Build & Evaluate GDCN-Final Fusion Agent on Criteo
Gated Deep Cross Network (GDCN) enhances Click-Through Rate (CTR) prediction in recommender systems by improving interpretability. Implement the state-of-the-art GDCN-Final Fusion Agent architecture from scratch, leveraging its dual-gated GDCN stream, feature-selected MLP stream, and bilinear fusion. The challenge involves developing a robust data pipeline for the Criteo dataset, including log-binning for numerical features, training the model, and establishing a rigorous AUC evaluation harness. Practitioners will demonstrate their ability to translate a complex architectural description into a working deep learning model and rigorously assess its performance. This task simulates a real-world scenario where an ML engineer must reproduce a research paper's findings, ensuring all nuanced components are correctly implemented and evaluated on a large-scale industrial dataset. The focus is on correctness, efficiency, and achieving competitive AUC scores while maintaining a reproducible training and evaluation pipeline.
What you are building
The core problem, expected build, and operating context for this challenge.
Implement the described GDCN-Final Fusion Agent, train it on the Criteo dataset, and rigorously evaluate its AUC performance.
How work is evaluated
The solution will be evaluated based on the correctness of the GDCN-Final Fusion Agent implementation, the stability of the training process, and the achieved AUC performance on a held-out validation set of the Criteo dataset. A Docker container will be used to ensure reproducible execution and environment consistency.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Understand and debug advanced PyTorch model components (custom layers, gates).
Preprocess large-scale CTR datasets (Criteo) including numerical feature binning.
Set up and execute a robust training and evaluation pipeline for CTR models.
Implement complex deep learning architectures from a conceptual description.
Accurately measure and interpret AUC for ranking tasks.
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
The solution will be evaluated based on the correctness of the GDCN-Final Fusion Agent implementation, the stability of the training process, and the achieved AUC performance on a held-out validation set of the Criteo...
Challenge input
{ "train_data_path": "<path_to_train_csv>", "test_data_path": "<path_to_test_csv>", "config": { "embedding_dim": 16, "hidden_units": [256, 128], "c...
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{ "auc_score": <float>, "model_weights_path": "<path_to_saved_model.pth>", "training_log": "<full_training_log_string>" }
- The provided code must execute the training and evaluation steps without critical runtime errors.
- The trained model weights must be successfully saved to the specified path and be loadable.
- Final Auc target: 0.815
- Docker execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- The provided code must execute the training and evaluation steps without critical runtime errors.
- The trained model weights must be successfully saved to the specified path and be loadable.
Metrics
- final_auc ยท Target 0.815
Workflow tasks
- Task 1
Input: { "train_data_path": "<path_to_train_csv>", "test_data_path": "<path_to_test_csv>", "config": { "embedding_dim": 16, "hidden_units": [256, 128], "cross_layer_num": 3, "mlp_dropout": 0.2, "learning_rate": 0.001, "batch_size": 2048, "epochs": 5, "log_bins": 16, "hash_dim": 100000 } }
Output: { "auc_score": <float>, "model_weights_path": "<path_to_saved_model.pth>", "training_log": "<full_training_log_string>" }
Execution harness: Docker
Recipe freeze marker
No recipe freeze time is recorded.