Cross-domain time series data conversion apparatus, methods, and systems
Apparatus, methods, and systems for cross-domain time series data conversion are disclosed. In an example embodiment, a first time series of a first type of data is received and stored. The first time series of the first type of data is encoded as a first distributed representation for the first type of data. The first distributed representation is converted to a second distributed representation for a second type of data which is different from the first type of data. The second distributed representation for the second type of data is decoded as a second time series of the second type of data.
1. A data conversion device comprising:
at least one memory; and
at least one processor configured to:
obtain first data;
encode, by using an encoder, the first data to a first distributed representation;
generate, by using a converter trained using a neural network and after the encoding of the first data, a second distributed representation using the first distributed representation; and
decode, by using a decoder and after the generation of the second distributed distribution, the second distributed representation to second data; and
output the second data as data generated from the first data, wherein
the first data is image data and the second data is non-image data.
2. The data conversion device according to claim 1 , wherein
the encoder includes an input layer to which the first data is inputted, at least one middle layer, and an output layer outputting the first distributed representation; and
the decoder includes an input layer to which the second distributed representation is inputted, at least one middle layer, and an output layer outputting the second data.
3. The data conversion device according to claim 1 , wherein
the converter is trained by using the first distributed representation for training and the second distributed representation for training.
4. The data conversion device according to claim 1 , wherein
the encoder is trained by using an autoencoder.
5. The data conversion device according to claim 1 , wherein the at least one processor is further configured to:
generate, by using a second converter, a third distributed representation using the first distributed representation; and
generate, by using a third converter, the second distributed representation using the third distributed representation.
6. The data conversion device according to claim 1 , wherein the first data and the second data are of types of data collected by respective data collectors.
7. A data conversion method comprising:
obtaining, by at least one processor, first data;
encoding, by the at least one processor using an encoder, the first data to a first distributed representation;
generating, by the at least one processor using a converter trained using a neural network and after the encoding of the first data, a second distributed representation using the first distributed representation;
decoding, by the at least one processor using a decoder and after the generation of the second distributed distribution, the second distributed representation to second data; and
outputting the second data as data generated from the first data, wherein
the first data is image data and the second data is non-image data.
8. The data conversion method according to claim 7 , wherein
the encoder includes an input layer to which the first data is inputted, at least one middle layer, and an output layer outputting the first distributed representation; and
the decoder includes an input layer to which the second distributed representation is inputted, at least one middle layer, and an output layer outputting the second data.
9. The data conversion method according to claim 7 , wherein
the converter is trained by using the first distributed representation for training and the second distributed representation for training.
10. The data conversion method according to claim 7 , wherein
the encoder is trained by using an autoencoder.
11. The data conversion method according to claim 7 further comprising:
generating, by the at least one processor using a second converter, a third distributed representation using the first distributed representation; and
generating, by the at least one processor using a third converter, the second distributed representation using the third distributed representation.
12. The data conversion method according to claim 7 , wherein the first data and the second data are of types of data collected by respective data collectors.
13. A non-transitory computer readable medium storing a program which, when executed by at least one processor, performs a method comprising:
obtaining first data;
encoding, by using an encoder, the first data to a first distributed representation;
generating, by using a converter trained using a neural network and after the encoding of the first data, a second distributed representation using the first distributed representation; and
decoding, by using a decoder and after the generation of the second distributed distribution, the second distributed representation to second data; and
outputting the second data as data generated from the first data, wherein
the first data is image data and the second data is non-image data.
14. The non-transitory computer readable medium according to claim 13 , wherein
the encoder includes an input layer to which the first data is inputted, at least one middle layer, and an output layer outputting the first distributed representation; and
the decoder includes an input layer to which the second distributed representation is inputted, at least one middle layer, and an output layer outputting the second data.
15. The non-transitory computer readable medium according to claim 13 , wherein
the encoder is trained by using an autoencoder.
16. The non-transitory computer readable medium according to claim 13 , wherein the method further comprises:
generating, by using a second converter, a third distributed representation using the first distributed representation; and
generating, by using a third converter, the second distributed representation using the third distributed representation.
17. The data conversion device according to claim 1 , wherein the decoder is trained by using an autoencoder.
18. The data conversion method according to claim 7 , wherein the decoder is trained by using an autoencoder.
19. The data conversion device according to claim 1 , wherein a dimensionality of the first distributed representation is lower than a dimensionality of the first data.
20. The data conversion method according to claim 7 , wherein a dimensionality of the first distributed representation is lower than a dimensionality of the first data.
21. The non-transitory computer readable medium according to claim 13 , wherein a dimensionality of the first distributed representation is lower than a dimensionality of the first data.
22. The data conversion device according to claim 1 , wherein
the at least one processor is configured to:
generate, by fusing the first data and third data, the first distributed representation, wherein the third data is a different type of data than the first data, and
output the second data as data generated from the first data and the third data.
23. The data conversion method according to claim 7 , wherein
the encoding includes generating the first distributed representation by fusing the first data and third data,
the third data is a different type of data than the first data, and
the second data is outputted as data generated from the first data and the third data.
24. The non-transitory computer readable medium according to claim 13 , wherein
the encoding includes generating the first distributed representation by fusing the first data and third data,
the third data is a different type of data than the first data, and
the second data is outputted as data generated from the first data and the third data.
25. The data conversion device according to claim 1 , wherein
the first data is time-series data, and
the first distributed representation represents time dependencies of the time-series data.
26. The data conversion method according to claim 7 , wherein
the first data is time-series data, and
the first distributed representation represents time dependencies of the time-series data.
27. The non-transitory computer readable medium according to claim 13 , wherein,
the first data is time-series data, and
the first distributed representation represents time dependencies of the time-series data.