Traditional approaches to machine learning focus on a centralised setting in which all the relevant data is brought to a central computational node, which is then trained on it. However, in many real-world systems, data is continuously generated across a wide region, and it is impractical or costly to bring it to one location. In addition, there may be security or privacy concerns related to fully sharing data. These motivate approaches to spreading the learning task across a large number of local agents while constraining communication between them.

This workshop will bring together researchers in a focused study group to discuss open problems related to the mathematical foundations of decentralised learning and applications in diverse domains. The workshop is supported by a joint research grant under the UK-India Educational and Research Initiative (UKIERI).

Attendance is by invitation only. If you are interested in participating, please email a.ganesh@bristol.ac.uk

Programme:

Monday

9.25:  Opening remarks and Welcome 

9.30-9.55:  Po-Ling Loh (Cambridge)

9.55-10.20:  Parimal Parag (IISc)

10.20-10.45:  Song Liu (Bristol)

10.45-11.10:  Jonathan Lawry (Bristol)

11.10-11.40:  Coffee break

11.40-12.05:  Chandramani Singh (IISc)

12.05-12.30:  Sidharth Jaggi (Bristol)

12.30-12.55:  Varun Jog (Cambridge)

13.00:  Lunch

Afternoon:  Collaboration 

1500:  Coffee

Tuesday

9.30-9.55:  Ghurumuruhan Ganesan (Bristol)

9.55-10.20:  Henry Reeve (Bristol)

10.20-10.45:  Ayalvadi Ganesh (Bristol)

10.45-11.10:  Ankita Koley (IISc)

11.10-11.40:  Coffee

11.40-12.05:  Aniket Mukherjee (IISc)

12.05-13.00:  More student talks?

13.00:  Lunch

Afternoon: Collaboration

15.00: Coffee

18.30: Conference dinner

Wednesday

Morning:  Collaboration

11.00:  Coffee

13.00:  Lunch

14.00-15.00:  Progress reports

15.00:  Coffee

15.30:  Collaboration