2021 01 06 Gamma noisy sine - WojciechMigda/TruRL GitHub Wiki
Experiment parameters:
Episodes: 100
max_episode_steps: 200
Memory capacity: 100000
GAMMA: <#####>
NEPOCHS(20)
KBinsDiscretizer({
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{34, -0.300000, 0.300000},
{10, 0.000000, 100.000000},
{10, 0.000000, 100.000000},
{10, 0.000000, 200.000000},})
Scaler({[-50.000000, 50.000000], [0, 10000]})
TsetliniClassifierBitwise({
"threshold": 10000,
"s": 4.000000,
"number_of_regressor_clauses": 3200,
"number_of_states": 127,
"boost_true_positive_feedback": 1,
"random_state": 1,
"n_jobs": 6,
"clause_output_tile_size": 16,
"weighted": true,
"loss_fn": "MSE",
"loss_fn_C1": 0.000000,
"max_weight": 2147483647,
"verbose": false
})
Gym: <TimeLimit<WavyMarketEnv, gen_fn=03_noisy_sine Actions=[<Actions.HOLD: 0>, <Actions.BUY100: 1>, <Actions.SELL100: 2>]>>
All other parameters were the same as in the baseline experiment.
There were 10 separate runs, each consisted of 100 episodes and each episode ran for 200 steps.

Each tested gamma value displayed very similar learning performance. Differences are practically none.

Location: /experiments/2021-01-06_wavy_gamma_noisy
Scripts are versioned in the folder above.