Small sample test

Ran a small sample test to check if network could train. Used the same 43 neutrinos for training and testing. Wanted to show that it could reduce the loss and hit a very high accuracy.

RMSProp solver used. Parameters:

# In the case of MNIST, we have test batch size 100 and 100 test iterations,
iter_size: 10
# covering the full 10,000 testing images.
test_iter: 40
# Carry out testing every 500 training iterations.
test_interval: 100
# The base learning rate, momentum and the weight decay of the network.
base_lr: 0.00001
momentum: 0.0
weight_decay: 0.0
# The learning rate policy
lr_policy: "inv"
gamma: 0.0002
power: 0.75
# Display every 20 iterations
display: 10
# The maximum number of iterations
max_iter: 100000
# snapshot intermediate results
snapshot: 1000
snapshot_prefix: "snapshot_rmsprop_uburn"
# solver mode: CPU or GPU
solver_mode: GPU
type: "RMSProp"
rms_decay: 0.90

Note that accuracy is for all pixels, which are mostly empty. Next step for confirmation is to write an analysis script, evaluating the accuracy of pixels with actual charge in them.