I’ve seen tf.random.set_seed(1234), tf.random.set_seed(22), etc.
Do these values follow any criteria? Or any value is placed and the results are tested.
When setting a value for the random seed in tf.random.set_seed, the criteria followed is to select a seed value that is sufficiently random to ensure that the generated random numbers are unpredictable, but also reproducible.
The seed value should be a non-negative integer, and different seed values will produce different sequences of random numbers. However, setting the same seed value will generate the same sequence of random numbers every time the program is run.
It’s important to note that the seed value only affects the initial state of the random number generator, and not the distribution of the generated numbers itself. In other words, different seed values will still produce random numbers that follow the same distribution, but the sequence of the generated numbers will differ.
When selecting a seed value, it’s recommended to use a value that is unique to the program or experiment, and to avoid using common seed values like 0 or 1. It’s also a good practice to log the seed value used in each experiment, to ensure that the results are reproducible.