文章目录
1.机器学习
1.1 通过内容进行年龄和性别预测
样本数据约7000w,90万用户三个月内点击广告的日志来预测用户性别和年龄
机器学习:lightgbm(入门级1.3左右)
深度学习:推荐RNN+NLP相关算法(霸榜大佬1.4+)
1.2 垃圾邮件分类
KNN-SVM-DT-RF-GBDT-Bayes
1.3 金融反欺诈二分类
1. XGBOOST
2. GBDT
3. LR
1.4 音乐推荐系统
SVM
1.5 手写数字识别
model 成绩 耗时
1. KNN 0.97282 258.80秒
2. LR 0.88603 44.04秒
3. RF-pca 0.94782 55.12秒
4. SVM 0.88375 534.41秒
5. XGBoost 0.97282 2000秒
6. RF-only 0.96796 50秒
2.深度学习
2.1 图像多分类104
There are over 5,000 species of mammals, 10,000 species of birds, 30,000 species of fish – and astonishingly, over 400,000 different types of flowers.
In this competition, you’re challenged to build a machine learning model that identifies the type of flowers in a dataset of images (for simplicity, we’re sticking to just over 100 types).
tf.keras.applications.DenseNet201()
1. EfficientNetB7 ~0.96-50epochs/bs=64/lr=0.00001
2. DenseNet201
2.2 吸烟打电话检测
2020中国华录杯·数据湖算法大赛—定向算法赛(吸烟打电话检测)-tensorflow2.3-python3.6-迁移学习
1. inceptionV3
2. resnet50
3.计算机视觉CV
3.1目标检测
1. yolov5
2. Resnet50
3. EfficientDet
4. Mobilenet
4.自然语言处理NLP
1. Word2vec
2. Seq2Seq
3. Attention
4. Transformer
5. Bert