Data Science fwdays'20 online
2020-08-08..2020-08-15
-- online --
| title | author | length |
|---|---|---|
| Building a state-of-the-art approach | Kostya Omelyanchuk, Oleksandr Skurzhanskiy | 55m |
Відео з онлайн-конференції Data Science fwdays'20, яка пройшла 8 серпня 2020 року.
Опис доповіді "Building a state-of-the-art approach to Grammatical Error… | ||
| Active learning and weak supervision in NLP projects | Mariia Havrylovych | 45m |
This is a video from the Data Science fwdays'20 online conference, that was held on August 8, 2020.
Talk description:
Successful artificial intelligence… | ||
| BioNLP: Problems of Current Solutions. Solving Current Problems | Mikhail Sokolov | 29m |
Survey of the current state of BioNLP amid the Covid-19-driven surge in research, outlining practical problems the BioNLP community faces with datasets… | ||
| Towards fast and coherent long-form text generation | Mohit Iyyer | 52m |
Claims that replacing learned self-attention with hard-coded attention yields similar quality while cutting decode time substantially; experiments use GPT-2… | ||
| Applying Metric Learning: A Logo Recognition Case Study | Yulya Goncharenko | 38m |
Describes applying metric learning to logo recognition at YouScan: training an embedding model once so new logo classes can be added by providing 5–10… | ||
| Applications of Multimodal Learning in Media Search Engines | Dmitriy Voytekh | 56m |
Describes a production media search system that trains a single joint vector space for queries, tags, and GIFs from billions of anonymized events using… | ||
| Personal data de-identification for data science tasks | Halyna Oliinyk | 38m |
Explains concrete tradeoffs: RSA ElGamal and Paillier style homomorphic encryption preserves additive or multiplicative ops so models can train on encrypted… | ||
| Named-Entity Recognition and Name Resolution Using Similarity Queries | Aleksey Shashlyuk | 49m |
A pipeline processes 20,000 news items per day to extract named entities and resolve them against a database of 8 million companies to detect about 20 event… | ||
| Deep Learning approaches meet 3D data | Ivan Sahumbaiev | 42m |
Covers concrete methods: PointNet with MLP + max‑pool and T‑net, and PointNet++ with sampling + grouping for local context; shows segmentation and… | ||
| An Introduction to Transfer Learning and Hugging Face | Thomas Wolf | 58m |
Recent NLP breakthroughs are tied to pretraining on large corpora with Transformer architectures and then adapting them; Hugging Face supplies the actual… | ||
| Natural Language Processing 2020 Discussion | Artem Chernodub, Kyryl Truskovskyi, Maria Nadejde, Nikita Lukianets, Svitlana Galeshchuk | 57m |
Panel assesses whether 2020’s transformer wave changes practice and gives concrete takeaways: model families and deployments named (BERT, RoBERTa, TinyBERT… | ||
| ML Beyond Metrics: Challenges of Using ML Systems in the Real World | Maks Ladislav | 54m |
The talk examines practical challenges of deploying machine learning systems, insisting on concrete evidence like dataset size/composition, which features… | ||
| The democratization discussion of NLP and what it means for the industry | Thomas Wolf | 01h 10m |
Panel reports that democratization is concrete: the Hugging Face transformers hub puts pretrained models and code into three lines of use, an API for hosted… | ||
| nothing matches | ||