Systems and methods for deep learning enhanced garbage collection

A method comprising:

• receiving, at a controller of a memory array, a time series dataset including data describing a number of write operations to the memory array and a number of delete operations to the memory array during a time period;
• generating, by the controller, a tensor by encoding the time series data using a plurality of frequencies;
• determining a pattern in the time series data from the tensor;
• generating a forecast of write operations to the memory array based on the pattern;
• determining, based on the forecast, a number of write operations to the memory array at forecasted time interval;
• during the forecasted time interval, comparing a threshold number of write operations with a number of actual write operations to the memory array, wherein the threshold number of write operations is based on the forecasted number of write operations;
• and
• based on a determination that the number of actual write operations is greater than the threshold number of write operations, scheduling a garbage collection (GC) algorithm to run on the memory array and run the GC algorithm to delete data stored to a block of the memory array marked for GC after a write operation on the memory array.

A non-transitory computer-readable medium storing instructions that, when executed on one or more processors, cause the one or more processors to:

• receive, at a controller of a memory array, a time series dataset including data describing a number of write operations to the memory array and a number of delete operations to the memory array during a time period;
• generate, by the controller, a tensor by encoding the time series data using a plurality of frequencies;
• determine a pattern in the time series data from the tensor;
• generate a forecast of write operations and delete operations to the memory array based on the pattern;
• determine, based on the forecast, a number of delete operations to the memory array at a forecasted time interval and a number of write operations to the memory array at the forecasted time interval;
• at the forecasted time interval, compare a threshold number of delete operations with a number of actual delete operations to the memory array and compare a threshold number of write operations with a number of actual write operations to the memory array, wherein the threshold number of delete operations is based on the forecasted number of delete operations and the threshold number of write operations is based on the forecasted number of write operations;
• and
• based on a determination that one or more of:
• the number of actual delete operations is greater than the threshold number of delete operations or the number of actual write operations is greater than the threshold number of write operations, schedule a garbage collection (GC) algorithm to run on the memory array and run the GC algorithm to delete data stored to a block of the memory array marked for GC after a delete operation or a write operation on the memory array.