|
|
--- |
|
|
license: apache-2.0 |
|
|
tags: |
|
|
- onnx |
|
|
- flock |
|
|
- dark-pool |
|
|
- trading |
|
|
- reinforcement-learning |
|
|
--- |
|
|
|
|
|
# Flock.io Task 18: Dark Pool Trading Model |
|
|
|
|
|
This is a neural network model trained for the Flock.io Task 18 - Dark Pool Trading. |
|
|
|
|
|
## Model Details |
|
|
|
|
|
- **Task**: Dark Pool Trading Prediction |
|
|
- **Framework**: PyTorch → ONNX |
|
|
- **Input**: 34 features (market state) |
|
|
- **Output**: 1 value (predicted fill rate / action value) |
|
|
- **Parameters**: ~1.78M (under 3M limit) |
|
|
|
|
|
## Architecture |
|
|
|
|
|
Multi-Layer Perceptron (MLP): |
|
|
- Input: 34 features |
|
|
- Hidden layers: 1024 → 1024 → 512 → 256 → 128 |
|
|
- Output: 1 value |
|
|
- Activation: ReLU |
|
|
- Normalization: BatchNorm |
|
|
- Regularization: Dropout (0.2) |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
import onnxruntime as ort |
|
|
import numpy as np |
|
|
|
|
|
# Load the model |
|
|
session = ort.InferenceSession("model.onnx") |
|
|
|
|
|
# Prepare input (34 features) |
|
|
input_data = np.random.randn(1, 34).astype(np.float32) |
|
|
|
|
|
# Run inference |
|
|
outputs = session.run(None, {"input": input_data}) |
|
|
prediction = outputs[0] |
|
|
print(f"Prediction: {prediction}") |
|
|
``` |
|
|
|
|
|
## Training |
|
|
|
|
|
Trained on Flock.io Task 18 dataset: |
|
|
- Training samples: 1200 |
|
|
- Validation samples: 400 |
|
|
- Best validation loss: ~0.001 |
|
|
|
|
|
## License |
|
|
|
|
|
Apache 2.0 |
|
|
|
