DSSTNE: A New Deep Learning Framework For Large Sparse Datasets

Szilard is an expert data scientist and the organizer/co-organizer of most data science meetups in LA (R, DataVis, Data Science/Machine Learning, Big Data).

Szilard Pafka

DSSTNE: A New Deep Learning Framework for Large Sparse Datasets
Speaker: Scott Le Grand
This talk is a gentle introduction to Amazon open sourced deep learning library <A HREF="https://github.com/amznlabs/amazon-dsstne" TARGET="_blank" DATA-MSYS-CLICKTRACK="0" REL="nofollow noopener noreferrer">DSSTNE</A>. The creation of DSSTNE is motivated by the massive scale and low latency computing needed for Amazon scale catalog and recommendation production problem. It is also motivated by the inherent extreme sparseness of recommendation or data sets in Amazon. In this talk Scott will walk through such requirements for a new (better) deep learning frame suitable for large scale production use. He will delineate a few considerations and solutions of how DSSTNE address those challenges. He will also briefly introduce DSSTNE’s API and how it can be used to easily reproduce the state of art deep nets topologies e.g. Alex Net.  This talk also includes DSSTNE’s performance compared to TensorFlow and compared to other machine learning algorithms on benchmark datasets.
Bio:  Scott Le Grand is currently a senior scientist in Tezza Technology. He is the lead author of DSSTNE (Deep Sparse Scalable Tensor Network Engine). DSSTNE is an Amazon developed library for building Deep Learning (DL) machine learning (ML) models, and in one of forces powering the personalization of the gigantic Amazon retail business. Besides the creation of DSSTNE, Scott is also leading the effort of optimizing Deep Neural Network (DNN) performance on GPUs and GPU clusters and applying DSSTNE to general machine learning, deep learning and analytical applications.
Scott has been an expert in developing large scale distributed algorithms and systems. He developed life science services on Amazon's Elastic Compute Cloud (EC2) as a principal engineer at Amazon AWS. Before Amazon he was a distinguished Principal Engineer and a CUDA Fellow in NVIDIA. There he was best known for his work porting the AMBER molecular dynamics package to CUDA, attaining record-breaking performance in the process. In a previous life, Scott picked up a B.S. in biology from Siena College and a Ph.D. in biochemistry from the Pennsylvania State University.
Date: October 27, 2016
Timeline: – 6:30pm arrival, food/drinks and networking – 7:30pm introductions, main talk  – 8:30pm after party: more networking, drinks, games, demos etc – 10:00pm closing 
You must have a confirmed RSVP and please arrive by 7:25pm the latest. Please RSVP here on Eventbrite.
Venue: Red Bull Media House, 1740 Stewart Street, Santa Monica Parking: You can park for free in the gated Red Bull Media House parking lot (see picture below). 
Thanks Red Bull Media House for hosting at this awesome venue.

Red Bull Media House

Our {communityGuidelinesLink} describe the sort of content we prohibit on Eventbrite. If you suspect an event may be in violation, you can report it to us so we can investigate.

To report other categories of prohibited or illegal content, submit {link}

DSSTNE: A New Deep Learning Framework For Large Sparse Datasets

More events from Szilard Pafka

Follow organizers to get events picked for you

Still looking for the right event?

Explore all events in Santa Monica and filter by date, category, and more to find the perfect fit.

DSSTNE: A New Deep Learning Framework For Large Sparse Datasets

More events from Szilard Pafka

Follow organizers to get events picked for you

You might also like...

Sign-In for personalized recommendations

Still looking for the right event?

Explore all events in Santa Monica and filter by date, category, and more to find the perfect fit.