Forecasting using fuzzy kernels in convolutional neural networks

A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations for forecasting changes in quantities over time, the operations comprising:

• obtaining training data;
• initializing a first continuous function;
• after obtaining the training data and initializing the first continuous function, training a convolutional neural network to obtain a trained convolutional neural network, wherein the convolutional neural network is trained using the training data that comprises first irregular time intervals over a first period of time, the training data comprises first time series data having first samples of a first quantity of the quantities, the first samples are collected at the first irregular time intervals, and the first irregular time intervals are used-as is during the training of the convolutional neural network without fitting the first irregular time intervals to a predefined time interval structure, wherein the training of the convolutional neural network further comprises
• obtaining a convolutional neural network architecture, the convolutional neural network architecture comprising parameters of the convolutional neural network;
• ingesting the training data into the convolutional neural network;
• performing learning using the convolutional neural network, the first continuous function, and the training data to obtain an output, the output comprising modifications to the parameters of the convolutional neural network and to the first continuous function;
• and
• optimizing the output to obtain the trained convolutional neural network and to generate a fuzzy kernel for the trained convolutional neural network;
• obtaining, after obtaining the trained convolutional neural network, a request for a future value of the first quantity;
• obtaining, in response to obtaining the request, second time series data associated with the first quantity, wherein the second time series data comprises second samples of the first quantity, and the second samples are collected at second irregular time intervals over a second period of time;
• processing the second time series data using the trained convolutional neural network to obtain a prediction without fitting the second irregular time intervals of the second time series data to the predefined time interval structure such that the second irregular time intervals are used as-is to obtain the prediction during the processing of the second time series data by the trained convolution neural network;
• and
• providing the prediction to a downstream consumer for use in providing computer-implemented services.

A data processing system, comprising:

• a processor;
• and
• a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations for forecasting changes in quantities over time, the operations comprising:
• obtaining training data;
• initializing a first continuous function;
• after obtaining the training data and initializing the first continuous function, training a convolutional neural network to obtain a trained convolutional neural network, wherein the convolutional neural network is trained using the training data that comprises first irregular time intervals over a first period of time, the training data comprises first time series data having first samples of a first quantity of the quantities, the first samples are collected at the first irregular time intervals, and the first irregular time intervals are used-as is during the training of the convolutional neural network without fitting the first irregular time intervals to a predefined time interval structure, wherein the training of the convolutional neural network further comprises
• obtaining a convolutional neural network architecture, the convolutional neural network architecture comprising parameters of the convolutional neural network;
• ingesting the training data into the convolutional neural network;
• performing learning using the convolutional neural network, the first continuous function, and the training data to obtain an output, the output comprising modifications to the parameters of the convolutional neural network and to the first continuous function;
• and
• optimizing the output to obtain the trained convolutional neural network and to generate a fuzzy kernel for the trained convolutional neural network;
• obtaining, after obtaining the trained convolutional neural network, a request for a future value of the first quantity;
• obtaining, in response to obtaining the request, second time series data associated with the first quantity, wherein the second time series data comprises second samples of the first quantity, and the second samples are collected at second irregular time intervals over a second period of time;
• processing the second time series data using the trained convolutional neural network to obtain a prediction without fitting the second irregular time intervals of the second time series data to the predefined time interval structure such that the second irregular time intervals are used as-is to obtain the prediction during the processing of the second time series data by the trained convolution neural network;
• and
• providing the prediction to a downstream consumer for use in providing computer-implemented services.