No, the sequence in input data is not mandatory however it is mandatory to consider all the possible sets of data. We cannot train partial data set in the first attempt and rest in the second attempt as a system cannot remember the previous dataset.
In order to get the desired output and generalize the network, it is extremely mandatory to randomize the data set this will reduces the chance of overfitting.
Nevertheless, in some cases, it is essential that data must be in sequence or order so that machines can learn correctly and generate the desired output.
Example:
In order to get the desired output and generalize the network, it is extremely mandatory to randomize the data set this will reduces the chance of overfitting.
Nevertheless, in some cases, it is essential that data must be in sequence or order so that machines can learn correctly and generate the desired output.
Example:
- NLP - Natural Language Processing, where each word must be in correct order so that machine can give the correct or desired output.
- Data based on time series
- Audio or Video data
- Stock Market, etc.
0 Comments