Progress is infinitesimal most of the time: how do we measure it? Do we add to progress combinatorially? Are relationships between highly-connected data, people, ideas, products, and services more crucial to understanding the big picture than knowing each element alone? How can we leverage technology to not only understand the world, but make the world a better place?
Let's hack the future.
I have used Apache Spark extensively for data science projects, particularly making use of MLLib for doing supervised and unsupervised learning on large, sparse data formats stored in HDFS. I tend to use both the Scala and Python APIs, depending on how fast I want to iterate and also where the current functionality is exposed.
The new dataframe abstraction is great, and works well with existing Python libraries when using the PySpark API, such as Pandas and Graphlab-Create.
This summer I have given three public talks about doing Deep Learning using Python. I am familiar with several Python-related deep learning interfaces, including, but not limited to, caffe, Theano, Torch, Graphlab-Create, Nervana Sys, etc.
Additionally, I am studying the theoretical properties of different kinds of neural networks and am applying this knowledge to analyze interesting data sets and solve interesting data problems, such as computer vision and transfer learning.
Slides from one of my talks:
Video from one of my talks:
Very impressed with Alex's work here. He's really making an effort to help me with my project.
Sep 30, 2015
Really like Alex's thoughtful approach to the problem. Enjoy the fact that he is clearly passionate about machine learning and up-to-date with the literature. Also really like that he is willing to put in some preparation for our next session. Impressed.