Evan Sparks

Results 45 comments of Evan Sparks

I have started some work on this here's a summary of some early results. I am benchmarking imaages of various sizes (given as "bytes") and have three implementations: `whi` which...

Just for posterity here - `whi` `iter` and `sum` measure time taken (in ns) to do each operation. `iterwhile`, and `itersum` are slowdown of `iter/sum` and `iter/while` - that is,...

In the ImageLoaderUtils class we have a function that takes in a filename and produces a label which is dataset specific (e.g. VOC and ImageNet have different labelsMap functions). Right...

I have a strong preference for avoiding "zip" in preference of a logical "join" with several physical implementations - one of which might be zip. I've been giving some thought...

Thanks @tomerk - this is useful stuff. I don't fully grok the use case for the `Concatenate` operator, so let's chat about it in person. We should also discuss semantics...

Can you enumerate the set of types you care about - are they basically, String, Enum, Double? Are there more? I ask because I'm worried about enforcing type-safety if possible...

Spoke with @shivaram and @tomerk about this - it looks like we can support the pattern you're asking for with the Concat operator and assuming the input tables can be...

So the basic idea right now is that each operator to _.concat will take in the entire row (a single object) and that we provide some simple syntax for selecting...

Also, @sjyk - @shivaram mentioned that there may be a .toCaseClass (or something) method on DataFrames coming soon, but I can't find a reference to it, so this may not...