evaluation.json

{
  "experiment": "Liquid time-constant irregular forecasting",
  "learned_time_constants": {
    "minimum": 0.12157730013132095,
    "mean": 0.45884114503860474,
    "maximum": 0.7155194282531738
  },
  "results": {
    "liquid_time_constant": {
      "parameters": 1887,
      "best_step": 2500,
      "normal_gaps": {
        "rmse": 0.08929167687892914,
        "mae": 0.03883583843708038,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.02,
          0.12
        ]
      },
      "unseen_large_gaps": {
        "rmse": 1.5136159658432007,
        "mae": 1.1862934827804565,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.12,
          0.4
        ]
      }
    },
    "matched_gru": {
      "parameters": 1887,
      "best_step": 2500,
      "normal_gaps": {
        "rmse": 0.08689922839403152,
        "mae": 0.040582284331321716,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.02,
          0.12
        ]
      },
      "unseen_large_gaps": {
        "rmse": 0.27418622374534607,
        "mae": 0.21493969857692719,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.12,
          0.4
        ]
      }
    },
    "matched_rnn": {
      "parameters": 1887,
      "best_step": 2500,
      "normal_gaps": {
        "rmse": 0.08811825513839722,
        "mae": 0.04227612167596817,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.02,
          0.12
        ]
      },
      "unseen_large_gaps": {
        "rmse": 0.2811814546585083,
        "mae": 0.21727129817008972,
        "examples": 2000,
        "sequence_length": 128,
        "delta_time_range": [
          0.12,
          0.4
        ]
      }
    }
  }
}