Improvements

[Dragjon]

• More data in dataset
• Removed 'lr' = 0.001 which was causing problems
• Changed rmsprop --> adam optimiser
• Changed <= threshold to > threshold for the new model
• Some minor changes to code
• Changed threshold to 0.5 (the estimated threshold is not accurate)

[kidscoots101]

Hey Dragjon!

Thanks for your help in improving our algorithm. Really appreciate it :)

Hmmm was just wondering about what the rationale was between switching from RMSProp to Adam Optimiser?

Thanks!

[Dragjon]

I used the Adam optimiser instead as it is known to adapt better to different types of data taking the advantages of both Momentum and RMSProp, usually the first optimiser used is the optimiser as it requires little parameter tuning to achieve results