Test: sample weights in tabular data

Data

  • XOR Data: xor_data.csv
  • Data configuration: data.txt
  • The weights are set such that the model should learn to accurately predict the first 3 examples. The fourth example has zero weight, so its error does not factor into the loss (and the model will typically assign a 1 as the output). However, the metrics take all examples into account, so we expect loss to go to zero, but the metrics will reflect the error in the last example.

Network

  • Two binary inputs
  • Two hidden layers: 20 and 10 hidden units
  • One binary output
  • Network configuration: network.txt

Training Details

Training Process: Command Line

Used for testing (assuming that the python environment is already active):

./exec.sh

Training Process: SLURM Batch File


This site uses Just the Docs, a documentation theme for Jekyll.