Upload README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Bidding Algorithms Benchmark
|
| 2 |
+
|
| 3 |
+
> Complete comparison framework for real-time bidding (RTB) algorithms in online advertising.
|
| 4 |
+
> Optimizing for clicks under budget constraints using Lagrangian dual methods.
|
| 5 |
+
|
| 6 |
+
## Research Resources
|
| 7 |
+
|
| 8 |
+
- **[RESEARCH_RESOURCES.md](RESEARCH_RESOURCES.md)** — Full literature survey: 32 papers across bidding algorithms, CTR prediction, and clearing price models
|
| 9 |
+
- **[AUDIT_TRAIL.md](AUDIT_TRAIL.md)** — Every paper, dataset, codebase, and external resource consulted (44 items)
|
| 10 |
+
|
| 11 |
+
## Problem Setup
|
| 12 |
+
|
| 13 |
+
- **Objective**: Maximize number of clicks
|
| 14 |
+
- **Constraints**: Total spend ≤ Budget, with k% minimum spend guarantee
|
| 15 |
+
- **Auction Types**: First-price and second-price
|
| 16 |
+
- **Core Approach**: Lagrangian dual multiplier with online error gradient descent
|
| 17 |
+
|
| 18 |
+
## Algorithms
|
| 19 |
+
|
| 20 |
+
| Algorithm | Type | Auction | Paper |
|
| 21 |
+
|-----------|------|---------|-------|
|
| 22 |
+
| **DualOGD** | Adaptive | First-price | Wang et al. 2023 [2304.13477] |
|
| 23 |
+
| **DualMirrorDescent** | Adaptive | Second-price | Balseiro et al. 2020 [2011.10124] |
|
| 24 |
+
| **DualRoS** | Adaptive | Second-price | Feng et al. 2022 [2208.13713] |
|
| 25 |
+
| **TwoSidedDual** | Adaptive | First-price | Extension (cap + floor) |
|
| 26 |
+
| **RLB** | RL+MDP | Both | Cai et al. 2017 [1701.02490] |
|
| 27 |
+
| **Linear** | Static | Both | Baseline |
|
| 28 |
+
| **ORTB** | Static | Second-price | Zhang et al. 2014 (KDD) |
|
| 29 |
+
|
| 30 |
+
## Models
|
| 31 |
+
|
| 32 |
+
| Model | Task | Architecture | Dataset |
|
| 33 |
+
|-------|------|-------------|---------|
|
| 34 |
+
| **FinalMLP** | CTR Prediction | Two-stream MLP + Feature Gating | Criteo_x4 |
|
| 35 |
+
| **DeepFM** | CTR Prediction | FM + DNN (baseline) | Criteo_x4 |
|
| 36 |
+
| **TorchSurv** | Clearing Price | Deep Survival (Cox PH) | Simulated |
|
| 37 |
+
| **EmpiricalCDF** | Win Probability | Non-parametric | Online |
|
| 38 |
+
|
| 39 |
+
## Structure
|
| 40 |
+
|
| 41 |
+
```
|
| 42 |
+
bidding_algorithms_benchmark/
|
| 43 |
+
├── README.md
|
| 44 |
+
├── RESEARCH_RESOURCES.md
|
| 45 |
+
├── AUDIT_TRAIL.md
|
| 46 |
+
├── src/
|
| 47 |
+
│ ├── ctr/
|
| 48 |
+
│ │ ├── train_finalmlp.py
|
| 49 |
+
│ │ └── train_deepfm.py
|
| 50 |
+
│ ├── price/
|
| 51 |
+
│ │ ├── empirical_cdf.py
|
| 52 |
+
│ │ └── torchsurv_model.py
|
| 53 |
+
│ ├── algorithms/
|
| 54 |
+
│ │ ├── dual_ogd.py
|
| 55 |
+
│ │ ├── dual_mirror_descent.py
|
| 56 |
+
│ │ ├── dual_ros.py
|
| 57 |
+
│ │ ├── two_sided_dual.py
|
| 58 |
+
│ │ ├── rlb.py
|
| 59 |
+
│ │ └── baselines.py
|
| 60 |
+
│ └── benchmark/
|
| 61 |
+
│ ├── auction_simulator.py
|
| 62 |
+
│ ├── run_comparison.py
|
| 63 |
+
│ └── sweep.py
|
| 64 |
+
├── configs/
|
| 65 |
+
│ ├── finalmlp_criteo.yaml
|
| 66 |
+
│ └── sweep_config.yaml
|
| 67 |
+
├── results/
|
| 68 |
+
└── requirements.txt
|
| 69 |
+
```
|