Apr-24-2022, 05:06 AM
Hello,
I am looking for a field of AI to explore that will be able to help me generate a sequence of numbers from existing sequences of numbers. For example the data i will have will be groupings of numbers such as:
{65, 68, 69, 72, 64, 65}
{72, 74, 69, 68 , 66, 65}
{65, 67, 65, 64, 62, 60}
And i would like to find a process that could take in many sequences of these numbers, then output different sequences based on the sequences given. For example an output would be:
{65, 67, 69, 72, 64, 62} (purely an example no logic to ordering of numbers)
During my initial searching i have came across Recurrent Neural Networks and more specifically LSTM's. Under this category i have came across Sequence to Sequence models too.
However my understanding is that these act more to convert sequence of type x to sequence of type y. For example converting a sentence in french to english, or decoding a message.
I am after something that can look at multiple sequences, then generate different iterations of the same type of sequence. So it would look at many iterations of sequence of type x. Then given an input, for example a starting point, or sequence length, it would generate a sequence of type x based on all the sequences of type x it has been given. It would need to generate different sequences rather than the same thing each time. (I understand how varied these are would depend on the size of the data set)
Does anyone know if this sort of things exists? Or can anyone offer advice on what type of AI i should be looking into for this type of thing? As AI is such a big field i would like to narrow my research where possible. I'm hoping that i can test out a few algorithms, gain a basic understanding of them, and then use them as a 'black box' as i have limited time to dedicate to studying AI. However I am aware that i may not be able to get away with this!
I am looking for a field of AI to explore that will be able to help me generate a sequence of numbers from existing sequences of numbers. For example the data i will have will be groupings of numbers such as:
{65, 68, 69, 72, 64, 65}
{72, 74, 69, 68 , 66, 65}
{65, 67, 65, 64, 62, 60}
And i would like to find a process that could take in many sequences of these numbers, then output different sequences based on the sequences given. For example an output would be:
{65, 67, 69, 72, 64, 62} (purely an example no logic to ordering of numbers)
During my initial searching i have came across Recurrent Neural Networks and more specifically LSTM's. Under this category i have came across Sequence to Sequence models too.
However my understanding is that these act more to convert sequence of type x to sequence of type y. For example converting a sentence in french to english, or decoding a message.
I am after something that can look at multiple sequences, then generate different iterations of the same type of sequence. So it would look at many iterations of sequence of type x. Then given an input, for example a starting point, or sequence length, it would generate a sequence of type x based on all the sequences of type x it has been given. It would need to generate different sequences rather than the same thing each time. (I understand how varied these are would depend on the size of the data set)
Does anyone know if this sort of things exists? Or can anyone offer advice on what type of AI i should be looking into for this type of thing? As AI is such a big field i would like to narrow my research where possible. I'm hoping that i can test out a few algorithms, gain a basic understanding of them, and then use them as a 'black box' as i have limited time to dedicate to studying AI. However I am aware that i may not be able to get away with this!