diff --git a/08.recommender_system/README.en.md b/08.recommender_system/README.en.md index 75bed8af0665a202ce0cfcb165dc8305f42e1722..6c9be597e8aaf2c41125c4281ec2332fc5228669 100644 --- a/08.recommender_system/README.en.md +++ b/08.recommender_system/README.en.md @@ -185,33 +185,37 @@ paddle.init(use_gpu=False) ```python uid = paddle.layer.data( - name='user_id', - type=paddle.data_type.integer_value( - paddle.dataset.movielens.max_user_id() + 1)) + name='user_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_user_id() + 1)) usr_emb = paddle.layer.embedding(input=uid, size=32) +usr_fc = paddle.layer.fc(input=usr_emb, size=32) usr_gender_id = paddle.layer.data( - name='gender_id', type=paddle.data_type.integer_value(2)) + name='gender_id', type=paddle.data_type.integer_value(2)) usr_gender_emb = paddle.layer.embedding(input=usr_gender_id, size=16) +usr_gender_fc = paddle.layer.fc(input=usr_gender_emb, size=16) usr_age_id = paddle.layer.data( - name='age_id', - type=paddle.data_type.integer_value( - len(paddle.dataset.movielens.age_table))) + name='age_id', + type=paddle.data_type.integer_value( + len(paddle.dataset.movielens.age_table))) usr_age_emb = paddle.layer.embedding(input=usr_age_id, size=16) +usr_age_fc = paddle.layer.fc(input=usr_age_emb, size=16) usr_job_id = paddle.layer.data( - name='job_id', - type=paddle.data_type.integer_value(paddle.dataset.movielens.max_job_id( - ) + 1)) + name='job_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_job_id() + 1)) usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) +usr_job_fc = paddle.layer.fc(input=usr_job_emb, size=16) ``` As shown in the above code, the input is four dimension integers for each user, that is, `user_id`,`gender_id`, `age_id` and `job_id`. In order to deal with these features conveniently, we use the language model in NLP to transform these discrete values into embedding vaules `usr_emb`, `usr_gender_emb`, `usr_age_emb` and `usr_job_emb`. ```python usr_combined_features = paddle.layer.fc( - input=[usr_emb, usr_gender_emb, usr_age_emb, usr_job_emb], + input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], size=200, act=paddle.activation.Tanh()) ``` @@ -226,16 +230,14 @@ mov_id = paddle.layer.data( type=paddle.data_type.integer_value( paddle.dataset.movielens.max_movie_id() + 1)) mov_emb = paddle.layer.embedding(input=mov_id, size=32) +mov_fc = paddle.layer.fc(input=mov_emb, size=32) mov_categories = paddle.layer.data( name='category_id', type=paddle.data_type.sparse_binary_vector( len(paddle.dataset.movielens.movie_categories()))) - mov_categories_hidden = paddle.layer.fc(input=mov_categories, size=32) - -movie_title_dict = paddle.dataset.movielens.get_movie_title_dict() mov_title_id = paddle.layer.data( name='movie_title', type=paddle.data_type.integer_value_sequence(len(movie_title_dict))) @@ -244,7 +246,7 @@ mov_title_conv = paddle.networks.sequence_conv_pool( input=mov_title_emb, hidden_size=32, context_len=3) mov_combined_features = paddle.layer.fc( - input=[mov_emb, mov_categories_hidden, mov_title_conv], + input=[mov_fc, mov_categories_hidden, mov_title_conv], size=200, act=paddle.activation.Tanh()) ``` diff --git a/08.recommender_system/README.md b/08.recommender_system/README.md index 812042ef9e12304a08ec112e4f4ffbd32da161c3..5ac44a6745efdeec54d1e07b94b4a661409e230f 100644 --- a/08.recommender_system/README.md +++ b/08.recommender_system/README.md @@ -20,7 +20,7 @@ ## 效果展示 -我们使用包含用户信息、电影信息与电影评分的数据集作为个性化推荐的应用场景。当我们训练好模型后，只需要输入对应的用户ID和电影ID，就可以得出一个匹配的分数（范围[1,5]，分数越高视为兴趣越大），然后根据所有电影的推荐得分排序，推荐给用户可能感兴趣的电影。 +我们使用包含用户信息、电影信息与电影评分的数据集作为个性化推荐的应用场景。当我们训练好模型后，只需要输入对应的用户ID和电影ID，就可以得出一个匹配的分数（范围[0,5]，分数越高视为兴趣越大），然后根据所有电影的推荐得分排序，推荐给用户可能感兴趣的电影。 ``` Input movie_id: 1962 @@ -189,26 +189,30 @@ print "User %s rates Movie %s with Score %s"%(user_info[uid], movie_info[mov_id] ```python uid = paddle.layer.data( - name='user_id', - type=paddle.data_type.integer_value( - paddle.dataset.movielens.max_user_id() + 1)) + name='user_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_user_id() + 1)) usr_emb = paddle.layer.embedding(input=uid, size=32) +usr_fc = paddle.layer.fc(input=usr_emb, size=32) usr_gender_id = paddle.layer.data( - name='gender_id', type=paddle.data_type.integer_value(2)) + name='gender_id', type=paddle.data_type.integer_value(2)) usr_gender_emb = paddle.layer.embedding(input=usr_gender_id, size=16) +usr_gender_fc = paddle.layer.fc(input=usr_gender_emb, size=16) usr_age_id = paddle.layer.data( - name='age_id', - type=paddle.data_type.integer_value( - len(paddle.dataset.movielens.age_table))) + name='age_id', + type=paddle.data_type.integer_value( + len(paddle.dataset.movielens.age_table))) usr_age_emb = paddle.layer.embedding(input=usr_age_id, size=16) +usr_age_fc = paddle.layer.fc(input=usr_age_emb, size=16) usr_job_id = paddle.layer.data( - name='job_id', - type=paddle.data_type.integer_value(paddle.dataset.movielens.max_job_id( - ) + 1)) + name='job_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_job_id() + 1)) usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) +usr_job_fc = paddle.layer.fc(input=usr_job_emb, size=16) ``` 如上述代码所示，对于每个用户，我们输入4维特征。其中包括`user_id`,`gender_id`,`age_id`,`job_id`。这几维特征均是简单的整数值。为了后续神经网络处理这些特征方便，我们借鉴NLP中的语言模型，将这几维离散的整数值，变换成embedding取出。分别形成`usr_emb`, `usr_gender_emb`, `usr_age_emb`, `usr_job_emb`。 @@ -216,7 +220,7 @@ usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) ```python usr_combined_features = paddle.layer.fc( - input=[usr_emb, usr_gender_emb, usr_age_emb, usr_job_emb], + input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], size=200, act=paddle.activation.Tanh()) ``` @@ -232,16 +236,14 @@ mov_id = paddle.layer.data( type=paddle.data_type.integer_value( paddle.dataset.movielens.max_movie_id() + 1)) mov_emb = paddle.layer.embedding(input=mov_id, size=32) +mov_fc = paddle.layer.fc(input=mov_emb, size=32) mov_categories = paddle.layer.data( name='category_id', type=paddle.data_type.sparse_binary_vector( len(paddle.dataset.movielens.movie_categories()))) - mov_categories_hidden = paddle.layer.fc(input=mov_categories, size=32) - -movie_title_dict = paddle.dataset.movielens.get_movie_title_dict() mov_title_id = paddle.layer.data( name='movie_title', type=paddle.data_type.integer_value_sequence(len(movie_title_dict))) @@ -250,7 +252,7 @@ mov_title_conv = paddle.networks.sequence_conv_pool( input=mov_title_emb, hidden_size=32, context_len=3) mov_combined_features = paddle.layer.fc( - input=[mov_emb, mov_categories_hidden, mov_title_conv], + input=[mov_fc, mov_categories_hidden, mov_title_conv], size=200, act=paddle.activation.Tanh()) ``` diff --git a/08.recommender_system/index.en.html b/08.recommender_system/index.en.html index 07ebae273e2282f01779d8ac1e49a0495510ffad..2a281b6675b30c7bca2d5b48d293932ab97999fa 100644 --- a/08.recommender_system/index.en.html +++ b/08.recommender_system/index.en.html @@ -227,33 +227,37 @@ paddle.init(use_gpu=False) ```python uid = paddle.layer.data( - name='user_id', - type=paddle.data_type.integer_value( - paddle.dataset.movielens.max_user_id() + 1)) + name='user_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_user_id() + 1)) usr_emb = paddle.layer.embedding(input=uid, size=32) +usr_fc = paddle.layer.fc(input=usr_emb, size=32) usr_gender_id = paddle.layer.data( - name='gender_id', type=paddle.data_type.integer_value(2)) + name='gender_id', type=paddle.data_type.integer_value(2)) usr_gender_emb = paddle.layer.embedding(input=usr_gender_id, size=16) +usr_gender_fc = paddle.layer.fc(input=usr_gender_emb, size=16) usr_age_id = paddle.layer.data( - name='age_id', - type=paddle.data_type.integer_value( - len(paddle.dataset.movielens.age_table))) + name='age_id', + type=paddle.data_type.integer_value( + len(paddle.dataset.movielens.age_table))) usr_age_emb = paddle.layer.embedding(input=usr_age_id, size=16) +usr_age_fc = paddle.layer.fc(input=usr_age_emb, size=16) usr_job_id = paddle.layer.data( - name='job_id', - type=paddle.data_type.integer_value(paddle.dataset.movielens.max_job_id( - ) + 1)) + name='job_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_job_id() + 1)) usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) +usr_job_fc = paddle.layer.fc(input=usr_job_emb, size=16) ``` As shown in the above code, the input is four dimension integers for each user, that is, `user_id`,`gender_id`, `age_id` and `job_id`. In order to deal with these features conveniently, we use the language model in NLP to transform these discrete values into embedding vaules `usr_emb`, `usr_gender_emb`, `usr_age_emb` and `usr_job_emb`. ```python usr_combined_features = paddle.layer.fc( - input=[usr_emb, usr_gender_emb, usr_age_emb, usr_job_emb], + input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], size=200, act=paddle.activation.Tanh()) ``` @@ -268,16 +272,14 @@ mov_id = paddle.layer.data( type=paddle.data_type.integer_value( paddle.dataset.movielens.max_movie_id() + 1)) mov_emb = paddle.layer.embedding(input=mov_id, size=32) +mov_fc = paddle.layer.fc(input=mov_emb, size=32) mov_categories = paddle.layer.data( name='category_id', type=paddle.data_type.sparse_binary_vector( len(paddle.dataset.movielens.movie_categories()))) - mov_categories_hidden = paddle.layer.fc(input=mov_categories, size=32) - -movie_title_dict = paddle.dataset.movielens.get_movie_title_dict() mov_title_id = paddle.layer.data( name='movie_title', type=paddle.data_type.integer_value_sequence(len(movie_title_dict))) @@ -286,7 +288,7 @@ mov_title_conv = paddle.networks.sequence_conv_pool( input=mov_title_emb, hidden_size=32, context_len=3) mov_combined_features = paddle.layer.fc( - input=[mov_emb, mov_categories_hidden, mov_title_conv], + input=[mov_fc, mov_categories_hidden, mov_title_conv], size=200, act=paddle.activation.Tanh()) ``` diff --git a/08.recommender_system/index.html b/08.recommender_system/index.html index 26514f81b6b97a13259e70083463372e9fbab8c9..d383bf1fcb9db834417293a86ef9382ef76f3754 100644 --- a/08.recommender_system/index.html +++ b/08.recommender_system/index.html @@ -62,7 +62,7 @@ ## 效果展示 -我们使用包含用户信息、电影信息与电影评分的数据集作为个性化推荐的应用场景。当我们训练好模型后，只需要输入对应的用户ID和电影ID，就可以得出一个匹配的分数（范围[1,5]，分数越高视为兴趣越大），然后根据所有电影的推荐得分排序，推荐给用户可能感兴趣的电影。 +我们使用包含用户信息、电影信息与电影评分的数据集作为个性化推荐的应用场景。当我们训练好模型后，只需要输入对应的用户ID和电影ID，就可以得出一个匹配的分数（范围[0,5]，分数越高视为兴趣越大），然后根据所有电影的推荐得分排序，推荐给用户可能感兴趣的电影。 ``` Input movie_id: 1962 @@ -231,26 +231,30 @@ print "User %s rates Movie %s with Score %s"%(user_info[uid], movie_info[mov_id] ```python uid = paddle.layer.data( - name='user_id', - type=paddle.data_type.integer_value( - paddle.dataset.movielens.max_user_id() + 1)) + name='user_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_user_id() + 1)) usr_emb = paddle.layer.embedding(input=uid, size=32) +usr_fc = paddle.layer.fc(input=usr_emb, size=32) usr_gender_id = paddle.layer.data( - name='gender_id', type=paddle.data_type.integer_value(2)) + name='gender_id', type=paddle.data_type.integer_value(2)) usr_gender_emb = paddle.layer.embedding(input=usr_gender_id, size=16) +usr_gender_fc = paddle.layer.fc(input=usr_gender_emb, size=16) usr_age_id = paddle.layer.data( - name='age_id', - type=paddle.data_type.integer_value( - len(paddle.dataset.movielens.age_table))) + name='age_id', + type=paddle.data_type.integer_value( + len(paddle.dataset.movielens.age_table))) usr_age_emb = paddle.layer.embedding(input=usr_age_id, size=16) +usr_age_fc = paddle.layer.fc(input=usr_age_emb, size=16) usr_job_id = paddle.layer.data( - name='job_id', - type=paddle.data_type.integer_value(paddle.dataset.movielens.max_job_id( - ) + 1)) + name='job_id', + type=paddle.data_type.integer_value( + paddle.dataset.movielens.max_job_id() + 1)) usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) +usr_job_fc = paddle.layer.fc(input=usr_job_emb, size=16) ``` 如上述代码所示，对于每个用户，我们输入4维特征。其中包括`user_id`,`gender_id`,`age_id`,`job_id`。这几维特征均是简单的整数值。为了后续神经网络处理这些特征方便，我们借鉴NLP中的语言模型，将这几维离散的整数值，变换成embedding取出。分别形成`usr_emb`, `usr_gender_emb`, `usr_age_emb`, `usr_job_emb`。 @@ -258,7 +262,7 @@ usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) ```python usr_combined_features = paddle.layer.fc( - input=[usr_emb, usr_gender_emb, usr_age_emb, usr_job_emb], + input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], size=200, act=paddle.activation.Tanh()) ``` @@ -274,16 +278,14 @@ mov_id = paddle.layer.data( type=paddle.data_type.integer_value( paddle.dataset.movielens.max_movie_id() + 1)) mov_emb = paddle.layer.embedding(input=mov_id, size=32) +mov_fc = paddle.layer.fc(input=mov_emb, size=32) mov_categories = paddle.layer.data( name='category_id', type=paddle.data_type.sparse_binary_vector( len(paddle.dataset.movielens.movie_categories()))) - mov_categories_hidden = paddle.layer.fc(input=mov_categories, size=32) - -movie_title_dict = paddle.dataset.movielens.get_movie_title_dict() mov_title_id = paddle.layer.data( name='movie_title', type=paddle.data_type.integer_value_sequence(len(movie_title_dict))) @@ -292,7 +294,7 @@ mov_title_conv = paddle.networks.sequence_conv_pool( input=mov_title_emb, hidden_size=32, context_len=3) mov_combined_features = paddle.layer.fc( - input=[mov_emb, mov_categories_hidden, mov_title_conv], + input=[mov_fc, mov_categories_hidden, mov_title_conv], size=200, act=paddle.activation.Tanh()) ``` diff --git a/08.recommender_system/train.py b/08.recommender_system/train.py index e57599254cf2fb67309012db24febff28c66e6ed..2adb802f0e79cebac3ed08d70429ba3b35da307c 100644 --- a/08.recommender_system/train.py +++ b/08.recommender_system/train.py @@ -11,26 +11,29 @@ def main(): type=paddle.data_type.integer_value( paddle.dataset.movielens.max_user_id() + 1)) usr_emb = paddle.layer.embedding(input=uid, size=32) + usr_fc = paddle.layer.fc(input=usr_emb, size=32) usr_gender_id = paddle.layer.data( name='gender_id', type=paddle.data_type.integer_value(2)) usr_gender_emb = paddle.layer.embedding(input=usr_gender_id, size=16) + usr_gender_fc = paddle.layer.fc(input=usr_gender_emb, size=16) usr_age_id = paddle.layer.data( name='age_id', type=paddle.data_type.integer_value( len(paddle.dataset.movielens.age_table))) usr_age_emb = paddle.layer.embedding(input=usr_age_id, size=16) + usr_age_fc = paddle.layer.fc(input=usr_age_emb, size=16) usr_job_id = paddle.layer.data( name='job_id', type=paddle.data_type.integer_value( paddle.dataset.movielens.max_job_id() + 1)) - usr_job_emb = paddle.layer.embedding(input=usr_job_id, size=16) + usr_job_fc = paddle.layer.fc(input=usr_job_emb, size=16) usr_combined_features = paddle.layer.fc( - input=[usr_emb, usr_gender_emb, usr_age_emb, usr_job_emb], + input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], size=200, act=paddle.activation.Tanh()) @@ -39,12 +42,12 @@ def main(): type=paddle.data_type.integer_value( paddle.dataset.movielens.max_movie_id() + 1)) mov_emb = paddle.layer.embedding(input=mov_id, size=32) + mov_fc = paddle.layer.fc(input=mov_emb, size=32) mov_categories = paddle.layer.data( name='category_id', type=paddle.data_type.sparse_binary_vector( len(paddle.dataset.movielens.movie_categories()))) - mov_categories_hidden = paddle.layer.fc(input=mov_categories, size=32) mov_title_id = paddle.layer.data( @@ -55,7 +58,7 @@ def main(): input=mov_title_emb, hidden_size=32, context_len=3) mov_combined_features = paddle.layer.fc( - input=[mov_emb, mov_categories_hidden, mov_title_conv], + input=[mov_fc, mov_categories_hidden, mov_title_conv], size=200, act=paddle.activation.Tanh()) @@ -106,9 +109,6 @@ def main(): feature = user.value() + movie.value() - def reader(): - yield feature - infer_dict = copy.copy(feeding) del infer_dict['score']