sklearned
Surrogate experiments
Experimental choices
- Choose a skater from timemachines
- Choose an optimizer from humpday
- Choose an embedding from the cube into the space of keras models, such as keras_mostly_linear
- Choose a search strategy and data augmentation strategy
- Choose a forecast horizon k, such as k=1
- Choose an input vector length, such as 80 or 160
- Choose a subset of live data streams from stream listing
- Run the skater to generate training data
- Challenge the existing out-of-sample champion model
The names of experiment files indicate choices made.
M1
Maybe..
brew install cmake protobuf
pip3 install --pre -i https://pypi.anaconda.org/scipy-wheels-nightly/simple numpy
conda install -c apple tensorflow-deps
pip install tensorflow-macos
pip3 install onnx
There might be a really screwed up mess with protobuf. You might have to use the builder.py file from a different version. See
this stack overflow question