{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\nimport time\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nfrom matplotlib import pyplot as plt\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 5GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Loading...","execution_count":null},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train = pd.read_csv(\"../input/conways-reverse-game-of-life-2020/train.csv\")\ntest = pd.read_csv(\"../input/conways-reverse-game-of-life-2020/test.csv\")\nss = pd.read_csv(\"../input/conways-reverse-game-of-life-2020/sample_submission.csv\")","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Defining Class for Data Generation","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"class GameOfLife():\n    \n    def __init__(self, start_array):\n        '''Start array is a 25x25 array of ones and zeros.'''\n        self.array = start_array\n        self.forward_pass()\n        self.padded_array = self.pad(self.array)\n    \n    def pad(self, array):\n        temp = np.zeros((27, 27))\n        for y in range(27):\n            for x in range(27):\n                temp[y, x] = array[(y-1)%25, (x-1)%25]\n        return temp\n    \n    def forward_pass(self):\n        try:\n            self.Y = self.X\n        except:\n            self.Y = self.array\n        \n        self.X = np.zeros((25, 25))\n        \n        for x in range(25):\n            for y in range(25):\n                self.X[y][x] = self.get_surround(self.Y, x, y)\n        return self.X, self.Y\n    \n    def get_surround(self, array, x, y):\n        count = 0 \n        for cx in range(x-1, x+2):\n            for cy in range(y-1, y+2):\n                if x == cx and y == cy:\n                    pass\n                else:\n                    count += array[cy%25][cx%25]\n        if count == 2:\n            return array[y][x]\n        elif count == 3:\n            return 1\n        else:\n            return 0\n    \n    def printxy(self):\n        print('X Values:')\n        print(*self.X.astype('uint8').tolist(), sep='\\n')\n        print('Y Values:')\n        print(*self.Y.astype('uint8').tolist(), sep='\\n')\n        ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"Test for binary values:\")\ncols = [col for col in train.columns if col.startswith('start')]\nx = np.array(train[cols].values).reshape(-1, 25, 25)[3]\nh = GameOfLife(x)\nh.printxy()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"Test visualization and forward_pass:\")\nplt.subplot(231)\n\nfor i in range(3):\n    _, _ = h.forward_pass()\n    plt.subplot(2, 3, i+1, label='X Values:')\n    plt.imshow(h.X)\n    plt.subplot(2, 3, i+4, label='Y Values:')\n    plt.imshow(h.Y)\nplt.show()\nplt.close()","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Defining the Train Generator","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"def train_batch_generator(size, df):\n    batch_size = size//5\n    cols = [x for x in df.columns if x.startswith('start')]\n    delta = df.delta\n    ids = df.id\n    zeroX = np.zeros((27, 27, 1))\n    zeroY = np.zeros(625)\n    all_values = df[cols].values.reshape(-1, 25, 25)\n    \n    for i in range(0, len(ids), batch_size):\n        X = [zeroX for _ in range(5*batch_size)]\n        Y = [zeroY for _ in range(5*batch_size)]\n        \n        for j in range(batch_size):\n            try:\n                item = GameOfLife(all_values[i+j])\n                for k in range(5):\n                    X[5*j+k] = item.pad(item.X).reshape((27, 27, 1))\n                    Y[5*j+k] = item.Y.reshape(625)\n                    item.forward_pass()\n            except Exception as e:\n                print('computing error...', e)\n                yield np.array(X), np.array(Y)\n                return\n        yield np.array(X), np.array(Y)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"Generator test:\")\nimport time\nstart, end = time.time(), time.time()\nfor x, y in train_batch_generator(50, train[:120]):\n    start, end = end, time.time()\n    print(x.shape, end - start)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Defining the Test Generator","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"def test_batch_generator(batch_size, df):\n    cols = [x for x in df.columns if x.startswith('stop')]\n    delta = df.delta\n    ids = df.id\n    zeroX = np.zeros((27, 27, 1))\n    all_values = df[cols].values.reshape(-1, 25, 25)\n    \n    for i in range(0, len(ids), batch_size):\n        X = []\n        \n        for j in range(batch_size):\n            try:\n                item = GameOfLife(all_values[i+j])\n                X.append(item.pad(item.X).reshape((27, 27, 1)))\n                \n            except Exception as e:\n                print('computing error...', e)\n                yield np.array(X)\n                return\n        yield np.array(X)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"Generator test:\")\nimport time\nstart, end = time.time(), time.time()\nfor x in test_batch_generator(50, train[:120]):\n    start, end = end, time.time()\n    print(x.shape, end - start)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Defining the Model for Learning","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.layers import Dense, Conv2D, Input, Flatten, multiply, add, MaxPooling2D, Reshape\nfrom tensorflow.keras.losses import BinaryCrossentropy as BC\nfrom tensorflow.keras.losses import CategoricalCrossentropy as CC\nfrom tensorflow.keras.optimizers import Adam\n\nnp.random.seed(171)\n\ndef custom_loss_func(y_true, y_pred):\n    error = (y_true - y_pred)\n    return error**2\n\ndef predictor(lr=0.01):\n    \n    i = Input((27, 27, 1))\n    x1 = Conv2D(64, (3, 3), activation='relu', padding='valid')(i)\n    x2 = Conv2D(64, (3, 3), activation='relu', padding='valid')(i)\n    x3 = Conv2D(64, (3, 3), activation='sigmoid', padding='valid')(i)\n    x = multiply([x3, add([x1, x2])])\n    \n    x = MaxPooling2D((3, 3))(x)\n    x = Reshape((64, 64, 1))(x)\n    \n    x = Conv2D(128, (9, 9), activation='relu', padding='valid')(x)\n    x1 = Conv2D(32, (9, 9), activation='relu', padding='valid')(x)\n    x2 = Conv2D(32, (9, 9), activation='sigmoid', padding='valid')(x)\n    x = multiply([x1, x2])\n    x = MaxPooling2D((3, 3))(x)\n    x = Flatten()(x)\n    \n    x = Dense(5000, activation='relu')(x)\n    x = Dense(2500, activation='relu')(x)\n    x = Dense(1250, activation='relu')(x)\n    \n    o = Dense(625, activation='sigmoid')(x)\n    \n    model = Model(inputs=i, outputs=o)\n    opt = Adam(lr=lr)\n    #loss = BC(from_logits=False, reduction=tf.keras.losses.Reduction.NONE)\n    loss = CC(from_logits=False, reduction=tf.keras.losses.Reduction.NONE)\n    model.compile(optimizer=opt,\n                  loss=loss,\n                  metrics=['accuracy', 'mse'])\n    \n    return model\n\nmodel = predictor()\nmodel.summary()\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def roundoff(array):\n    #Array should be of shape (625)\n    arr = array.reshape((625))\n    mean = np.quantile(arr, 0.5) if max(arr) < 0.3 else 0.5\n    f = lambda x: x>mean\n    v = np.vectorize(f)\n    return v(arr)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def test_model(model, n=1):\n    y = np.array([[np.random.randint(0, 2) for _ in range(25)] for _ in range(25)])\n    test_acc = GameOfLife(y)\n    \n    for _ in range(5): #Preconditioning\n        _, _ = test_acc.forward_pass()\n        \n    for _ in range(n):\n        x, y = test_acc.forward_pass()\n        pred_y = model.predict(test_acc.pad(x).reshape(-1, 27, 27, 1))\n        plt.subplot(221)\n        plt.subplot(2, 2, 1, label=f'X Values:')\n        plt.imshow(x.reshape((25, 25)))\n        plt.subplot(2, 2, 2, label=f'Y Values:')\n        plt.imshow(y.reshape((25, 25)))\n        plt.subplot(2, 2, 3, label=f'Predicted Y Values:')\n        plt.imshow(pred_y.reshape((25, 25)))\n        plt.subplot(2, 2, 4, label='Rounded Predicted Y Values:')\n        plt.imshow(roundoff(pred_y).reshape((25, 25)))\n        plt.show()\n        print(\"Accuracy:\", sum(sum(np.equal(roundoff(pred_y).reshape((25, 25)), y)))/625)\ntest_model(model, 10)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.model_selection import train_test_split as tts\nfrom tensorflow.keras.callbacks import ReduceLROnPlateau as rlrp\nfrom tensorflow.keras.callbacks import EarlyStopping as es\n\ncall_rlrp = rlrp(monitor='val_accuracy', patience=50, min_lr=1e-8, factor=0.2, verbose=1)\ncall_es = es(monitor='val_mse', patience=100, mode='auto', verbose=1)\n\nprint(\"Begin training:\\t%s\\n\" %time.ctime())\n\nmodel = predictor(.0005)\nstart, end = time.time(), time.time()\nfor n, (x, y) in enumerate(train_batch_generator(250000, train)):\n    print(f\"Training new batch {n+1}...\")\n    tx, vx, ty, vy =tts(x, y, test_size=0.2, shuffle=True, random_state=172)\n    model.fit(tx, ty,\n              validation_data=(vx, vy),\n              batch_size=256, epochs=300,\n              callbacks=[call_rlrp, call_es],\n              verbose=1)\n    model.evaluate(vx, vy, verbose=1)\n    start, end = end, time.time()\n    print(\"#\"*72)\n    print(\"Total time for batch Training: %.2f\" %(end - start))\n    test_model(model, 3)\n    print(\"#\"*72)\nmodel.save('train_model.h5')\nprint(\"\\nEnd training:\\t%s\" %time.ctime())","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"preds = []\ntest_x = []\nfor x in test_batch_generator(10, train[:100]):\n    temp = model.predict(x)\n    try:\n        preds = np.concatenate([preds, temp], axis=0)\n        test_x = np.concatenate([test_x, x], axis=0)\n    except Exception as e:\n        print(e)\n        preds = temp\n        test_x = x\n        \npreds = np.array(preds)\nprint(preds.shape)\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"Test visualization of predictions:\")\n\nfor i in range(0, 96, 3):\n    print(f\"Test data from i={i} to i={i+3}:\")\n    plt.subplot(331)\n    plt.subplot(3, 3, 1, label=f'X {i} Values:')\n    plt.imshow(test_x[i][1:26, 1:26].reshape((25, 25)))\n    plt.subplot(3, 3, 4, label=f'Y {i} Values:')\n    plt.imshow(preds[i].reshape((25, 25)))\n    plt.subplot(3, 3, 7, label=f'Y {i} Values:')\n    plt.imshow(roundoff(preds[i]).reshape((25, 25)))\n    \n    plt.subplot(3, 3, 2, label=f'X {i} Values:')\n    plt.imshow(test_x[i+1][1:26, 1:26].reshape((25, 25)))\n    plt.subplot(3, 3, 5, label=f'Y {i} Values:')\n    plt.imshow(preds[i+1].reshape((25, 25)))\n    plt.subplot(3, 3, 8, label=f'Y {i} Values:')\n    plt.imshow(roundoff(preds[i+1]).reshape((25, 25)))\n    \n    plt.subplot(3, 3, 3, label=f'X {i} Values:')\n    plt.imshow(test_x[i+2][1:26, 1:26].reshape((25, 25)))\n    plt.subplot(3, 3, 6, label=f'Y {i} Values:')\n    plt.imshow(preds[i+2].reshape((25, 25)))\n    plt.subplot(3, 3, 9, label=f'Y {i} Values:')\n    plt.imshow(roundoff(preds[i+2]).reshape((25, 25)))\n    \n    plt.show()\n    plt.close()","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Predictions","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"preds = []\nprint(\"Compute for all delta >= 1...\")\nfor x in test_batch_generator(10000, test):\n    temp = model.predict(x, verbose=2)\n    try:\n        preds = np.concatenate([preds, temp], axis=0)\n    except Exception as e:\n        print(e)\n        preds = temp\n\n\npreds = np.array(preds)\nprint(preds.shape)\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(2, 6):\n    print(f\"Computing for all delta >= {i}...\")\n    indices = test.index[test.delta >= i].tolist()\n    preds[indices] = [roundoff(x) for x in model.predict(np.array(\n        [h.pad(array.reshape((25, 25))) for array in preds[indices]])\n                                                        )]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"ids = test.id\n\nsubmission = pd.DataFrame([roundoff(x) for x in preds], index=ids)\nsubmission.columns = ['start_'+str(c) for c in range(625)]\nsubmission = submission.astype('uint32')\nsubmission.head()\nsubmission.to_csv(\"submission.csv\", index=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"submission.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"submission.describe()","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}