Learn about two takes on neural networks: Word2vec and TensorFlow

On November 9, join 180 other Pythonistas in diving deeper into neural networks.

PLEASE RSVP HERE (https://ti.to/sf-python/LearnAboutTwoTakesOnNeuralNetworks): https://ti.to/sf-python/LearnAboutTwoTakesOnNeuralNetworks

Lightning talks:

If you'd like to give a lightning talk, sign-up here. (https://docs.google.com/forms/d/1LtV839ktupRboMUSXlXqoJ9lLFvpe-TZtLCf2q6jUpY/viewform?usp=send_form#start=invite)

Talk #1: Word2vec algorithm: made as simple as possible, but no simpler

SUMMARY

This talk will give a Pythonic introduction to the word2vec algorithm. Word2vec, translating words (strings) to vectors (lists of floats), is a relatively new algorithm which has proved to be very useful for making sense of text data. You will gain a conceptual understanding of the algorithm and be empowered to try it out on your favorite collection of text data.

DESCRIPTION

“You shall know a word by the company it keeps” is a common refrain in natural language processing (NLP). word2vec is a simple neural network that learns which words tend to co-occur and embeds the words in a vector space. From these word embeddings, it is possible to use distance measures to compare words, find neighbors by clustering, and add/subtract words to explore relationships between concepts. Actually, word2vec is a general purpose algorithm that allows any sequential data to be encoded into meaningful vectors - including emojis!

BIO

Dr. Brian Spiering is a faculty member at GalvanizeU, which offers a Master of Science in Data Science. His passions are natural language processing (NLP), deep learning, and building data products. He is active in the San Francisco data science community through volunteering and mentoring.

Talk #2: Recurrent neural networks with TensorFlow

SUMMARY

In this presentation you will be introduced to recurrent neural networks and the TensorFlow routines used for this type of machine learning. We will setup a model to classify two types of earthquakes. At the end of the talk you will have learned what an RNN is and when to use it.

DESCRIPTION

Time-series data is seen in many places, from a heart rate monitor to the variations in a star’s brightness. A great tool to model time-series is a recurrent neural network. This presentation will describe this type of model and why it works with this type of data. As an example, we will use the earthquakes dataset from the UCR Time Series Classification Archive. Using TensorFlow, we will set up an RNN model to classify types of earthquakes. In addition, we will use TensorBoard to visualize the steps of the modeling process.

BIO

David Clark has a background in astrophysics, where he used Python extensively to analyze astronomical data. He recently transitioned careers to data science. Currently he is doing consulting for two startups. At Palo Alto Scientific, Inc., he uses the machine learning library TensorFlow to model sensor data from a wearable and infer a runner’s performance. He is also doing work for Quantea, Inc., making a dashboard using the Python libraries Bokeh and Pandas.

Agenda:

6:00p - Check-in and mingle, with pizza and beer provided by our generous sponsor Yelp!

7:05p - Welcome

7:10p - Talk #1 and Q&A

7:50p - Announcements and lightning talks

8:10p - Talk #2 and Q&A

8:50p - More mingling

9:30p - Doors close

**SF Python is run by volunteers aiming to foster the Python Community in the bay area. Please consider making a donation (https://secure.meetup.com/sfpython/contribute/) to SF Python and saying a big thank you to Yelp for providing food, drinks, and the venue for this Wed's meetup.