Logistic Regression with L2 regularization.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"Logistic Regression with L2 regularization.ipynb","provenance":[],"mount_file_id":"1kTvJchftU_Ur2yLijovZTk31BpWanz61","authorship_tag":"ABX9TyOGIyUyCksOEys6lgN/dAuX"},"kernelspec":{"name":"python3","display_name":"Python 3"}},"cells":[{"cell_type":"markdown","metadata":{"id":"pMdeUxn0XC4V","colab_type":"text"},"source":["Logistic Regression with L2 regularization"]},{"cell_type":"code","metadata":{"id":"4QGND9XCWFzt","colab_type":"code","colab":{}},"source":["pip install turicreate"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"5gkf_CrLW_a1","colab_type":"text"},"source":["Fire up Turi Create"]},{"cell_type":"code","metadata":{"id":"1qkCD76lWWL3","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849298146,"user_tz":-330,"elapsed":3790,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["from __future__ import division\n","import turicreate as tc"],"execution_count":15,"outputs":[]},{"cell_type":"code","metadata":{"id":"-q83muWQWX4h","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":458},"executionInfo":{"status":"ok","timestamp":1599849298150,"user_tz":-330,"elapsed":3767,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"b8e1597a-96b0-4a0e-c270-e63c23b18512"},"source":["products = tc.SFrame('/content/drive/My Drive/amazon_baby_subset.sframe')\n","products"],"execution_count":16,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n","    <tr>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">name</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">review</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">rating</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">sentiment</th>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Stop Pacifier Sucking<br>without tears with ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">All of my kids have cried<br>non-stop when I tried to ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Nature&#x27;s Lullabies Second<br>Year Sticker Calendar ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">We wanted to get<br>something to keep track ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Nature&#x27;s Lullabies Second<br>Year Sticker Calendar ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">My daughter had her 1st<br>baby over a year ago. ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Lamaze Peekaboo, I Love<br>You ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">One of baby&#x27;s first and<br>favorite books, and i ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">4.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">SoftPlay Peek-A-Boo<br>Where&#x27;s Elmo A Childr ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Very cute interactive<br>book! My son loves this ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Our Baby Girl Memory Book</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Beautiful book, I love it<br>to record cherished t ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Hunnt&amp;reg; Falling<br>Flowers and Birds Kids ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Try this out for a spring<br>project !Easy ,fun and ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Blessed By Pope Benedict<br>XVI Divine Mercy Full ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">very nice Divine Mercy<br>Pendant of Jesus now on ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Cloth Diaper Pins<br>Stainless Steel ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">We bought the pins as my<br>6 year old Autistic son ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">4.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">Cloth Diaper Pins<br>Stainless Steel ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">It has been many years<br>since we needed diaper ...</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.0</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n","    </tr>\n","</table>\n","[53072 rows x 4 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n","</div>"],"text/plain":["Columns:\n","\tname\tstr\n","\treview\tstr\n","\trating\tfloat\n","\tsentiment\tint\n","\n","Rows: 53072\n","\n","Data:\n","+-------------------------------+-------------------------------+--------+-----------+\n","|              name             |             review            | rating | sentiment |\n","+-------------------------------+-------------------------------+--------+-----------+\n","| Stop Pacifier Sucking with... | All of my kids have cried ... |  5.0   |     1     |\n","| Nature's Lullabies Second ... | We wanted to get something... |  5.0   |     1     |\n","| Nature's Lullabies Second ... | My daughter had her 1st ba... |  5.0   |     1     |\n","|  Lamaze Peekaboo, I Love You  | One of baby's first and fa... |  4.0   |     1     |\n","| SoftPlay Peek-A-Boo Where'... | Very cute interactive book... |  5.0   |     1     |\n","|   Our Baby Girl Memory Book   | Beautiful book, I love it ... |  5.0   |     1     |\n","| Hunnt&reg; Falling Flowers... | Try this out for a spring ... |  5.0   |     1     |\n","| Blessed By Pope Benedict X... | very nice Divine Mercy Pen... |  5.0   |     1     |\n","| Cloth Diaper Pins Stainles... | We bought the pins as my 6... |  4.0   |     1     |\n","| Cloth Diaper Pins Stainles... | It has been many years sin... |  5.0   |     1     |\n","+-------------------------------+-------------------------------+--------+-----------+\n","[53072 rows x 4 columns]\n","Note: Only the head of the SFrame is printed.\n","You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."]},"metadata":{"tags":[]},"execution_count":16}]},{"cell_type":"markdown","metadata":{"id":"TRYgFQAUXE8T","colab_type":"text"},"source":["Load and process review dataset"]},{"cell_type":"code","metadata":{"id":"0XnpBix8Wtzf","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849300152,"user_tz":-330,"elapsed":5746,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["# The same feature processing (same as the previous assignments)\n","# ---------------------------------------------------------------\n","import json\n","with open('/content/drive/My Drive/important_words.json', 'r') as f: # Reads the list of most frequent words\n","    important_words = json.load(f)\n","important_words = [str(s) for s in important_words]\n","\n","import string \n","def remove_punctuation(text):\n","  translator = text.maketrans('', '', string.punctuation)\n","  text = text.translate(translator)\n","  return text\n","\n","# Remove punctuation.\n","products['review_clean'] = products['review'].apply(remove_punctuation)\n","\n","# Split out the words into individual columns\n","for word in important_words:\n","    products[word] = products['review_clean'].apply(lambda s : s.split().count(word))"],"execution_count":17,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"Oror9DJCXTz9","colab_type":"text"},"source":["Train-Validation split"]},{"cell_type":"code","metadata":{"id":"Z94Tvav0XJOi","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":54},"executionInfo":{"status":"ok","timestamp":1599849377177,"user_tz":-330,"elapsed":82754,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"679b8ecb-dff9-48f2-aade-4eb22584d2d5"},"source":["train_data, validation_data = products.random_split(.8, seed=2)\n","\n","print('Training set   : %d data points' % len(train_data))\n","print('Validation set : %d data points' % len(validation_data))"],"execution_count":18,"outputs":[{"output_type":"stream","text":["Training set   : 42361 data points\n","Validation set : 10711 data points\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"jLvqC7hsYOua","colab_type":"text"},"source":["Convert SFrame to NumPy array"]},{"cell_type":"code","metadata":{"id":"-ifBEShOXvAg","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849377199,"user_tz":-330,"elapsed":82755,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["import numpy as np\n","\n","def get_numpy_data(data_sframe, features, label):\n","    data_sframe['intercept'] = 1\n","    features = ['intercept'] + features\n","    features_sframe = data_sframe[features]\n","    feature_matrix = features_sframe.to_numpy()\n","    label_sarray = data_sframe[label]\n","    label_array = label_sarray.to_numpy()\n","    return(feature_matrix, label_array)"],"execution_count":19,"outputs":[]},{"cell_type":"code","metadata":{"id":"NsD554gDXyVf","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849382046,"user_tz":-330,"elapsed":87578,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["feature_matrix_train, sentiment_train = get_numpy_data(train_data, important_words, 'sentiment')\n","feature_matrix_valid, sentiment_valid = get_numpy_data(validation_data, important_words, 'sentiment') "],"execution_count":20,"outputs":[]},{"cell_type":"code","metadata":{"id":"uzUV8IA4YLBZ","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849382082,"user_tz":-330,"elapsed":87597,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["'''\n","produces probablistic estimate for P(y_i = +1 | x_i, w).\n","estimate ranges between 0 and 1.\n","'''\n","def predict_probability(feature_matrix, coefficients):\n","    predictions = []\n","    # Take dot product of feature_matrix and coefficients  \n","    scores  = np.dot(feature_matrix,coefficients)\n","    # Compute P(y_i = +1 | x_i, w) using the link function\n","    for value in scores:\n","      pred = 1/(1 + np.exp(-value))\n","      predictions.append(pred)\n","    \n","    # return predictions\n","    return predictions"],"execution_count":21,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"BHOw4HQoZZiJ","colab_type":"text"},"source":["Adding L2 penalty"]},{"cell_type":"code","metadata":{"id":"u_dd8tiDZaWu","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849382123,"user_tz":-330,"elapsed":87611,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["def feature_derivative_with_L2(errors, feature, coefficient, l2_penalty, feature_is_constant): \n","    \n","    # Compute the dot product of errors and feature\n","    derivative = np.dot(errors, feature)\n","\n","    # add L2 penalty term for any feature that isn't the intercept.\n","    if not feature_is_constant: \n","        derivative -= (2 * l2_penalty * coefficient)\n","        \n","    return derivative\n","\n"],"execution_count":23,"outputs":[]},{"cell_type":"code","metadata":{"id":"IpX3-LQhcxxI","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599849442720,"user_tz":-330,"elapsed":1052,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["def compute_log_likelihood_with_L2(feature_matrix, sentiment, coefficients, l2_penalty):\n","    indicator = (sentiment==+1)\n","    scores = np.dot(feature_matrix, coefficients)\n","    \n","    lp = np.sum((indicator-1)*scores - np.log(1. + np.exp(-scores))) - l2_penalty*np.sum(coefficients[1:]**2)\n","    \n","    return lp\n","\n"],"execution_count":24,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"Elfwx3SodUVX","colab_type":"text"},"source":["Logisic Regression With L2 penalty"]},{"cell_type":"code","metadata":{"id":"o-i7P9XJdTMc","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599850037687,"user_tz":-330,"elapsed":1262,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["def logistic_regression_with_L2(feature_matrix, sentiment, initial_coefficients, step_size, l2_penalty, max_iter):\n","    coefficients = np.array(initial_coefficients) # make sure it's a numpy array\n","    for itr in range(max_iter):\n","        # Predict P(y_i = +1|x_i,w) using your predict_probability() function\n","        predictions = predict_probability(feature_matrix,coefficients)\n","        \n","        # Compute indicator value for (y_i = +1)\n","        indicator = (sentiment==+1)\n","        \n","        # Compute the errors as indicator - predictions\n","        errors = indicator - predictions\n","        for j in range(len(coefficients)): # loop over each coefficient\n","            is_intercept = (j == 0)\n","            # Recall that feature_matrix[:,j] is the feature column associated with coefficients[j].\n","            # Compute the derivative for coefficients[j]. Save it in a variable called derivative\n","            derivative = feature_derivative_with_L2(errors,feature_matrix[:,j],coefficients[j],l2_penalty,is_intercept)\n","            \n","            # add the step size times the derivative to the current coefficient\n","            coefficients[j] += step_size * derivative\n","        \n","        # Checking whether log likelihood is increasing\n","        if itr <= 15 or (itr <= 100 and itr % 10 == 0) or (itr <= 1000 and itr % 100 == 0) \\\n","        or (itr <= 10000 and itr % 1000 == 0) or itr % 10000 == 0:\n","            lp = compute_log_likelihood_with_L2(feature_matrix, sentiment, coefficients, l2_penalty)\n","            print('iteration %*d: log likelihood of observed labels = %.8f' % \\\n","                (int(np.ceil(np.log10(max_iter))), itr, lp))\n","    return coefficients"],"execution_count":27,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"sIPxKL2Re6jN","colab_type":"text"},"source":["Explore effects of L2 regularization"]},{"cell_type":"code","metadata":{"id":"rzR029SWe4w8","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850151748,"user_tz":-330,"elapsed":112795,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"25c8fbf1-806b-4e31-dc56-33c7c0cd7d02"},"source":["# run with L2 = 0\n","coefficients_0_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                     initial_coefficients=np.zeros(194),\n","                                                     step_size=5e-6, l2_penalty=0, max_iter=501)"],"execution_count":28,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29179.39138303\n","iteration   1: log likelihood of observed labels = -29003.71259047\n","iteration   2: log likelihood of observed labels = -28834.66187288\n","iteration   3: log likelihood of observed labels = -28671.70781507\n","iteration   4: log likelihood of observed labels = -28514.43078198\n","iteration   5: log likelihood of observed labels = -28362.48344665\n","iteration   6: log likelihood of observed labels = -28215.56713122\n","iteration   7: log likelihood of observed labels = -28073.41743783\n","iteration   8: log likelihood of observed labels = -27935.79536396\n","iteration   9: log likelihood of observed labels = -27802.48168669\n","iteration  10: log likelihood of observed labels = -27673.27331484\n","iteration  11: log likelihood of observed labels = -27547.98083656\n","iteration  12: log likelihood of observed labels = -27426.42679977\n","iteration  13: log likelihood of observed labels = -27308.44444728\n","iteration  14: log likelihood of observed labels = -27193.87673876\n","iteration  15: log likelihood of observed labels = -27082.57555831\n","iteration  20: log likelihood of observed labels = -26570.43059938\n","iteration  30: log likelihood of observed labels = -25725.48742389\n","iteration  40: log likelihood of observed labels = -25055.53326910\n","iteration  50: log likelihood of observed labels = -24509.63590026\n","iteration  60: log likelihood of observed labels = -24054.97906083\n","iteration  70: log likelihood of observed labels = -23669.51640848\n","iteration  80: log likelihood of observed labels = -23337.89167628\n","iteration  90: log likelihood of observed labels = -23049.07066021\n","iteration 100: log likelihood of observed labels = -22794.90974921\n","iteration 200: log likelihood of observed labels = -21283.29527353\n","iteration 300: log likelihood of observed labels = -20570.97485473\n","iteration 400: log likelihood of observed labels = -20152.21466944\n","iteration 500: log likelihood of observed labels = -19876.62333410\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"hf_p-33Kfq95","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850292769,"user_tz":-330,"elapsed":111432,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"7ff34ed8-65af-4bed-ca54-45542c602b82"},"source":["# run with L2 = 4\n","coefficients_4_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                      initial_coefficients=np.zeros(194),\n","                                                      step_size=5e-6, l2_penalty=4, max_iter=501)"],"execution_count":29,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29179.39508175\n","iteration   1: log likelihood of observed labels = -29003.73417180\n","iteration   2: log likelihood of observed labels = -28834.71441858\n","iteration   3: log likelihood of observed labels = -28671.80345068\n","iteration   4: log likelihood of observed labels = -28514.58077957\n","iteration   5: log likelihood of observed labels = -28362.69830317\n","iteration   6: log likelihood of observed labels = -28215.85663259\n","iteration   7: log likelihood of observed labels = -28073.79071393\n","iteration   8: log likelihood of observed labels = -27936.26093762\n","iteration   9: log likelihood of observed labels = -27803.04751805\n","iteration  10: log likelihood of observed labels = -27673.94684207\n","iteration  11: log likelihood of observed labels = -27548.76901327\n","iteration  12: log likelihood of observed labels = -27427.33612958\n","iteration  13: log likelihood of observed labels = -27309.48101569\n","iteration  14: log likelihood of observed labels = -27195.04624253\n","iteration  15: log likelihood of observed labels = -27083.88333261\n","iteration  20: log likelihood of observed labels = -26572.49874392\n","iteration  30: log likelihood of observed labels = -25729.32604153\n","iteration  40: log likelihood of observed labels = -25061.34245801\n","iteration  50: log likelihood of observed labels = -24517.52091982\n","iteration  60: log likelihood of observed labels = -24064.99093939\n","iteration  70: log likelihood of observed labels = -23681.67373669\n","iteration  80: log likelihood of observed labels = -23352.19298741\n","iteration  90: log likelihood of observed labels = -23065.50180166\n","iteration 100: log likelihood of observed labels = -22813.44844580\n","iteration 200: log likelihood of observed labels = -21321.14164794\n","iteration 300: log likelihood of observed labels = -20624.98634439\n","iteration 400: log likelihood of observed labels = -20219.92048845\n","iteration 500: log likelihood of observed labels = -19956.11341777\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"zObniIYEgljp","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850421351,"user_tz":-330,"elapsed":113428,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"73ef5fbc-2535-4bec-e90c-00b81b832fc5"},"source":["# run with L2 = 10\n","coefficients_10_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                      initial_coefficients=np.zeros(194),\n","                                                      step_size=5e-6, l2_penalty=10, max_iter=501)"],"execution_count":30,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29179.40062984\n","iteration   1: log likelihood of observed labels = -29003.76654163\n","iteration   2: log likelihood of observed labels = -28834.79322654\n","iteration   3: log likelihood of observed labels = -28671.94687528\n","iteration   4: log likelihood of observed labels = -28514.80571589\n","iteration   5: log likelihood of observed labels = -28363.02048079\n","iteration   6: log likelihood of observed labels = -28216.29071186\n","iteration   7: log likelihood of observed labels = -28074.35036891\n","iteration   8: log likelihood of observed labels = -27936.95892966\n","iteration   9: log likelihood of observed labels = -27803.89576265\n","iteration  10: log likelihood of observed labels = -27674.95647005\n","iteration  11: log likelihood of observed labels = -27549.95042714\n","iteration  12: log likelihood of observed labels = -27428.69905549\n","iteration  13: log likelihood of observed labels = -27311.03455140\n","iteration  14: log likelihood of observed labels = -27196.79890162\n","iteration  15: log likelihood of observed labels = -27085.84308528\n","iteration  20: log likelihood of observed labels = -26575.59697506\n","iteration  30: log likelihood of observed labels = -25735.07304608\n","iteration  40: log likelihood of observed labels = -25070.03447306\n","iteration  50: log likelihood of observed labels = -24529.31188025\n","iteration  60: log likelihood of observed labels = -24079.95349572\n","iteration  70: log likelihood of observed labels = -23699.83199186\n","iteration  80: log likelihood of observed labels = -23373.54108747\n","iteration  90: log likelihood of observed labels = -23090.01500055\n","iteration 100: log likelihood of observed labels = -22841.08995135\n","iteration 200: log likelihood of observed labels = -21377.25595328\n","iteration 300: log likelihood of observed labels = -20704.63995428\n","iteration 400: log likelihood of observed labels = -20319.25685307\n","iteration 500: log likelihood of observed labels = -20072.16321721\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"RBbl_6cBgo04","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850639503,"user_tz":-330,"elapsed":111646,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"279c13b2-5a17-4e04-ed6b-6da5fd2a340e"},"source":["# run with L2 = 1e2\n","coefficients_1e2_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                       initial_coefficients=np.zeros(194),\n","                                                       step_size=5e-6, l2_penalty=1e2, max_iter=501)"],"execution_count":31,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29179.48385120\n","iteration   1: log likelihood of observed labels = -29004.25177457\n","iteration   2: log likelihood of observed labels = -28835.97382190\n","iteration   3: log likelihood of observed labels = -28674.09410083\n","iteration   4: log likelihood of observed labels = -28518.17112932\n","iteration   5: log likelihood of observed labels = -28367.83774654\n","iteration   6: log likelihood of observed labels = -28222.77708939\n","iteration   7: log likelihood of observed labels = -28082.70799392\n","iteration   8: log likelihood of observed labels = -27947.37595368\n","iteration   9: log likelihood of observed labels = -27816.54738615\n","iteration  10: log likelihood of observed labels = -27690.00588850\n","iteration  11: log likelihood of observed labels = -27567.54970126\n","iteration  12: log likelihood of observed labels = -27448.98991327\n","iteration  13: log likelihood of observed labels = -27334.14912742\n","iteration  14: log likelihood of observed labels = -27222.86041863\n","iteration  15: log likelihood of observed labels = -27114.96648229\n","iteration  20: log likelihood of observed labels = -26621.50201299\n","iteration  30: log likelihood of observed labels = -25819.72803950\n","iteration  40: log likelihood of observed labels = -25197.34035501\n","iteration  50: log likelihood of observed labels = -24701.03698195\n","iteration  60: log likelihood of observed labels = -24296.66378580\n","iteration  70: log likelihood of observed labels = -23961.38842316\n","iteration  80: log likelihood of observed labels = -23679.38088853\n","iteration  90: log likelihood of observed labels = -23439.31824267\n","iteration 100: log likelihood of observed labels = -23232.88192018\n","iteration 200: log likelihood of observed labels = -22133.50726528\n","iteration 300: log likelihood of observed labels = -21730.03957488\n","iteration 400: log likelihood of observed labels = -21545.87572145\n","iteration 500: log likelihood of observed labels = -21451.95551390\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"3CPvOnkWgplH","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850790850,"user_tz":-330,"elapsed":111507,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"5ba56529-494b-48f0-8188-e2919b2b9efd"},"source":["# run with L2 = 1e3\n","coefficients_1e3_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                       initial_coefficients=np.zeros(194),\n","                                                       step_size=5e-6, l2_penalty=1e3, max_iter=501)"],"execution_count":32,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29180.31606471\n","iteration   1: log likelihood of observed labels = -29009.07176112\n","iteration   2: log likelihood of observed labels = -28847.62378912\n","iteration   3: log likelihood of observed labels = -28695.14439397\n","iteration   4: log likelihood of observed labels = -28550.95060743\n","iteration   5: log likelihood of observed labels = -28414.45771129\n","iteration   6: log likelihood of observed labels = -28285.15124375\n","iteration   7: log likelihood of observed labels = -28162.56976044\n","iteration   8: log likelihood of observed labels = -28046.29387744\n","iteration   9: log likelihood of observed labels = -27935.93902900\n","iteration  10: log likelihood of observed labels = -27831.15045502\n","iteration  11: log likelihood of observed labels = -27731.59955260\n","iteration  12: log likelihood of observed labels = -27636.98108219\n","iteration  13: log likelihood of observed labels = -27547.01092670\n","iteration  14: log likelihood of observed labels = -27461.42422295\n","iteration  15: log likelihood of observed labels = -27379.97375625\n","iteration  20: log likelihood of observed labels = -27027.18208317\n","iteration  30: log likelihood of observed labels = -26527.22737267\n","iteration  40: log likelihood of observed labels = -26206.59048765\n","iteration  50: log likelihood of observed labels = -25995.96903148\n","iteration  60: log likelihood of observed labels = -25854.95710284\n","iteration  70: log likelihood of observed labels = -25759.08109950\n","iteration  80: log likelihood of observed labels = -25693.05688014\n","iteration  90: log likelihood of observed labels = -25647.09929349\n","iteration 100: log likelihood of observed labels = -25614.81468705\n","iteration 200: log likelihood of observed labels = -25536.20998919\n","iteration 300: log likelihood of observed labels = -25532.57691220\n","iteration 400: log likelihood of observed labels = -25532.35543765\n","iteration 500: log likelihood of observed labels = -25532.33970049\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"CyJrim2XgrYQ","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":550},"executionInfo":{"status":"ok","timestamp":1599850902177,"user_tz":-330,"elapsed":222259,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"ebc8b0db-3af8-4c5d-ef87-fbfe64877837"},"source":["# run with L2 = 1e5\n","coefficients_1e5_penalty = logistic_regression_with_L2(feature_matrix_train, sentiment_train,\n","                                                       initial_coefficients=np.zeros(194),\n","                                                       step_size=5e-6, l2_penalty=1e5, max_iter=501)"],"execution_count":33,"outputs":[{"output_type":"stream","text":["iteration   0: log likelihood of observed labels = -29271.85955115\n","iteration   1: log likelihood of observed labels = -29271.71006589\n","iteration   2: log likelihood of observed labels = -29271.65738833\n","iteration   3: log likelihood of observed labels = -29271.61189923\n","iteration   4: log likelihood of observed labels = -29271.57079975\n","iteration   5: log likelihood of observed labels = -29271.53358505\n","iteration   6: log likelihood of observed labels = -29271.49988440\n","iteration   7: log likelihood of observed labels = -29271.46936584\n","iteration   8: log likelihood of observed labels = -29271.44172890\n","iteration   9: log likelihood of observed labels = -29271.41670149\n","iteration  10: log likelihood of observed labels = -29271.39403722\n","iteration  11: log likelihood of observed labels = -29271.37351294\n","iteration  12: log likelihood of observed labels = -29271.35492661\n","iteration  13: log likelihood of observed labels = -29271.33809523\n","iteration  14: log likelihood of observed labels = -29271.32285309\n","iteration  15: log likelihood of observed labels = -29271.30905015\n","iteration  20: log likelihood of observed labels = -29271.25729150\n","iteration  30: log likelihood of observed labels = -29271.20657205\n","iteration  40: log likelihood of observed labels = -29271.18775997\n","iteration  50: log likelihood of observed labels = -29271.18078247\n","iteration  60: log likelihood of observed labels = -29271.17819447\n","iteration  70: log likelihood of observed labels = -29271.17723457\n","iteration  80: log likelihood of observed labels = -29271.17687853\n","iteration  90: log likelihood of observed labels = -29271.17674648\n","iteration 100: log likelihood of observed labels = -29271.17669750\n","iteration 200: log likelihood of observed labels = -29271.17666862\n","iteration 300: log likelihood of observed labels = -29271.17666862\n","iteration 400: log likelihood of observed labels = -29271.17666862\n","iteration 500: log likelihood of observed labels = -29271.17666862\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"-TyR0yVFjWX4","colab_type":"text"},"source":["Compare coefficients"]},{"cell_type":"code","metadata":{"id":"jT_y7Ob7jYBq","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599851049201,"user_tz":-330,"elapsed":1049,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["table = turicreate.SFrame({'word': ['(intercept)'] + important_words})\n","def add_coefficients_to_table(coefficients, column_name):\n","    table[column_name] = coefficients\n","    return table"],"execution_count":34,"outputs":[]},{"cell_type":"code","metadata":{"id":"UGD5JiO0jaEe","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":499},"executionInfo":{"status":"ok","timestamp":1599851057024,"user_tz":-330,"elapsed":1013,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"4091c3f0-9421-4499-e04f-8d8e93ce80bb"},"source":["add_coefficients_to_table(coefficients_0_penalty, 'coefficients [L2=0]')\n","add_coefficients_to_table(coefficients_4_penalty, 'coefficients [L2=4]')\n","add_coefficients_to_table(coefficients_10_penalty, 'coefficients [L2=10]')\n","add_coefficients_to_table(coefficients_1e2_penalty, 'coefficients [L2=1e2]')\n","add_coefficients_to_table(coefficients_1e3_penalty, 'coefficients [L2=1e3]')\n","add_coefficients_to_table(coefficients_1e5_penalty, 'coefficients [L2=1e5]')"],"execution_count":36,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n","    <tr>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">word</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=0]</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=4]</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=10]</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=1e2]</th>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">(intercept)</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.06374213522751736</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.06314308770737208</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.06225594377004484</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.05043836519527793</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">baby</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0740730059216435</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.07399385414054363</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.07387735348036835</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.07236036022182919</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">one</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.012752505778363732</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.01249497044812448</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.012115252953356623</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.00724728331869879</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">great</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.8016249897781762</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.7968969330033088</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.7899351472214042</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.7014250736748336</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">love</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1.0585539820695435</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1.050855680993631</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1.0395285158465861</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.8966441331010778</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">use</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.00010415219124811945</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0001628576561772408</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0005557109757555857</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.005481032679461016</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">would</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.28702144353429704</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.2860272029746515</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.28456403556187243</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.2659927462255519</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">like</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.0033844739929333313</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.003442085770451767</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.0035272944496608154</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.004634992832527627</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">easy</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.9845588198731924</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.9776001497817036</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.9673618366314701</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.8382448329554505</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">little</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.5244194563643343</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.5213847261068162</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.5169173924914707</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.4602345779024191</td>\n","    </tr>\n","</table>\n","<table frame=\"box\" rules=\"cols\">\n","    <tr>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=1e3]</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=1e5]</th>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">5.3867532695823166e-05</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.011361751184438747</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.05975168883639244</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0017841492162980102</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.008760917620035742</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.0018268556802280068</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.37601171422167073</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.008949560497355635</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.4183536441339269</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.009041737297695288</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.01732648744612427</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.000417863934615973</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.18866242204876657</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.008127027099004498</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.007043075843533989</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">-0.0008266502700305295</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.40190397136319134</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.00880768121210408</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.2512206539585926</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.005940513640376537</td>\n","    </tr>\n","</table>\n","[194 rows x 7 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n","</div>"],"text/plain":["Columns:\n","\tword\tstr\n","\tcoefficients [L2=0]\tfloat\n","\tcoefficients [L2=4]\tfloat\n","\tcoefficients [L2=10]\tfloat\n","\tcoefficients [L2=1e2]\tfloat\n","\tcoefficients [L2=1e3]\tfloat\n","\tcoefficients [L2=1e5]\tfloat\n","\n","Rows: 194\n","\n","Data:\n","+-------------+-------------------------+-----------------------+\n","|     word    |   coefficients [L2=0]   |  coefficients [L2=4]  |\n","+-------------+-------------------------+-----------------------+\n","| (intercept) |   -0.06374213522751736  |  -0.06314308770737208 |\n","|     baby    |    0.0740730059216435   |  0.07399385414054363  |\n","|     one     |   0.012752505778363732  |  0.01249497044812448  |\n","|    great    |    0.8016249897781762   |   0.7968969330033088  |\n","|     love    |    1.0585539820695435   |   1.050855680993631   |\n","|     use     | -0.00010415219124811945 | 0.0001628576561772408 |\n","|    would    |   -0.28702144353429704  |  -0.2860272029746515  |\n","|     like    |  -0.0033844739929333313 | -0.003442085770451767 |\n","|     easy    |    0.9845588198731924   |   0.9776001497817036  |\n","|    little   |    0.5244194563643343   |   0.5213847261068162  |\n","+-------------+-------------------------+-----------------------+\n","+------------------------+-----------------------+------------------------+\n","|  coefficients [L2=10]  | coefficients [L2=1e2] | coefficients [L2=1e3]  |\n","+------------------------+-----------------------+------------------------+\n","|  -0.06225594377004484  |  -0.05043836519527793 | 5.3867532695823166e-05 |\n","|  0.07387735348036835   |  0.07236036022182919  |  0.05975168883639244   |\n","|  0.012115252953356623  |  0.00724728331869879  | -0.008760917620035742  |\n","|   0.7899351472214042   |   0.7014250736748336  |  0.37601171422167073   |\n","|   1.0395285158465861   |   0.8966441331010778  |   0.4183536441339269   |\n","| 0.0005557109757555857  |  0.005481032679461016 |  0.01732648744612427   |\n","|  -0.28456403556187243  |  -0.2659927462255519  |  -0.18866242204876657  |\n","| -0.0035272944496608154 | -0.004634992832527627 | -0.007043075843533989  |\n","|   0.9673618366314701   |   0.8382448329554505  |  0.40190397136319134   |\n","|   0.5169173924914707   |   0.4602345779024191  |   0.2512206539585926   |\n","+------------------------+-----------------------+------------------------+\n","+------------------------+\n","| coefficients [L2=1e5]  |\n","+------------------------+\n","|  0.011361751184438747  |\n","| 0.0017841492162980102  |\n","| -0.0018268556802280068 |\n","|  0.008949560497355635  |\n","|  0.009041737297695288  |\n","|  0.000417863934615973  |\n","| -0.008127027099004498  |\n","| -0.0008266502700305295 |\n","|  0.00880768121210408   |\n","|  0.005940513640376537  |\n","+------------------------+\n","[194 rows x 7 columns]\n","Note: Only the head of the SFrame is printed.\n","You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."]},"metadata":{"tags":[]},"execution_count":36}]},{"cell_type":"code","metadata":{"id":"6P7-JelKjd7-","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":284},"executionInfo":{"status":"ok","timestamp":1599851209737,"user_tz":-330,"elapsed":930,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"7dc442ee-a3b3-42e8-bdf7-e0f98499fa8a"},"source":["table[['word','coefficients [L2=0]']].sort('coefficients [L2=0]',ascending = False)"],"execution_count":38,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n","    <tr>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">word</th>\n","        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">coefficients [L2=0]</th>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">love</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1.0585539820695435</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">loves</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1.0524840477983537</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">easy</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.9845588198731924</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">perfect</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.8356932076383464</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">great</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.8016249897781762</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">happy</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.5573952023634463</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">best</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.535033889771777</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">little</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.5244194563643343</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">fits</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.4876522532652972</td>\n","    </tr>\n","    <tr>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">well</td>\n","        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.4538664867675504</td>\n","    </tr>\n","</table>\n","[194 rows x 2 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n","</div>"],"text/plain":["Columns:\n","\tword\tstr\n","\tcoefficients [L2=0]\tfloat\n","\n","Rows: 194\n","\n","Data:\n","+---------+---------------------+\n","|   word  | coefficients [L2=0] |\n","+---------+---------------------+\n","|   love  |  1.0585539820695435 |\n","|  loves  |  1.0524840477983537 |\n","|   easy  |  0.9845588198731924 |\n","| perfect |  0.8356932076383464 |\n","|  great  |  0.8016249897781762 |\n","|  happy  |  0.5573952023634463 |\n","|   best  |  0.535033889771777  |\n","|  little |  0.5244194563643343 |\n","|   fits  |  0.4876522532652972 |\n","|   well  |  0.4538664867675504 |\n","+---------+---------------------+\n","[194 rows x 2 columns]\n","Note: Only the head of the SFrame is printed.\n","You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."]},"metadata":{"tags":[]},"execution_count":38}]},{"cell_type":"code","metadata":{"id":"ve2Azk-DkEzL","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599851294744,"user_tz":-330,"elapsed":3158,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["positive_words = table.topk('coefficients [L2=0]',k=5,reverse=False)['word']\n"],"execution_count":39,"outputs":[]},{"cell_type":"code","metadata":{"id":"RNobkjvYkZZU","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":72},"executionInfo":{"status":"ok","timestamp":1599851311791,"user_tz":-330,"elapsed":2002,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"1d93da05-8338-440f-8e43-94f4af0d872d"},"source":["positive_words"],"execution_count":40,"outputs":[{"output_type":"execute_result","data":{"text/plain":["dtype: str\n","Rows: 5\n","['love', 'loves', 'easy', 'perfect', 'great']"]},"metadata":{"tags":[]},"execution_count":40}]},{"cell_type":"code","metadata":{"id":"Z83umXAAkffy","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":72},"executionInfo":{"status":"ok","timestamp":1599851346238,"user_tz":-330,"elapsed":3171,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"5fa56e1f-ece3-46f1-ab51-22a3f20ec5bf"},"source":["negative_words = table.topk('coefficients [L2=0]',k=5,reverse=True)['word']\n","negative_words"],"execution_count":41,"outputs":[{"output_type":"execute_result","data":{"text/plain":["dtype: str\n","Rows: 5\n","['disappointed', 'money', 'return', 'waste', 'returned']"]},"metadata":{"tags":[]},"execution_count":41}]},{"cell_type":"markdown","metadata":{"id":"p8WJeAnOkn4B","colab_type":"text"},"source":["Let us observe the effect of increasing L2 penalty on the 10 words just selected. We provide you with a utility function to plot the coefficient path."]},{"cell_type":"code","metadata":{"id":"9C5u0aa3kk8m","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599851390029,"user_tz":-330,"elapsed":989,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["import matplotlib.pyplot as plt\n","%matplotlib inline\n","plt.rcParams['figure.figsize'] = 10, 6\n","\n","def make_coefficient_plot(table, positive_words, negative_words, l2_penalty_list):\n","    cmap_positive = plt.get_cmap('Reds')\n","    cmap_negative = plt.get_cmap('Blues')\n","    \n","    xx = l2_penalty_list\n","    plt.plot(xx, [0.]*len(xx), '--', lw=1, color='k')\n","    \n","    table_positive_words = table.filter_by(column_name='word', values=positive_words)\n","    table_negative_words = table.filter_by(column_name='word', values=negative_words)\n","    del table_positive_words['word']\n","    del table_negative_words['word']\n","    \n","    for i in range(len(positive_words)):\n","        color = cmap_positive(0.8*((i+1)/(len(positive_words)*1.2)+0.15))\n","        plt.plot(xx, table_positive_words[i:i+1].to_numpy().flatten(),\n","                 '-', label=positive_words[i], linewidth=4.0, color=color)\n","        \n","    for i in range(len(negative_words)):\n","        color = cmap_negative(0.8*((i+1)/(len(negative_words)*1.2)+0.15))\n","        plt.plot(xx, table_negative_words[i:i+1].to_numpy().flatten(),\n","                 '-', label=negative_words[i], linewidth=4.0, color=color)\n","        \n","    plt.legend(loc='best', ncol=3, prop={'size':16}, columnspacing=0.5)\n","    plt.axis([1, 1e5, -1, 2])\n","    plt.title('Coefficient path')\n","    plt.xlabel('L2 penalty ($\\lambda$)')\n","    plt.ylabel('Coefficient value')\n","    plt.xscale('log')\n","    plt.rcParams.update({'font.size': 18})\n","    plt.tight_layout()"],"execution_count":42,"outputs":[]},{"cell_type":"code","metadata":{"id":"290ry9LMksEv","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":424},"executionInfo":{"status":"ok","timestamp":1599851394554,"user_tz":-330,"elapsed":2768,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"ede4e251-af1a-4529-be8a-080e1ad87585"},"source":["make_coefficient_plot(table, positive_words, negative_words, l2_penalty_list=[0, 4, 10, 1e2, 1e3, 1e5])"],"execution_count":43,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 720x432 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"Wfe2JDiBmGrC","colab_type":"text"},"source":["Measuring accuracy"]},{"cell_type":"code","metadata":{"id":"dZ4Rqu2Elwui","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599851802755,"user_tz":-330,"elapsed":1003,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["def get_classification_accuracy(feature_matrix, sentiment, coefficients):\n","    scores = np.dot(feature_matrix, coefficients)\n","    apply_threshold = np.vectorize(lambda x: 1. if x > 0  else -1.)\n","    predictions = apply_threshold(scores)\n","    \n","    num_correct = (predictions == sentiment).sum()\n","    accuracy = num_correct / len(feature_matrix)    \n","    return accuracy"],"execution_count":44,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"ePJ5jRScmv_e","colab_type":"text"},"source":["Below, we compare the accuracy on the training data and validation data for all the models that were trained in this assignment. We first calculate the accuracy values and then build a simple report summarizing the performance for the various models."]},{"cell_type":"code","metadata":{"id":"JPPzetAZmYW9","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1599851915004,"user_tz":-330,"elapsed":1063,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}}},"source":["train_accuracy = {}\n","train_accuracy[0]   = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_0_penalty)\n","train_accuracy[4]   = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_4_penalty)\n","train_accuracy[10]  = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_10_penalty)\n","train_accuracy[1e2] = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_1e2_penalty)\n","train_accuracy[1e3] = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_1e3_penalty)\n","train_accuracy[1e5] = get_classification_accuracy(feature_matrix_train, sentiment_train, coefficients_1e5_penalty)\n","\n","validation_accuracy = {}\n","validation_accuracy[0]   = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_0_penalty)\n","validation_accuracy[4]   = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_4_penalty)\n","validation_accuracy[10]  = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_10_penalty)\n","validation_accuracy[1e2] = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_1e2_penalty)\n","validation_accuracy[1e3] = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_1e3_penalty)\n","validation_accuracy[1e5] = get_classification_accuracy(feature_matrix_valid, sentiment_valid, coefficients_1e5_penalty)"],"execution_count":45,"outputs":[]},{"cell_type":"code","metadata":{"id":"uV7dGa10myps","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":348},"executionInfo":{"status":"ok","timestamp":1599851982804,"user_tz":-330,"elapsed":1238,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"597f26ce-fa9e-4399-ffbc-01ca78478a6b"},"source":["# Build a simple report\n","for key in sorted(validation_accuracy.keys()):\n","    print(\"L2 penalty = %g\" % key)\n","    print(\"train accuracy = %s, validation_accuracy = %s\" % (train_accuracy[key], validation_accuracy[key]))\n","    print(\"--------------------------------------------------------------------------------\")"],"execution_count":46,"outputs":[{"output_type":"stream","text":["L2 penalty = 0\n","train accuracy = 0.7851561577866434, validation_accuracy = 0.781439641490057\n","--------------------------------------------------------------------------------\n","L2 penalty = 4\n","train accuracy = 0.7851089445480512, validation_accuracy = 0.7815330034543927\n","--------------------------------------------------------------------------------\n","L2 penalty = 10\n","train accuracy = 0.7849909114515711, validation_accuracy = 0.7817197273830642\n","--------------------------------------------------------------------------------\n","L2 penalty = 100\n","train accuracy = 0.7839758268218409, validation_accuracy = 0.781066193632714\n","--------------------------------------------------------------------------------\n","L2 penalty = 1000\n","train accuracy = 0.7758551497839994, validation_accuracy = 0.7713565493417982\n","--------------------------------------------------------------------------------\n","L2 penalty = 100000\n","train accuracy = 0.6803663747314747, validation_accuracy = 0.667818130893474\n","--------------------------------------------------------------------------------\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"ir-S-tdXnIei","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":400},"executionInfo":{"status":"ok","timestamp":1599852069258,"user_tz":-330,"elapsed":1087,"user":{"displayName":"hrushikesh malpekar","photoUrl":"","userId":"12958357008200083198"}},"outputId":"315287d0-0dca-41a0-a4c7-b29a4c105948"},"source":["# Optional. Plot accuracy on training and validation sets over choice of L2 penalty.\n","import matplotlib.pyplot as plt\n","%matplotlib inline\n","plt.rcParams['figure.figsize'] = 10, 6\n","\n","sorted_list = sorted(train_accuracy.items(), key=lambda x:x[0])\n","plt.plot([p[0] for p in sorted_list], [p[1] for p in sorted_list], 'bo-', linewidth=4, label='Training accuracy')\n","sorted_list = sorted(validation_accuracy.items(), key=lambda x:x[0])\n","plt.plot([p[0] for p in sorted_list], [p[1] for p in sorted_list], 'ro-', linewidth=4, label='Validation accuracy')\n","plt.xscale('symlog')\n","plt.axis([0, 1e3, 0.78, 0.786])\n","plt.legend(loc='lower left')\n","plt.rcParams.update({'font.size': 18})\n","plt.tight_layout"],"execution_count":48,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<function matplotlib.pyplot.tight_layout>"]},"metadata":{"tags":[]},"execution_count":48},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 720x432 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]}]}