my solutions to the kaggle competition I have attended
the-2nd-youtube-8m-video-understanding-challenge: computer vision project, I got gold medal (7th place) using tensorflow models. Code available at https://github.com/boliu61/youtube-8m, Paper accepted by ECCV18 Youtube-8M Workshop available at: https://arxiv.org/abs/1808.06739
2018-data-science-bowl: computer vision project, I got silver medal (top 3%) using Mask-RCNN.
talkingdata-adtracking-fraud-detection: classification problem with time series. I got silver medal (top 2%) using light GBM.
google-landmark-recognition-challenge: classification problem with image. I got silver medal (top 6%) using ResNet50 and code from google-landmark-retrieval-challenge.
google-landmark-retrieval-challenge: image retrieval problem in computer vision. I got bronze medal (48/218) using DEep Local Features (DELF) for feature extraction and Faiss (a library developed by Facebook AI Research) for efficient similarity search.
avito-demand-prediction: regression problem with all-in-one features, including text, image, numerical, and categorical features. I got silver medal (top 5%) using light GBM with several engineered features on text, image, and categorical aggregation.