LINGUIST List 32.2508

Wed Jul 28 2021

Summer Schools: Deep Learning for Linguists / Online

Editor for this issue: Sarah Robinson <srobinsonlinguistlist.org>



Date: 25-Jul-2021
From: Associação Brasileira de Linguística <abralinabralin.org>
Subject: Deep Learning for Linguists / Online
E-mail this message to a friend

Host Institution:
Website: https://ead.abralin.org/

Dates: 04-Aug-2021 - 06-Aug-2021
Location: Online, Online, Brazil

Focus: Backpropagation is the essence of generalization in deep learning, and the course will start with a thorough introduction to this algorithm so that students get a feel for what is happening in the seemingly black box of neural networks. Most of the course will focus on the most successful neural networks: Transformers. Recent work on how transformers extract abstract linguistic structure from unlabeled data will serve as the culmination of the course. Outcomes: -Learn what a neural network does. - Learn the backpropagation algorithm. - Learn how to encourage neural networks to generalize, rather than memorize. - Learn to use Google’s Tensorflow to write simple neural networks, without any background prior background in programming. - Learn how the state-of-the-art neural networks, transformers work. - Learn how transformers learn some of the same structures from data as linguists, and maybe humans do. For more information about the course, access our e-learning platform: https://ead.abralin.org.
Minimum Education Level: No Minimum

Description:
Starting Aug 04, 2021: Deep Learning for Linguists
A free online (Zoom) course with Khalil Iskarous

Deep Learning algorithms have made profound progress over the last few years in speech and language applications. Linguists often assume that deep learning neural networks just memorize surface patterns in data. This course will introduce deep learning, from scratch (arithmetic), and show the many ways in which generalizations about speech and language corpora are learned. Special attention will be paid to the possibility that endangered language documentation work can benefit from deep learning. One of the most central concepts to be conveyed in the course is that the traditional symbolic methods of investigation employed by linguists and the numerical algorithms of deep learning are not as different from each other as they may seem on the surface. They are both about generalization extraction from data.
Tuition Explanation: Free course

Registration: 25-Jul-2021 to 03-Aug-2021
Contact Person: Associação Brasileira de Linguística
                Email: abralinabralin.org


Registration Instructions:



Page Updated: 28-Jul-2021