Flames of Winter and Rapid Fire Talks: Co-Hosted with Burlington Data Scientists

Come huddle around the projector light and join us for some data driven warmth in the middle of winter!

Continuing our 5 year tradition of brightening the darkest nights of winter with a party, on Wednesday January 22nd Burlington Data Scientists are hosting "Rapid Fire" data science talks.

Continuing our 2 year tradition, we are co-hosting with Women in Machine Learning and Data Science!

7-by-7!

We have 7 exciting 7-minute talks planned for this event!

After the talks we'll all move to Skinny Pancake to continue our discussions. We're working on pulling together some funding to provide free drinks.

Our Speakers...

JANE ADAMS
I'd like to introduce the idea of network graphs for exploratory analysis of high-dimensional data. Nodes are features, and edges are drawn based on similarity, measured in various ways (correlation, conditional probability, mutual information).

BRENT SITTERLY
Data visualization is cool...but is it practical. Alternatively, what is a scatterplot worth? I will spend 7 minutes talking about how data visualization can be an invaluable tool to showcases the value that data science can bring to an organization.

MARK EHLER
The goal of my presentation would be to briefly introduce audio modeling using the Freesound dataset. I think this presentation would be useful for musicians with no experience in data modeling and data scientists with little to no experience working with audio. This presentation introduces audio specific challenges and how to address them such as, data augmentation using specific formulas (MFCCs, Spectrogram, DFT), and selecting for the purest signal. This presentation will also go over a couple familiar use cases for audio processing and briefly looks at how they work (speech recognition, shazam, google magenta)

SCOTT THIBAULT
I would like to introduce people to the BTV Waterfront's Low Power Environmental Sensor Network. It is an open public wireless network that sensors on the waterfront can connect to and send sensor data. The network is up and now is time to start deploying sensors and figuring out the type data we should collect and what we can do with it.

CHARLES MASENAS
As a leisure time project, I have been messing with open source donkeycars (https://www.donkeycar.com, https://github.com/autorope/donkeycar). Donkeycars use ML to drive autonomously around a real world track. The vehicle has an onboard raspberry pi and training is done on a laptop or desktop. I have hacked the software and hardware to better negotiate the tight turns on my basement track. I can bring my vehicles for demonstration, one of which is the new AWS Deepracer which Amazon uses to promote their web services.

STEVE COMEAU
Would like to present on using Inductive Automation Ignition software as a tool for automated data collection field metering devices, such as counters, electricity meters, and sensors. Specifically will describe how Ignition "Data Bridge" can be used to get data and log data to an SQL Database by Inserting, Updating, or calling stored procedures based on data change or schedules. This topic is described more broadly in this blog post. https://www.hallam-ics.com/blog/data-sources-and-transactions-for-manufacturing-intelligence

TOM DINITZ
I'd like to give a quick overview of recent advances in NLP. Specifically the introduction of transformers in 2017, and BERT, the model released by Google in 2018. My aim is to provide some intuition into what these did, why they were so impactful, and how easy it is for anyone to start using BERT to help with their own NLP problems.