[R] Microsoft DeBERTa Tops Human Performance on SuperGLUE NLU Benchmark Computer Science |
- [R] Microsoft DeBERTa Tops Human Performance on SuperGLUE NLU Benchmark
- An answer about innovation behind a school project.
- What are some important discoveries in computer science so far?
- Why is Formal Methods so hard?
- Machine learning discord server
| [R] Microsoft DeBERTa Tops Human Performance on SuperGLUE NLU Benchmark Posted: 06 Jan 2021 01:03 PM PST SuperGLUE met its match this week when, for the first time, a new model surpassed human baseline performance on the challenging natural language understanding (NLU) benchmark. Dubbed DeBERTa (Decoding-enhanced BERT with disentangled attention), the breakthrough Transformer-based neural language model was initially introduced by a team of researchers from Microsoft Dynamics 365 AI and Microsoft Research in June of last year. Recently scaled up to 1.5 billion parameters, DeBERTa "substantially" outperformed the previous SuperGLUE leader — Google's 11 billion parameter T5 — and surpassed the human baseline with a score of 89.9 (vs. 89.8). Here is a quick read: Microsoft DeBERTa Tops Human Performance on SuperGLUE NLU Benchmark The paper DeBERTa: Decoding-enhanced BERT with Disentangled Attention is on arXiv. The team will update their GitHub code repository soon with the latest DeBERTa code and models. [link] [comments]  |
| An answer about innovation behind a school project. Posted: 07 Jan 2021 02:06 AM PST So my partner and I chose to create a Student Productivity App (A chatbot made from Python with some Machine Learning) that lets users type in their problems and the bot should answer corresponding questions. Our problem is (not just us, the whole class) is the professor keeps asking us what is the innovation behind our chatbot. Its been 2 months already and he isn't satiafied with any answer. We already tried mentioning that our application is supposed to be free, light-weight, and is UI friendly than other programs. But he still isn't satisfied with those answers. Computer Scientists of reddit, please help us. [link] [comments]  |
| What are some important discoveries in computer science so far? Posted: 06 Jan 2021 07:01 PM PST I've looked at some profound discoveries/developments within computer science but I want some more content to mull over, (specifically theories, not stuff like the creation of Linux or data structures for example). So far I've been really interested with the Church-Turing Thesis and what exactly constitutes a Turing machine. What are some other big theory things that I should explore? [link] [comments]  |
| Why is Formal Methods so hard? Posted: 06 Jan 2021 11:37 AM PST I'm doing my masters in Computer Engineering (embedded systems) and I have a few compulsory courses I must complete from Formal Methods. With great difficulty, I barely passed one and I'm taking another one in this quarter. Why is it FM so hard and mysterious. I'm not bad at math. I've done several other math heavy courses and I haven't struggled as much as I struggle with formal methods. Just wondering if it's just me or is it generally one of the most difficult things in life? [link] [comments]  |
| Machine learning discord server Posted: 06 Jan 2021 05:14 AM PST There is a dearth of analytics discord servers and I don't seem to find a machine learning discord focused on analytics and tabular data so I'm reaching out to compsc communities interested in joining an active text and voice chat discord community dedicated to business data and Analytics. This is a healthy and supportive Data community where we hold mock interviews, solve ML difficulties and more Whether you are beginner or an expert in the data field you will absolutely find something that interests you. Invite link : https://discord.gg/6uwvUzBH6Q [link] [comments]  |
| You are subscribed to email updates from Computer Science: Theory and Application. To stop receiving these emails, you may unsubscribe now. | Email delivery powered by Google |
| Google, 1600 Amphitheatre Parkway, Mountain View, CA 94043, United States | |
No comments:
Post a Comment