Fixed random seed for reproducible results

[Ruiyang Li]

Hi there,

I'm running a scala program using Keras-like API. I'd like to improve the performance of my deep learning model, but I get different results every time. Is there a way to set a fixed global random seed for reproducibility? Or any idea about how to get reproducible results with BigDL?

Thank you!

Ruiyang

[huangka...@gmail.com]

Hi Ruiyang,

Can you try adding the following to set seed?

```

import com.intel.analytics.bigdl.dllib.utils.RandomGenerator

RandomGenerator.RNG.setSeed(seed)

```

Thanks,

Kai

[Ruiyang Li]

Thank you, Kai! I will try this.

Best,

Ruiyang