PLE Tutorial: Progressive Layered Extraction

PLE Tutorial: Progressive Layered Extraction#

Sandbox Model — Experimental

This model is part of the CoreRec Sandbox — an experimental collection of algorithms for research and exploration. Sandbox models are not production-tested and may have incomplete implementations. For production-ready models, see the Production Models.

Introduction#

PLE improves MMoE by explicitly separating shared and task-specific experts to avoid negative transfer.

Paper: Tang et al. 2020 - Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations (Tencent)

How PLE Works#

Architecture#

CGC (Customized Gate Control):

  1. Task-Specific Experts: Only feed into specific task

  2. Shared Experts: Feed into all tasks

  3. Multi-Level: Stacked CGC modules for deep extraction

Key Innovation: Eliminates negative transfer by isolating task-specific knowledge

Architecture:

[Specific Exp A] [Shared Exp] [Specific Exp B]
       |              |              |
    [Gate A]          |           [Gate B]
       ↓              |              ↓
    [Tower A]         |           [Tower B]

Mathematical Foundation#

CGC Aggregation:

E^k = [SpecificExperts^k, SharedExperts]
y^k = Gate(x) · E^k(x)

Gate only selects from relevant experts + shared experts.

Loss:

L = Σ_k w_k L_k(y_k, label_k)

Reference Implementation#

Implementation Note

The code examples below are reference implementations for learning purposes. This is a sandbox model and has not been production-tested. Verify behavior thoroughly before using in production.

Step 1: Import and Load Data#

from corerec.sandbox.collaborative_full.nn_base.PLE_base import PLE_base as PLE
from cr_learn import ml_1m
from sklearn.model_selection import train_test_split
import numpy as np

# Load dataset with CORRECT API
data = ml_1m.load()  # Returns dict with 'ratings', 'users', 'movies'
ratings_df = data['ratings']  # DataFrame with user_id, movie_id, rating, timestamp

print(f"Loaded {len(ratings_df)} ratings")

# Split data
train_df, test_df = train_test_split(ratings_df, test_size=0.2, random_state=42)

# Extract arrays for model
train_users = train_df['user_id'].values
train_items = train_df['movie_id'].values
train_ratings = train_df['rating'].values

test_users = test_df['user_id'].values
test_items = test_df['movie_id'].values
test_ratings = test_df['rating'].values

Step 2: Initialize Model#

model = PLE(
    name="PLE_Model",
    num_shared_experts=2,
    num_specific_experts=2,
    task_names=['income', 'marital'],
    epochs=20,
    batch_size=256,
    learning_rate=0.001,
    verbose=True
)

print(f"Initialized {model.name}")

Step 3: Train#

model.fit(
    user_ids=train_users,
    item_ids=train_items,
    ratings=train_ratings
)

print("Training complete!")

Step 4: Predict#

# Single prediction
score = model.predict(user_id=1, item_id=100)
print(f"Predicted score: {score:.3f}")

# Batch predictions
test_predictions = model.batch_predict(list(zip(test_users[:100], test_items[:100])))

Step 5: Recommend#

# Get top-10 recommendations for user
user_id = 1
recommendations = model.recommend(
    user_id=user_id,
    top_k=10
)

print(f"Top-10 recommendations for User {user_id}:")
for rank, item_id in enumerate(recommendations, 1):
    print(f"  {rank}. Item {item_id}")

Step 6: Evaluate#

from sklearn.metrics import mean_squared_error
import numpy as np

# Predict all test ratings
test_pred = [model.predict(u, i) for u, i in zip(test_users, test_items)]
rmse = np.sqrt(mean_squared_error(test_ratings, test_pred))
print(f"Test RMSE: {rmse:.4f}")

Step 7: Save & Load#

# Save model
model.save('artifacts/ple_model')

# Load model
loaded = PLE.load('artifacts/ple_model')
test_score = loaded.predict(1, 100)
print(f"Loaded model prediction: {test_score:.3f}")

Key Takeaways#

When to Use PLE#

✅ Perfect For:

  • Complex multi-task scenarios

  • When MMoE suffers from negative transfer

  • Highly conflicting tasks (e.g., Like vs Share)

  • Deep multi-level representation learning

❌ Not For:

  • Simple tasks

  • Limited computational budget (more params than MMoE)

Best Practices#

  1. Levels: 1-2 CGC layers

  2. Expert Split: 50% shared, 50% specific

  3. Gate Activation: Softmax

  4. Gradient Analysis: Monitor gradient conflicts

Scaling & Production Considerations#

If you plan to move this sandbox model toward production use, consider:

  • Data Volume: Test with realistic dataset sizes; sandbox implementations may not be optimized for large-scale data.

  • Distributed Training: For datasets exceeding single-machine memory, consider wrapping the model in distributed training frameworks (e.g., PyTorch DDP, Horovod).

  • Hyperparameter Tuning: Use systematic search (Optuna, Ray Tune) rather than manual tuning. The defaults in sandbox models are not tuned for any specific domain.

  • Serving Latency: Profile inference time. Some architectures (e.g., attention-based) may need quantization or ONNX export for low-latency serving.

  • Monitoring: Implement A/B testing and online metrics (CTR, engagement, diversity) before replacing an existing system.

  • Validation: Write comprehensive unit tests and integration tests before deploying. Sandbox models do not have CI coverage.

Further Reading#

  • Paper: Tang et al. 2020 - Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations (Tencent)

  • GitHub Repository

Contents