Dimension reduction, State, Orientation, and Velocity.
Figuring out TensorFlow documentation and tutorials (with a focus on matrix operations, loading from hadoop, and clustering).
Really basic examples with tiny data sets like linear regression with gradient descent optimizers are EASY. Sessions, variables, placeholders, and other core artifacts all make sense. Across the room Phil’s hair is getting increasingly frizzy as he’s dealing with more complicated examples that are far less straightforward.
Test extraction of Hadoop records
Create TF tensors using Python against HBASE tables to see if the result is performant enough (otherwise recommend we write a MapReduce job to build out a proto file consumed by TF)
Test polar coordinates against client data
See if we can use k-means/DBSCAN against polar coordinates to generate the correct clusters with known data). If we cannot use polar coordinates for dimension reduction, what process is required to implement DBSCAN in TensorFlow?
The artifacts for this sprint’s completion are architecture diagrams and proposal for next sprint’s implementation. I haven’t gotten feedback from the customer about our proposed framework, but it will come up in our end-of-sprint activities. Design path and flow diagram are due on Wednesday.
I did my first 15.2 mile ride today. My everything hurts, and my average speed was way down from yesterday, but I finished.