machine learning

Come and have fun with local useRs at SatRday LA on April 6, 2019. This is the first SatRday in LA, and the second in the states. If you have not heard of SatRday, it is a one-day affordable, inclusive, non-profit R conference organized by local R users.

Update_1: Slides for the 4 talks added below (at each talk section separately). Update_2: Photos from the meetup and a short description of the event added below (at the bottom of the post). Update_3: Video recording of the meetup here:

Update: Video recording of the meetup here: ▶ Machine Learning in Production with Szilard Pafka Slides here:

Update: Slides here: Slides: https://speakerdeck.com/szilard/machine-learning-meetup-febr-2017 Our next meetup: No-Bullshit Data Science by Szilard Pafka Abstract: While extracting business value from data has been performed by practitioners for decades, the last several years have seen an unprecedented amount of hype in this field. This hype has created not only unrealistic expectations in results, but also glamour in the usage of the newest tools assumably capable of extraordinary feats. In this talk I will apply the much needed methods of critical thinking and quantitative measurements (that data scientists are supposed to use daily in solving problems for their companies) to assess the capabilities of the most widely used software tools for data science. I will discuss in details two such analyses, one concerning the size of datasets used for analytics and the other one regarding the performance of machine learning software used for supervised learning.

UPDATE: The meetup “sold out” in record time (35 minutes!) from announcing it on our mailing list. There are more than 250 people on the waiting list now. Machine Learning at Google # Come and hear about a range of the work that is being done in Google’s Los Angeles office in the area of machine learning:

Update: This event has ended. Slides are available here. # Update: Video recording by Carl Mullins: # This video is no longer available on YouTube (original link).

XGBoost is a fantastic open source implementation of Gradient Boosting Machines, a general purpose supervised learning method that achieves the highest accuracy on a wide range of datasets in practical applications. Deep learning is all the hype now, but apart from specific domains such as images, speech or text (i.e. problems with higher-level abstractions to be learnt and/or perception problems, where deep learning achieved some remarkable results indeed), it is usually outperformed by Gradient Boosting in a majority of general business domains and supervised learning applications. Proof of this and also because XGBoost has an easy-to-use interface from both R and Python, XGBoost has become a favorite tool in Kaggle competitions. Besides feature engineering, cross-validation and ensembling, XGBoost is a key ingredient for achieving the highest accuracy in many data science competitions and more importantly in practical applications.

XGBoost is a fantastic open source implementation of Gradient Boosting Machines, one of the most accurate general purpose supervised learning algorithms. With interfaces from both R and Python among others, it has become a favorite tool in Kaggle competitions, and it’s also widely used in real-world applications. In this meetup we’ll have a talk from the best person on the topic,Tianqi Chen, main author of XGBoost.

Deep Learning @Whisper talk by Nick Stucky-Mack Video now available, thank you to community member Carl Mullins for providing video services: ▶ Deep Learning @Whisper THU, MAY 12 AT 6 00 PM, VENICE, CA In this talk we will discuss how Whisper uses machine learning for natural language processing tasks. We will focus on how Whisper applies deep learning to moderate content created by its more than 20 million monthly active users. The talk will explore the problem of short text classification, and will cover Whisper’s choices in machine learning models, software development, and hardware.

What has Kaggle learned from 2 million machine learning models? Anthony Goldbloom, founder and CEO of Kaggle UPDATE:Video recording of the talk available here Kaggle is a community of almost 450K data scientists who have built almost 2MM machine learning models to participate in our competitions. Data scientists come to Kaggle to learn, collaborate and develop the state of the art in machine learning. This talk will cover some of the lessons on winning techniques we have learned from the Kaggle community.

Join PyDSLA for a talk by Erin LeDell Intro to Machine Learning with H2O and Python Get your tickets HERE Summary: This workshop will provide an overview of how to use H2O, the scalable open source machine learning library, from Python. The core algorithms of H2O are implemented in Java, however, fully-featured APIs are available in R, Python, Scala, and also through a web interface. The focus of this hands-on workshop will be the “h2o” Python module.

With so much hype about “big data” and the industry pushing for “big data” analytical tools for everyone, the question arises how many people have big data (for analytics) and how many of them really need these tools (which are more complex and often more immature compared to the traditional tools for analytics).

If you’ve never heard of kaggle, I guess the first question I have to ask is “how did you end up on DataScience.LA?” :) Kaggle is one of the foremost machine learning competition websites on the internet. Companies from GE to Criteo and everything in between create competitions. These competitions are in the form of a dataset, ranging from small to relatively large (up to many many gigabytes), a stated goal (such as predicting an event or finding a numeric value) and an evaluation criteria. The Kaggle community has grown in the last few years, so that now competitions often host thousands of participants!

I currently have the need for machine learning tools that can deal with observations of the order of 10 millions in the context of binary classification. That kind of data is a few GBs in size and it fits comfortably nowadays in the RAM of a decent single machine. It is a trivial task for linear models, as there are plenty of open source tools that can train a logistic regression with this amount of data on a single machine in a few seconds, even while using only 1 processor core (many of these tools are single-threaded). Linear models are also the gold standard of large-scale machine learning that can run on clusters, processing very large distributed datasets.

(Posting the slides and video recording from this event is long overdue - our apologies for the delay.) At the September LA Data Science/Machine Learning meetup we had IRIS.TV’s Chief Data Scientist Tom Sullivan and VP of Engineering Joel Spitalnik presenting. IRIS.TV is a video recommendation engine, and in their excellent talk Tom & Joel shared with us their approach to personalization and how they use machine learning to achieve their goals. They also shared the architecture of their systems, including how data is flowing between the various parts (i.e. the ETL pipeline). One thing that popped up as a question and I found interesting is that the usage of R and Python in their company is split 50-50. In fact, while Tom is using mainly R, Joel is using mainly Python. While this is perhaps unusual at a relatively small startup, our survey on software tools used for data science revealed that most data scientists in LA (at least those attending the meetups) are using R or Python, with a roughly equal split/small advantage for R. It just happens that for obvious reasons (such as network effects) that one tool dominates in most companies.

Data Science @ Datapop: The Marketing Knowledge Graph presented by Dr. Ben Zamanzadeh, CTO at DataPop Today’s eCommerce has evolved into a complex multi-dimensional optimization problem. DataPop’s semantic technology and Marketing Knowledge Graph is used to generate Omni-Channel advertising campaigns. This presentation will review our technology stack, machine learning models, discuss what worked what did not work for us and do a live demo of Marketing Knowledge Graph.

Last week, the Python Data Science LA meetup debuted with a fabulous event. It was already much anticipated, with more than 250 people showing interest (RSVP+waiting list), and a lucky 100 converged upon the hip Venice Arts venue to hear our speakers. Our sponsor OpenMail has also done a fantastic job in getting us this venue, bringing food+drinks and taking care of all the details. Thanks OpenMail Team!

I ♥ data. But I also ♥ privacy. Even though I have been running a few data meetups for the past few years, I have managed to keep myself out of public eye - even in my professional life. I must admit, I’m giving a bit away in an recent interview with Data Science Weekly. They had excellent questions to pose, and I’d like to take a moment here to summarize two main points from the interview: my thoughts on when and how much time it took me to became a data scientist and how I feel about the hype in the field today. If you’d like, you can read the full interview here.

I’m pleased to announce the first event of our new Python Data Science LA meetup, started by two of us here at DataScience.LA (our very own Szilard and Eduardo). We’ll have 5 speakers at this event, covering various aspects of the Python Data Science ecosystem (NumPy, SciPy, pandas, matplotlib, scikit-learn, NetworkX, NLTK, IPython notebook, etc.)

This is a re-publication of a blog post from a blog I created not long before we got the idea to start DataScience.LA, and in many aspects it is the genesis of DSLA (the blog’s name, the people getting together - thanks Eduardo, Amelia and Leigh). It was also the spark that lead to the creation of the Python Data Science LA meetup.