Configuration is given as a JSON file. We show each field below:


Specify unlearning strategy from below two.

Value Method
"random" Delete data randomly
"lru" Delete data upon Least-Recently-Updated strategy

Specify parameters for the algorithm.


Specify the upper-limit of data quantity. The smaller it is, the less memory usage, the more fast, the less accurately.

  • range of value: 0 < max_size < 2147483647

Specify pattern of IDs to exclude from unlearning. Patterns can be specified in the same format as in key of rules in Data Conversion. This parameter is only available when using lru. This parameter is optional. If not specified, no IDs are excluded from unlearning. (String)


Specify random seed when you use random for unlearning strategy. If you set the same seed and do the same learn, you will get the same result. It may be useful when you examine the result repeatedly. If not specified, system clock is used as seed parameter. So you will get different result each experiment.

  • range of value: 0 <= seed <= \(2^{32} - 1\)

unlearning based on Least-Recently-Used(LRU)

  "unlearner" : "lru",
  "unlearner_parameter" : {
    "max_size": 16777216

unlearning based on random

  "unlearner" : "random",
  "unlearner_parameter" : {
    "max_size": 16777216,
    "seed": 9193