trained models

trained models

trained models

ppo rnn active return with global advantage normalization + ppo rnn liquidity penalty at 5% cash

fix lambda initialization to -0.5

ppo rnn active return with liquidity buffer penalty global adv norm

ppo rnn active return with liquidity buffer penalty (10% cash) experiments

ppo liquidity buffer penalty with lagrangian relaxation

ppo rnn increased ent_coef, l2 weight regularization reward, Curriculum Learning (soft-start the liquidity penalty impl.)

fix fix ppo rnn + base momentum features + active reward + liquidity buffer penalty

ppo rnn active reward base momentum features with lower commition fees

rename models

ppo rnn active return with base momentum features and smaller window length

ppo rnn active return with price feature combos for market trend

initial commit