quantdata.dysanos.grid.torch_drift_changeV2_2#

quantdata.dysanos.grid.torch_drift_changeV2_2.torch_utility(utility='entropy', lmbda=1.0)[source]#
quantdata.dysanos.grid.torch_drift_changeV2_2.entropy(X, lmnda=1.0, dim=0)[source]#
class quantdata.dysanos.grid.torch_drift_changeV2_2.DRNConfig(label='', group='deephedging', user_version='0.2', utility='entropy', lmbda=1.0, delta_hedged=True, pf_config=FeedForwardConfig(activation='Softplus', num_depth=0, num_width=16, use_bias=True), cost=0.0, cost_std=1.0, seed=1231231, block_size=None, action_cap=1.0, learn_action_cap=True, cash_trades=False, boot_strap=1, boot_mode='outer', boot_lmbda=2.0, boot_std_scale=2.0, boot_lmbda_n=False, boot_random=False, _floor_P=1e-08)[source]#

Bases: object

label: str | Callable = ''#
group: str = 'deephedging'#
user_version: str = '0.2'#
utility: str = 'entropy'#
lmbda: float = 1.0#
delta_hedged: bool = True#
pf_config: FeedForwardConfig | None = FeedForwardConfig(activation='Softplus', num_depth=0, num_width=16, use_bias=True)#
cost: float = 0.0#
cost_std: float = 1.0#
seed: int = 1231231#
block_size: int | None = None#
action_cap: float = 1.0#
learn_action_cap: bool = True#
cash_trades: bool = False#
boot_strap: int = 1#
boot_mode: str = 'outer'#
boot_lmbda: float = 2.0#
boot_std_scale: float = 2.0#
boot_lmbda_n: bool = False#
boot_random: bool = False#
property has_some_cost: bool#
property qualified_label: str#
property quantiles: ndarray#
create_boot_strap(env, num_samples, num_steps)[source]#

Returns a bootstrap density for num_samples. The density adds up to num_total_samples (e.g. the baseline is (1,…,1)) which means to compute e.g. a mean use np.mean( boostrap * x ) not np.sum( bootstrap * x )

class quantdata.dysanos.grid.torch_drift_changeV2_2.OutputTracker(epochs=<factory>, loss=<factory>, utility=<factory>, util_mean=<factory>, util_std=<factory>, inst_mean=<factory>, inst_std=<factory>, cost_mean=<factory>, cost_std=<factory>, inst_excess_mean=<factory>, inst_excess_std=<factory>, inst_excess2_mean=<factory>, inst_excess2_std=<factory>, pnl_mean=<factory>, pnl_std=<factory>, raw_pnl_mean=<factory>, raw_pnl_std=<factory>, action_mean=<factory>, action_std=<factory>, action_l1=<factory>, action_l2=<factory>, action_max=<factory>, action_cap=<factory>)[source]#

Bases: object

class quantdata.dysanos.grid.torch_drift_changeV2_2.DRN(env, *, num_xstate, num_expiries, num_strikes, num_steps, config=DRNConfig(label='unknown', group='deephedging', user_version='0.2', utility='entropy', lmbda=1.0, delta_hedged=True, pf_config=FeedForwardConfig(activation='Softplus', num_depth=0, num_width=16, use_bias=True), cost=0.0, cost_std=1.0, seed=1231231, block_size=None, action_cap=1.0, learn_action_cap=True, cash_trades=False, boot_strap=1, boot_mode='outer', boot_lmbda=2.0, boot_std_scale=2.0, boot_lmbda_n=False, boot_random=False, _floor_P=1e-08))[source]#

Bases: BaseModel

effective_boot_lmbda(num_samples)[source]#

Outer-entropy lambda, optionally scaled by sqrt(number of paths).

compute_a_y(*, S_slc, TAU_slc, iS, iTAU, log_strikes, expiries, xstate)[source]#

Compute a,y

compute_cost_std_vector(p)[source]#
compute_bootstrap_normalizers(p)[source]#
compute_block(p, slc, cost_std, normalizer=None, *, utility=True, deriv=False)[source]#
compute_bootstrap_utilities(p, normalizer, cost_std, second=False)[source]#
backward_outer_objective(data)[source]#
__call__(data, *, training=False)[source]#

Call self as a function.

class quantdata.dysanos.grid.torch_drift_changeV2_2.VizDRNConfig(plot_prior_alpha=0.025, cost_act_std=2.0, seed=1231231)[source]#

Bases: object

plot_prior_alpha: float = 0.025#
cost_act_std: float = 2.0#
seed: int = 1231231#
class quantdata.dysanos.grid.torch_drift_changeV2_2.Visualizer(universe, total_epochs, *, num_trn_samples, num_vld_samples, col_size=6, row_size=4, ref_dQdP=None, viz_config, mdl_config)[source]#

Bases: object

compute_devs(dQdS, trn)[source]#

TODO: implement relative and delta hedged

__call__(tracker, trn_tracker, vld_tracker, trn_model_output, vld_model_output, *, final=False)[source]#

Update

quantdata.dysanos.grid.torch_drift_changeV2_2.train(universe, *, ref_dQdP=None, mdl_config=DRNConfig(label='unknown', group='deephedging', user_version='0.2', utility='entropy', lmbda=1.0, delta_hedged=True, pf_config=FeedForwardConfig(activation='Softplus', num_depth=0, num_width=16, use_bias=True), cost=0.0, cost_std=1.0, seed=1231231, block_size=None, action_cap=1.0, learn_action_cap=True, cash_trades=False, boot_strap=1, boot_mode='outer', boot_lmbda=2.0, boot_std_scale=2.0, boot_lmbda_n=False, boot_random=False, _floor_P=1e-08), trn_config=TrainConfig(num_epochs=5000, batch_size=None, val_batch_size=None, pcnt_validation=0.0, lr=0.001, final_lr=1e-05, weight_decay=0.0001, compile=False, clip_grad_norm=None, clip_grad=None, epoch_interval=200, time_interval=120), viz_config=VizDRNConfig(plot_prior_alpha=0.025, cost_act_std=2.0, seed=1231231), restart=False, just_load=False, env=<quantdata.torch.BaseEnvironment object>, verbose=<cdxcore.verbose.Context object>)[source]#

Train model