Schedule
The workshop will be a full day event from 9:00 am - 6:15 pm EDT (New York Time).Throughout the day we will be livestreaming invited talks, contributed talks, two poster sessions, and a live panel discussion.
The schedule is as follows. All times listed are in EDT (New York Time):
| Time | Talk Title | Speaker |
|---|---|---|
| 09:00 - 09:10 | Opening Remarks | |
| 09:10 - 09:40 | Invited Talk: Practical Machine Learning for Developing Countries: Lessons and Reflections | Charity Wayua |
| 09:40 - 09:45 | Invited Live Q&A: Practical Machine Learning for Developing Countries: Lessons and Reflections | Charity Wayua |
| 09:45 - 10:45 | Panel Discussion: on practical issues of AI for developing countries | Jade Abbot, Kommy Woldemariam, Geoffery Siwo, and Muthoni Wanyoike |
| 10:50 - 11:05 | Contributed Talk 14: Beyond Good Intentions: Data Sharing in and for Africa | Sara Kingsley (CMU) |
| 11:05 - 11:20 | Contributed Talk 08: Measuring Changes in Poverty with Deep Learning and Satellite Imagery | Lukas Kondmann (German Aerospace Center) |
| 11:20 - 11:35 | Contributed Talk 07: Rigging the Lottery: Making All Tickets Winners | Utku Evci (Google AI) |
| 11:35 - 11:50 | Contributed Talk 11: Distributed Learning: Sequential Decision Making in Resource-Constrained Environments | Udari Madhushani (Princeton University) |
| 11:50 - 12:05 | Contributed Talk 09: Distant Supervision and Noisy Label Learning for Low Resource Named Entity Recognition: A Study on Hausa and Yorùbá | Michael Hedderich (Saarland University) |
| 12:05 - 12:35 | Live Q&A for Contributed Talk Speakers 14, 08, 07, 11, and 09 | |
| 12:35 - 01:35 | Poster Session 1 | |
| 01:40 - 02:10 | Invited Talk: Lessons learned from practicing and teaching data science in Latin America | John Alexis Guerra Gómez |
| 02:10 - 02:15 | Invited Live Q&A: Lessons learned from practicing and teaching data science in Latin America | John Alexis Guerra Gómez |
| 02:20 - 02:35 | Contributed Talk 12: Streamlining Tensor and Network Pruning in PyTorch | Jessica Forde (Brown University) |
| 02:35 - 02:50 | Contributed Talk 05: Convolutional Neural Network for ECG-based Virtual Pathology Stethoscope Tracking in Patient Heart Auscultation | Haben Yhdego (Old Dominion University) |
| 02:50 - 03:05 | Contributed Talk 10: Spanish pre-trained BERT model and evaluation data | Gabriel Chaperon (Universidad de Chile) |
| 03:05 - 03:20 | Contributed Talk 06: LOW RESOURCE BREAST CANCER DETECTION WITH MAMMOGRAMS | Sara Ebrahim (AIMS Rwanda) |
| 03:20 - 03:35 | Contributed Talk 18: Predicting Legal Proceedings Status: an Approach Based on Sequential Texts | Felipe Polo (University of São Paulo) |
| 03:35 - 03:55 | Live Q&A for Contributed Talk Speakers 12, 05, 10, 06, and 18 | |
| 03:55 - 04:25 | Invited Talk: Learning with less resources: minimizing the labeling effort | Negar Rouzmateh |
| 04:25 - 04:30 | Invited Live Q&A: Learning with less resources: minimizing the labeling effort | Negar Rouzmateh |
| 04:30 - 05:30 | Poster Session 2 | |
| 05:35 - 05:45 | Closing Remarks |
The Panel Discussion will focus on training data creation and ownership issues: The tension between reducing resources such as the expense to annotate and gather datasets vs. hiring people to label and annotate datasets (many of the people labeling right now are from the developing world). Most training data labeling companies are hiring people from the developing world for this effort. But a lot of the data is not owned by people from the developing world.