Difference between revisions of "Meta"
Jump to navigation
Jump to search
Zeno Gantner (talk | contribs) (add paper ''Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models'', ISCA 2022 industry track) |
Zeno Gantner (talk | contribs) (→Papers) |
||
| Line 22: | Line 22: | ||
# {{ops}}{{performance}} [https://arxiv.org/pdf/2104.05158.pdf Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models], ISCA 2022 industry track | # {{ops}}{{performance}} [https://arxiv.org/pdf/2104.05158.pdf Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models], ISCA 2022 industry track | ||
# {{ops}}{{performance}} [https://dl.acm.org/doi/abs/10.1145/3534678.3539034 AutoShard: Automated Embedding Table Sharding for Recommender Systems], [[KDD 2022]] | # {{ops}}{{performance}} [https://dl.acm.org/doi/abs/10.1145/3534678.3539034 AutoShard: Automated Embedding Table Sharding for Recommender Systems], [[KDD 2022]] | ||
| + | # [https://arxiv.org/pdf/2203.11014.pdf DHEN: A Deep and Hierarchical Ensemble Network for Large-Scale Click-Through Rate Prediction], KDD 2022 | ||
# https://research.fb.com/publications/the-decoupled-extended-kalman-filter-for-dynamic-exponential-family-factorization-models/, [[JMLR]], 2021 | # https://research.fb.com/publications/the-decoupled-extended-kalman-filter-for-dynamic-exponential-family-factorization-models/, [[JMLR]], 2021 | ||
# [https://research.fb.com/publications/preference-amplification-in-recommender-systems/ Preference Amplification in Recommender Systems], [[KDD 2021]] | # [https://research.fb.com/publications/preference-amplification-in-recommender-systems/ Preference Amplification in Recommender Systems], [[KDD 2021]] | ||
Revision as of 08:08, 19 May 2023
Meta is the company behind Facebook, Instagram, and WhatsApp.
Contents
Blog posts
- The new AI-powered feature designed to improve Feed for everyone, 2022-10-05
- When do recommender systems amplify user preferences? A theoretical framework and mitigation strategies, 2021-08
- Efficient tuning of online systems using Bayesian optimization
- On the value of diversified recommendations, 2020-12-17
- How Instagram suggests new content, 2020-12-10
- Five things I learned about working on content quality at Instagram, 2020-01-25
- 🕴 Instagram’s Explore Recommender System, 2019-11-26 HackerNews discussion
- 3-part funnel (2 layers of candidate generation)
- domain-specific language (We have separation between model and filter config).
- account embeddings
- embedding-based
- “See fewer posts like this” – explicit feedback
- Offline Policy Evaluation: Run fewer, better A/B tests
- DLRM: An advanced, open source deep learning recommendation model, 2019-07-02
Papers
- 🔩🏃 Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models, ISCA 2022 industry track
- 🔩🏃 AutoShard: Automated Embedding Table Sharding for Recommender Systems, KDD 2022
- DHEN: A Deep and Hierarchical Ensemble Network for Large-Scale Click-Through Rate Prediction, KDD 2022
- https://research.fb.com/publications/the-decoupled-extended-kalman-filter-for-dynamic-exponential-family-factorization-models/, JMLR, 2021
- Preference Amplification in Recommender Systems, KDD 2021
- 🔍 Embedding-based Retrieval in Facebook Search, KDD 2020
- Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems, KDD 2020
- 👗 Modeling Fashion Influence from Photos, IEEE Transactions on Multimedia, 2020
- 🧠 Deep Learning Recommendation Model for Personalization and Recommendation Systems, 2019 (DLRM)
- 💵 Counterfactual Reasoning and Learning Systems: The Example of Computational Advertising
- Mixed Dimension Embeddings with Application to Memory-Efficient Recommendation Systems
- 💵 Practical lessons from predicting clicks on ads at facebook, workshop 2014
- Supervised Random Walks: Predicting and Recommending Links in Social Networks, WSDM 2011
Software
- DLRM recommender: click probability model
- Prophet: forecasting time series data based on an additive model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects.
- Facebook AI Performance Evaluation Platform: framework and backend agnostic benchmarking platform to compare machine learning inferencing runtime metrics on a set of models and on variety of backends.
- 📑 StarSpace: Learning embeddings for classification, retrieval and ranking.
- ReAgent/Horizon: end-to-end platform designed to solve industry applied RL problems where datasets are large (millions to billions of observations), the feedback loop is slow (vs. a simulator), and experiments must be done with care because they don’t run in a simulator. Tutorial contains e-commerce/recommendation example. paper
- faiss: library for efficient similarity search and clustering of dense vectors.
- pysparnn: approximate nearest neighbor search for sparse data in Python.
- Ax: adaptive experimentation platform, ax.dev
