Upsampling of compressed financial time-series data using a jointly trained Vector Quantized Variational Autoencoder neural network
A system and methods for upsampling compressed data using a jointly trained Vector Quantized Variational Autoencoder (VQ-VAE) and neural upsampler. The system compresses input data into a discrete latent space using a VQ-VAE encoder, reconstructs the data using a VQ-VAE decoder, and enhances the reconstructed data using a neural upsampler. The VQ-VAE and neural upsampler are jointly trained using a combined loss function, enabling end-to-end optimization. The system allows for efficient compression and high-quality reconstruction of various data types, including financial time-series, images, audio, video, sensor data, and text. The learned discrete latent space can be explored and manipulated using techniques such as interpolation, extrapolation, and vector arithmetic to generate new or modified data samples. The system finds applications in data storage, transmission, analysis, and generation across multiple domains.
1. A system for upsampling compressed data using a jointly trained vector quantized variational autoencoder (VQ-VAE) neural upsampler, comprising:
a computing device comprising at least a memory and a processor;
a plurality of programming instructions stored in the memory and operable on the processor, wherein the plurality of programming instructions, when operating on the processor, cause the computing device to:
compress input data into a discrete latent representation using a VQ-VAE encoder;
store the compressed representation in a discrete latent space;
reconstruct the compressed data from the latent representation using a VQ-VAE decoder;
enhance the reconstructed data using a neural upsampler to recover information lost during compression;
jointly train the VQ-VAE and neural upsampler by iteratively updating their parameters based on a joint loss function that combines the reconstruction loss of the VQ-VAE and the upsampling loss of the neural upsampler; and
explore and manipulate the discrete latent space learned by the VQ-VAE to generate new or modified data using techniques comprising interpolation, extrapolation, and vector arithmetic.
2. The system of claim 1 , wherein the input data comprises financial time-series data, the financial time-series data comprising stock prices, economic indicators, and other relevant data.
3. The system of claim 1 , wherein the input data comprises one or more of image data, audio data, video data, sensor data, or text data.
4. The system of claim 1 , wherein the joint training allows gradients to flow from the neural upsampler back to the VQ-VAE, enabling end-to-end training of the entire system.
5. The system of claim 1 , wherein the joint loss function takes into account both compression efficiency and reconstruction quality.
6. The system of claim 1 , wherein the VQ-VAE encoder comprises one or more convolutional layers, one or more pooling layers, and one or more fully connected layers.
7. The system of claim 1 , wherein the VQ-VAE decoder comprises one or more transposed convolutional layers, one or more upsampling layers, and one or more fully connected layers.
8. The system of claim 1 , wherein the neural upsampler comprises one or more convolutional layers, one or more residual blocks, and one or more attention mechanisms.
9. The system of claim 1 , wherein the discrete latent space is learned using a vector quantization technique that maps the continuous latent representations to a finite set of discrete codes.
10. The system of claim 1 , wherein the system further comprises a data preprocessing module for normalizing, scaling, and transforming the input data before compression.
11. A method for upsampling compressed data using a jointly trained vector quantized variational autoencoder (VQ-VAE) neural upsampler, comprising the steps of:
compressing input data into a discrete latent representation using a VQ-VAE encoder;
storing the compressed representation in a discrete latent space;
reconstructing the compressed data from the latent representation using a VQ-VAE decoder;
enhancing the reconstructed data using a neural upsampler to recover information lost during compression;
jointly training the VQ-VAE and neural upsampler by iteratively updating their parameters based on a joint loss function that combines the reconstruction loss of the VQ-VAE and the upsampling loss of the neural upsampler; and
exploring and manipulating the discrete latent space learned by the VQ-VAE to generate new or modified data using techniques comprising interpolation, extrapolation, and vector arithmetic.
12. The method of claim 11 , wherein the input data comprises financial time-series data, the financial time-series data comprising stock prices, economic indicators, and other relevant data.
13. The method of claim 11 , wherein the input data comprises one or more of image data, audio data, video data, sensor data, or text data.
14. The method of claim 11 , wherein the joint training allows gradients to flow from the neural upsampler back to the VQ-VAE, enabling end-to-end training of the entire system.
15. The method of claim 11 , wherein the joint loss function takes into account both compression efficiency and reconstruction quality.
16. The method of claim 11 , wherein the VQ-VAE encoder comprises one or more convolutional layers, one or more pooling layers, and one or more fully connected layers.
17. The method of claim 11 , wherein the VQ-VAE decoder comprises one or more transposed convolutional layers, one or more upsampling layers, and one or more fully connected layers.
18. The method of claim 11 , wherein the neural upsampler comprises one or more convolutional layers, one or more residual blocks, and one or more attention mechanisms.
19. The method of claim 11 , wherein the discrete latent space is learned using a vector quantization technique that maps the continuous latent representations to a finite set of discrete codes.
20. The method of claim 11 , further comprising the step of normalizing, scaling, and transforming the input data before compression.