Pie & AI: Amsterdam - Deep Learning and Formula 1

Event Information

Share this event

Date and Time

Location

Location

GoDataDriven

202 Wibautstraat

1091 GS Amsterdam

Netherlands

View Map

Event description
Pie & AI - Andrew Ng Initiative - Amsterdam Kick-Off - a deeplearning.ai meetup hosted by community groups

About this Event

Join us for the inaugural Pie & AI Amsterdam event. Pie & AI is an initiative from Deeplearning.ai, a global, diverse network of deep learners passionate about learning and building AI.

The meetup series Pie & AI typically includes conversations with AI leaders, thought-provoking discussions, networking opportunities, hands-on project practice, and pie.

For this first edition in Amsterdam, we are working on the exact program, but you can expect a good mixture of applied AI and academic research.

Confirmed speakers include:

Luuk Figdor, data scientist at AWS Professional Services, who will talk about machine learning to enhance viewer experience in the Formula 1

Didrik Nilsen, PhD student. Currently working on probabilistic deep generative models with Prof. Ole Winther at DTU Compute in Copenhagen, Denmark. He also works part-time at the start-up raffle.ai, developing corporate virtual assistants using information retrieval, natural language processing and deep learning.

Amirhossein Habibian, researcher from Qualcomm to talk about a video compression with rate-distortion autoencoders

Didrik Nilsen will talk about:

Normalizing flows provide an attractive approach to building deep generative models.

Flows rely on invertible transformations to map between a simple distribution to a more complex one.

Expressive models can be built by composing transformations in deep architectures.

The use of invertible transformations allows exact computation of the likelihood using the change-of-variables formula.

However, when applying flows to quantized data such as images and audio, dequantization have to be applied, yielding a lower bound on the likelihood.

I will first give a brief introduction to normalizing flows. Next, I will present recent work on dequantization and its effect on performance.

Share with friends

Date and Time

Location

GoDataDriven

202 Wibautstraat

1091 GS Amsterdam

Netherlands

View Map

Save This Event

Event Saved