search

ClimbsRocks / data-formatter

Takes raw csv input and formats it to be ready for neural networks
19 stars 7 forks source link

refactor neural network normalization to work with sparse matrices #30

Closed ClimbsRocks closed 9 years ago

ClimbsRocks commented 9 years ago

should be relatively easy...

convert to csc, get a row, work through that row to normalize it as we normally would we can even consider todense() for that single row. though then we need to think about converting back to sparse, and i'm not sure if that's easy or not, given that we're integrating into an existing sparse array

ClimbsRocks commented 9 years ago

using MaxAbsScaler from the (pre-release) 0.17 version of scikit-learn makes this a breeze.