Bath ML Mini-Conference
Bath Machine Learning Meetup
1pm - 5pm
We're organising a day of multiple talks by the members of our community. Note: The schedule may look slightly different on the day.
12:45 - Doors open
13:15 - 14:15 Deep Learning with Keras - Owen Jones
Do you like the idea of waking up on a Monday morning, looking in the bathroom mirror, smiling, and saying "I am a deep learning expert"?
Actually, you can do that whenever you like (well, as long as it's a Monday morning). But what if you genuinely DID know enough about deep learning to impress your friends/family/colleagues/household pets with some cool theory and kickass code?
BathML is here to help. We're going to use Keras to build a convolutional neural network for classifying time series data.
And whenever we organise these workshops, we have to choose a language; and whichever one we choose, people get upset because we haven't chosen their favourite. So we've made a decision, and we're going to do it in R.
At. The. Same. Time.
It's like Batman vs Superman, but if Batman and Superman were actually really great friends and also if they were both popular open source statistical programming languages. It's going to be epic. Don't miss it.
14:30 - 15:15 Thomas Perryman - Generating Images with Variational Autoencoders (VAE)
“Sampling from a latent space of images to create entirely new images or edit existing ones is currently the most popular and successful application of creative AI.”
- François Chollet (primary author and maintainer of Keras)
First we’ll look at how autoencoders can be created and used for data compression. With a tweak, these models can be converted into something that can generate new images from a dataset. These models are called Variational Autoencoders, which try to find the latent variables (the hidden parameters) which can be manipulated to generate images. See the following webpage for an interactive example: http://vdumoulin.github.io/morphing_faces/online_demo.html
Variational Autoencoders have been around since 2013, and much has been written about how to generate images, the next step in my research is to look at how to use this technique to generate audio timbres.
15:30 - 16:15 - Austeja P - Mathematics in ML and how I learnt it.
SVM, hierarchical clustering, logistic regression... there are a lot of mathematical terms that come up when reading literature on machine learning. Austeja will talk about how she learnt the ML maths and hopefully help you understand some of those terms. If you have any mathematical terms you'd like to understand better, please let us know here! https://goo.gl/forms/35PF6rDVPEET5AHz2
16:25 - 16:45 You - Lightning talks (schedule may be longer or shorter depending on interest)
Lightning talks are 5-10 minute talks where anyone can talk about something they're passionate about - assuming it's at least tangential to Machine Learning ;)
If you would like to do a lightning talk about anything even slightly related to ML please contact the organisers. We are more than happy to help you prepare and get everything set up if you are new to speaking.
16:45 - 17:00 Austeja P - The past, present and future of Bath ML.
It's been almost two years since we started the Meetup and we will soon be handing over the group to new organisers. We'd like to summarise the activities we had so far and share the experiences we had as organisers.
As always, we'll be providing refreshments on the day - and if people are not too exhausted from all the ML we might head to a local watering hole.