροποποιήσετε/αλλάξετε το αρχικό μικρό συνελικτικό δίκτυο του παραδείγματος. Μπορείτε να συμβουλευτείτε \n","- τη [βιβλιογραφία απο το leaderboard του CIFAR-100](https://benchmarks.ai/cifar-100) για αρχιτεκτονικές και παραμέτρους των δικτύων\n","- ή/και να πάρετε ιδέες [από σχετική αναζήτηση στο Google Scholar](https://scholar.google.gr/scholar?hl=en&as_sdt=0%2C5&q=cifar+100+cnn&oq=cifa)"]},{"cell_type":"markdown","metadata":{"id":"8LZIFj-AmlPv"},"source":["### Μεταφορά μάθησης\n","\n","Εναλλακτικά, μπορείτε να χρησιμοποιήσετ τη [μεταφορά μάθησης του tf2](https://www.tensorflow.org/tutorials/images/transfer_learning). Σε αντίθεση με τα μοντέλα \"from scratch\" η μεταφορά μάθησης μας επιστρέφει έτοιμα μοντέλα με προκαθορισμένη αρχιτεκτονική στην οποία μπορούμε γενικά μόνο να προσθέσουμε επίπεδα, τα οποία συνήθως περιορίζοντα σε πλήρως διασυνδεδεμένα επίπεδα που εξειδικεύονται στο συγκεκριμένο task ταξινόμησης που έχουμε να επιτελέσουμε. "]},{"cell_type":"markdown","metadata":{"id":"GrCxOjJ7ush3"},"source":["#### Εκπαίδευση βαρών\n","\n","Ταυτόχρονα με την αρχιτεκτονική, στη μεταφορά μάθησης εισάγουμε και τη γνώση που έχει αποκτήσει το μοντέλο, δηλαδή τις τιμές των βαρών του όπως έχουν προκύψει μετά από εκπαίδευση συνήθως στο (τεράστιο) ImageNet. Οταν εισάγουμε ένα μοντέλο με μεταφορά μάθησης έχουμε τρεις επιλογές για την εκπαίδευση:\n","- να παγώσουμε τη συνελικτική βάση και να εκπαιδεύσουμε την κεφαλή ταξινόμησης (classification head). Αυτό αντιστοιχεί στο να χρησιμοποιήσουμε τη συνελικτική βάση για εξαγωγή χαρακτηριστικών (feature extraction), σημαία trainable = False.\n","- να συνεχίσουμε να εκπαιδεύουμε όλα τα επίπεδα του δικτύου, σημαία trainable = True.\n","- να εκπαιδευτεί μόνο ένα ποσοστό των επιπέδων, εβρισκόμενο προς την έξοδο του δικτύου. Οι σημαίες trainable εδώ θα πρέπει να οριστούν ανά επίπεδο.\n"]},{"cell_type":"markdown","metadata":{"id":"6WFvmWr9xEUz"},"source":["\n","#### Διαθέσιμα μοντέλα για μεταφορά μάθησης στο tf2\n","\n","1. tf.keras.applications. Ο πιο απλός τρόπος για να κάνουμε μεταφορά μάθησης στο tf2 είναι μέσω του [tf.keras.applications](https://www.tensorflow.org/api_docs/python/tf/keras/applications) που παρέχει προεκπαιδευμένα μοντέλα από το Keras και συγκεκριμένα τα δίκτυα: DenseNet, Inception-ResNet V2, Inception V3, MobileNet v1, MobileNet v2, NASNet-A, ResNet, ResNet v2, VGG16, VGG19 και Xception V1. Η εισαγωγή των μοντέλων γίνεται παρόμοια με αυτή που δείξαμε παραπάνω για το VGG16.\n","\n","2. TensorFlow Hub. Μπορείτε επίσης να χρησιμοποιήσετε μοντέλα τα οποία είναι διαθέσιμα στο αποθετήριο [TensoFlow Hub](https://tfhub.dev/s?fine-tunable=yes&module-type=image-augmentation,image-classification,image-feature-vector,image-generator,image-object-detection,image-others,image-style-transfer,image-rnn-agent&tf-version=tf2) το οποίο περιλαμβάνει πάνω από 100 προεκπαιδευμένα μοντέλα.\n","\n","3. Αποθηκευμένα μοντέλα απο τρίτες πηγές. Μπορείτε επίσης να κάνετε μεταφορά μάθησης από τρίτες πηγές, είτε του συνόλου του νευρωνικού, αρχιτεκτονικής και βαρών, είτε μόνο της αρχιτεκτονικής ή των βαρών. Το μοντέλο θα πρέπει να έχει αποθηκευθεί σε ένα από τα δύο φορμάτ, Keras HDF5 format (.h5 ή .keras) ή στο SavedModel format που αναφέραμε στην εισαγωγή. Τα βάρη μπορούν να εισαχθούν και μόνα τους ως Checkpoints. Για περισσότερα, διαβάστε σχετικά τα λήμματα [\"Save and load models\"](https://www.tensorflow.org/tutorials/keras/save_and_load), [\"Save and serialize\"](https://www.tensorflow.org/guide/keras/save_and_serialize), [\"Using the SavedModel format\"](https://www.tensorflow.org/guide/saved_model) και δείτε για παράδειγμα πως μπορούμε να κάνουμε μεταφορά μάθησης από τα state-of-the-art EfficientNets ([1](https://www.dlology.com/blog/transfer-learning-with-efficientnet/), [2](https://github.com/tensorflow/tpu/tree/master/models/official/efficientnethttps://)).\n","\n","Σημειώστε ότι πολλά μοντέλα απαιτούν μεγαλύτερες διαστάσεις στην είσοδο από αυτές του CIFAR-100 και κατά συνέπεια τα δεδομένα πρέπει να [μετασχηματιστούν](https://www.tensorflow.org/api_docs/python/tf/image/resize). Προσέξτε ωστόσο τις απαιτήσεις σε μνήμη όταν αυτοί οι μετασχηματισμοί γίνονται απευθείας σε μεταβλητές (βλ. πιο κάτω \"Διαχείριση μνήμης\"). \n"]},{"cell_type":"markdown","metadata":{"id":"qTOi1CQsPOTT"},"source":["#### Zoos and Gardens\n","\n","![](https://camo.githubusercontent.com/ffc27b8ea00af45c0e71785dda0d04658e964a1f8df583634637c61123add357/68747470733a2f2f73746f726167652e676f6f676c65617069732e636f6d2f6d6f64656c5f67617264656e5f6172746966616374732f54465f4d6f64656c5f47617264656e2e706e67)\n","\n","Οι διάφορες βιβλιοθήκες -όχι μόνο η TensorFlow- αποκαλούν τη συλλογή εκπαιδευμένων μοντέλων τους \"Zoo\" ή \"Garden\". Παράδειγμα το [TensorFlow Model Garden](https://github.com/tensorflow/models)."]},{"cell_type":"markdown","metadata":{"id":"m9J7yIJrr4Ku"},"source":["### Επαύξηση δεδομένων\n","\n","Μια τεχνική που μπορεί να σας δώσει καλά αποτελέσματα είναι η επάυξηση δεδομένων (data augmentation). Η επαύξηση δεδομένων επιτρέπει να δημιουργήσουμε μεγαλύτερη ποικιλία στα δεδομένα εφαρμόζοντας τυχαίους αλλά ρεαλιστικούς μετασχηματισμούς στις εικόνες, όπως πχ η περιστροφη.\n","\n","Μπορούμε να κάνουμε data augmetation με δύο τρόπους: με επίπεδα προεπεξεργασίας του Keras, ή με χρήση του tf.image. Δείτε [εδώ](https://www.tensorflow.org/tutorials/images/data_augmentation) σχετικά από το documentation του TensorFlow και [εδώ](https://stepup.ai/train_data_augmentation_keras/) ένα πρακτικό παράδειγμα στο CIFAR-10."]},{"cell_type":"markdown","metadata":{"id":"SRI_3XhBQ7sb"},"source":["## Παρατηρήσεις ως προς τη βελτιστοποίηση"]},{"cell_type":"markdown","metadata":{"id":"mtVx8MsZRQrn"},"source":["### Διαχείριση μνήμης (TFRecord)\n","\n","Η φόρτωση δεδομένων με τον τρόπο που το κάναμε παραπάνω στο απλό παράδειγμα υλοποίησης είναι πολύ βολική αλλά δεν είναι καθόλου αποτελεσματική ως προς τη διαχείριση της μνήμης. Συγκεκριμένα, με τον τρόπο αυτό, τα δεδομένα αποθηκεύονται απευθείας σε μεταβλητές, οι οποίες όλες μαζί καταλαμβάνουν τη RAM της CPU ή της GPU, κάτι που κάνει αδύνατη τη διαχείριση μεγάλων datasets ή τον μεταχηματισμό των δεδομένων όπως όταν κάνουμε αύξηση δεδομένων (data augmentation).\n","\n","Για να παρακαμφθεί αυτό το πρόβλημα, υπάρχει η δυνατότητα της σειριοποίησης των δεδομένων (serialization) και της αποθήκευσής τους σε αρχεία μεσαίου μεγέθους (κάποιων MB) τα οποία μπορούνα να αναγνωστούν γραμμικά. Το φορμάτ TFRecord είναι ένα φορμάτ που επιτρέπει την αποθήκευση σειράς δυαδικών εγγραφών. Διαβάστε τα σχετικά λήμματα [TFRecord and tf.Example](https://www.tensorflow.org/tutorials/load_data/tfrecord) και [tf.data: Build TensorFlow input pipelines](https://www.tensorflow.org/guide/data). \n","\n","Σημειώστε ότι με τη μέθοδο αυτή θα πρέπει να γίνει import η `tensorflow_datasets` και να χρησιμοποιήσουμε την `tfds.load` ώστε να αποθηκευθεί το σύνολο δεδομένων σε αρχεία tfrecord στο δίσκο (δείτε [εδώ](https://colab.research.google.com/github/tensorflow/datasets/blob/master/docs/overview.ipynb) ένα παράδειγμα). Φυσικά μπορούμε να μετατρέψουμε και τα πρωτογενή δεδομένα (raw data) του dataset όπως αρχεία jpg σε φορματ tfrecord όπως [εδώ](https://towardsdatascience.com/working-with-tfrecords-and-tf-train-example-36d111b3ff4d).\n","\n"]},{"cell_type":"markdown","metadata":{"id":"yypACH_oZx_i"},"source":["### Υπερεκπαίδευση\n","\n","Μπορείτε να πειραματιστείτε ως προς τον έλεγχο της υπερεκπαίδευσης (overfitting) με διάφορους τρόπους. Μεταξύ αυτών μπορούμε να αναφέρουμε τους εξής:\n","- Πρόωρος τερματισμός (early stopping). Μια μέθοδος που τερματίζει την εκπαίδευση αν δεν υπάρχει βελτίωση ως προς τη μετρική απόδοσης που παρακολουθούμε. [tf.keras.callbacks.EarlyStopping](https://www.tensorflow.org/api_docs/python/tf/keras/callbacks/EarlyStoppinghttps://)\n","- Dropout. Μια άλλη τεχνική για τη μείωση της υπερεκπαίδευσης είναι το Dropout. Είναι ένα είδος ομαλοποίησης (regularization) που επιβάλλει στα βάρη του δικτύου να παίρνουν μόνο μικρές τιμές. Εάν εφαρμόσουε dropout σε ένα επίπεδο του δικτύου, τότε ένα ποσοστό των βαρών του γίνεται τυχαία μηδενικό κατά την εκπαίδευση. [Dropout](https://www.tensorflow.org/tutorials/images/classification#dropout)\n","- Επαύξηση δεδομένων. Η υπερεκπαίδευση συνήθως συμβαίνει όταν έχουμε λίγα ή/και πολύ όμοια δεδομένα εκπαίδευσης. Ένας τρόπος να διορθωθεί αυτό το πρόβλημα είναι να αυξήσουμε τα δεδομένα (data augmentation). Το data augmentation δημιουργεί νέα δεδομένα εκπαίδευσης με βάση τα υπάρχοντα εφαρμόζοντας τυχαίους μετασχηματισμούς ώστε να προκύπτουν αληθοφανείς εικόνες. [Data augmentation](https://www.tensorflow.org/tutorials/images/classification#data_augmentation), [ImageDataGenerator](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator#class_imagedatagenerator)\n","\n","Βλέπε επίσης [Image classification](https://www.tensorflow.org/tutorials/images/classification)."]},{"cell_type":"markdown","metadata":{"id":"MBLBjFFFmsUP"},"source":["### Χρόνος εκπαίδευσης\n","\n","Το TensorFlow 2 προσφέρει νέους ή βελτιώνει διάφορους μηχανισμούς βελτιστοποίησης της εκπαίδευσης. Μεταξύ αυτών έχουμε τους εξής:\n","- Data prefetching (το χρησιμοποιήσαμε παραπάνω)\n","- Data reading parallelization \n","- Map transformation parallelization\n","- Caching\n","- Reducing memory footprint\n","\n","Συμβουλευτείτε για τα παραπάνω το [Better performance with the tf.data API](https://www.tensorflow.org/guide/data_performance)"]},{"cell_type":"markdown","metadata":{"id":"s3ddu1ECoCGQ"},"source":["### Εργαλεία υψηλού επιπέδου\n","\n","Μεταξύ των εργαλείων βελτιστοποίησης υψηλού επιπέδου (high-level) του TensorFlow μπορούμε να αναφέρουμε τα ακόλουθα:\n","\n","- [TensorBoard](https://www.tensorflow.org/tensorboard/get_started) και [What-If Tool](https://www.tensorflow.org/tensorboard/what_if_tool) Επικουρικό εργαλείο οπτικοποίησης/ανάλυσης για τον πειραματισμό στη εκπαίδευση\n","- [tf-explain](https://tf-explain.readthedocs.io/en/latest/) Προσφέρει μεθόδους επεξηγισιμότητας για το tf2\n","- [Keras Tuner](https://github.com/keras-team/keras-tuner) Βελτιστοποίηση υπερπαραμέτρων του Keras στο TensorFlow 2.0\n","- [AutoAugment](https://github.com/tensorflow/models/tree/master/research/autoaugment) Εκμάθηση της πολιτικης επαύξησης από τα δεδομένα"]},{"cell_type":"markdown","metadata":{"id":"VNHFNS981Qh0"},"source":["# Για τις χριστουγεννιάτικες διακοπές\n"]},{"cell_type":"markdown","source":["\n","## What's next for Deep Learning\n","\n","Οι νονοί της Τεχνητής Νοημοσύνης και νικητές του βραβείου Turing 2018 της ACM, Geoffrey Hinton, Yann LeCun και Yoshua Bengio, μοιράστηκαν τη σκηνή στη Νέα Υόρκη το βράδυ της Κυριακής 9 Φεβρουαρίου σε μια εκδήλωση που διοργάνωσε το τριακοστό τέταρτο συνέδριο AAAI για την Τεχνητή Νοημοσύνη (AAAI 2020). \n","\n","Η τριάδα των ερευνητών έχει καταστήσει τα βαθιά νευρωνικά δίκτυα ένα κρίσιμο συστατικό της πληροφορικής, και σε μεμονωμένες ομιλίες και μια συζήτηση σε πάνελ συζήτησαν τις απόψεις τους σχετικά με τις τρέχουσες προκλήσεις που αντιμετωπίζει η βαθιά μάθηση και το πού θα πρέπει να κατευθυνθεί."],"metadata":{"id":"uzoS4hnnY1pN"}},{"cell_type":"code","metadata":{"id":"ECLw-wy_wGvm"},"source":["from IPython.display import IFrame\n","IFrame(src='https://www.youtube.com/embed/UX8OubxsY8w', width=640, height=480)"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"mwhiplJkzR1W"},"source":["## Αντίλογος\n"]},{"cell_type":"markdown","source":["\n","### Η κριτική του Schmidhuber για τους LBH\n","\n","(Ο Schmidhuber και ο Hochreiter [εισήγαγαν τα LSTM το 1997](https://www.bioinf.jku.at/publications/older/2604.pdf))\n","\n","*Machine learning is the science of credit assignment. The machine learning community itself profits from proper credit assignment to its members. The inventor of an important method should get credit for inventing it. She may not always be the one who popularizes it. Then the popularizer should get credit for popularizing it (but not for inventing it). Relatively young research areas such as machine learning should adopt the honor code of mature fields such as mathematics: if you have a new theorem, but use a proof technique similar to somebody else's, you must make this very clear. If you \"re-invent\" something that was already known, and only later become aware of this, you must at least make it clear later.*\n","\n","*As a case in point, let me now comment on a recent article in [Nature (2015) about \"deep learning\"](http://www.nature.com/nature/journal/v521/n7553/full/nature14539.html) in artificial neural networks (NNs), by LeCun & Bengio & Hinton (LBH for short), three CIFAR-funded collaborators who call themselves the \"deep learning conspiracy\" (e.g., LeCun, 2015). They heavily cite each other. Unfortunately, however, they fail to credit the pioneers of the field, which originated half a century ago. All references below are taken from the recent [deep learning overview](http://www.idsia.ch/~juergen/deep-learning-overview.html) (Schmidhuber, 2015), except for a few papers listed beneath this critique focusing on nine items.*\n","\n","[Full text](http://people.idsia.ch/~juergen/deep-learning-conspiracy.html)\n","\n","\n","Στο πρώτο του claim ο Schmidhuber γράφει:\n","\n","Η έρευνα του LBH δεν αναφέρει καν τον πατέρα της βαθιάς μάθησης, τον Alexey Grigorevich Ivakhnenko, ο οποίος δημοσίευσε τους πρώτους γενικούς, λειτουργικούς αλγορίθμους μάθησης για βαθιά δίκτυα (π.χ. Ivakhnenko and Lapa, 1965). \n","\n","Εντοπίσαμε το εν λόγω \"Ivakhnenko, A. G. and Lapa, V. G. (1965). Cybernetic Predicting Devices. CCM Information Corporation.\" μεταφρασμένο το 1966 [εδώ](https://drive.google.com/file/d/1cSKixS3_kaVghwETTvReQcIpFisJ-H6Y/view?usp=sharing).\n","\n","![](https://i.imgur.com/bghR7pk.png)\n","\n","Κοιτάξτε το κεφάλαιο 4, σελίδα 148 του κειμένου και 155 του PDF, για να δείτε αν έχει δίκιο ο Schmidhuber!\n"],"metadata":{"id":"j5jOycHJzT4F"}}]}