NN different results

hello ,why does Neural Network may produce different results when i train the same NN architecture many times even with same tensorflow random , what can i do to make it give me same result to make choice between models

Hi @taha_fawzy_anwar

Well, neural networks might yield varying results during training due to inherent randomness from factors like weight initialization, mini-batch sampling, and optimization methods.

To obtain consistent results for model comparison, ensure to set the same random seed using TensorFlow’s functions. Additionally, control other sources of variability and consider averaging results from multiple runs.

Techniques like regularization, dropout, and batch normalization can also stabilize learning and Cross-validation can provide a more robust estimate of performance.

I hope this help

best regards
elirod

i got it thanks

1 Like