quantdata.dysanos.grid.gendysanos.case_c_partition#

Parallel fixed-notional direction models for conditional-arbitrage searches.

class quantdata.dysanos.grid.gendysanos.case_c_partition.GatedParallelFeedForwardDirections(env, *, config, state_mean, state_whitener)[source]#

Bases: _ParallelDirectionBase

Independent statewise directions with a learned zero-to-one activity.

initialize_output_layers()[source]#

Start from neutral directions with small, recoverable activity.

forward(state)[source]#

Return unit directions multiplied by learned statewise activity.

class quantdata.dysanos.grid.gendysanos.case_c_partition.LearnedPartitionDirections(env, *, config, state_mean, state_whitener)[source]#

Bases: _ParallelDirectionBase

Learned soft state partitions with independent local portfolio directions.

initialize_partition()[source]#

Start with a uniform router and distinct random feasible directions.

project_parameters_()[source]#
responsibilities(factor)[source]#

Return learned state-only soft assignments and their logits.

expert_actions()[source]#

Return local directions with shape (N,K,Q,E,I).

forward(state)[source]#

Return interpolated actions, responsibilities, and router logits.

class quantdata.dysanos.grid.gendysanos.case_c_partition.ParallelDirectionConfig(*, label, source_cache_uid, architecture, state_features, notionals, partitions=16, kernels=64, deterministic_time=0, conditional_time=6, expiries=7, instruments_per_expiry=21, learn_oce_translation=False, action_radius=1.0, group='', user_version='0.1', network_config=FeedForwardConfig(activation='SiLU', num_depth=2, num_width=96, use_bias=True))[source]#

Bases: object

Logical geometry shared by the conditional-direction models.

partitions: int = 16#
kernels: int = 64#
deterministic_time: int = 0#
conditional_time: int = 6#
expiries: int = 7#
instruments_per_expiry: int = 21#
learn_oce_translation: bool = False#
action_radius: float = 1.0#
group: str = ''#
user_version: str = '0.1'#
network_config: FeedForwardConfig = FeedForwardConfig(activation='SiLU', num_depth=2, num_width=96, use_bias=True)#
property action_components: int#
property filename: str#
class quantdata.dysanos.grid.gendysanos.case_c_partition.ParallelFeedForwardDirections(env, *, config, state_mean, state_whitener)[source]#

Bases: _ParallelDirectionBase

Independent feed-forward conditional directions for every notional.

initialize_output_layers()[source]#

Cold-start every notional at the same neutral unit-gross direction.

forward(state)[source]#

Return actions with shape (S,N,2,Q,E,I) and raw logits.

class quantdata.dysanos.grid.gendysanos.case_c_partition.ParallelKernelDirections(env, *, config, state_mean, state_whitener, centers, inverse_bandwidth)[source]#

Bases: _ParallelDirectionBase

Independent local Gaussian-kernel portfolios for every direction head.

initialize_kernel()[source]#

Cold-start distinct local directions with small recoverable activity.

project_parameters_()[source]#
forward(state)[source]#

Return Gaussian-kernel mixtures and their normalized weights.

class quantdata.dysanos.grid.gendysanos.case_c_partition.ParallelRadialFeedForwardDirections(env, *, config, state_mean, state_whitener)[source]#

Bases: _ParallelDirectionBase

Independent zero-capable feed-forward actions for every objective head.

The network emits one signed position per instrument. Dividing by one plus its gross norm keeps the action strictly inside the unit L1 ball while retaining the identity derivative at the zero action. This avoids both a no-trade softmax coordinate and a hard clipping boundary.

initialize_zero_action()[source]#

Start every deterministic and conditional head at no trade.

forward(state)[source]#

Return radial unit-ball actions and unconstrained signed outputs.

quantdata.dysanos.grid.gendysanos.case_c_partition.project_rows_to_l1_ball_(value)[source]#

Euclidean-project direct nonnegative rows onto the unit L1 ball.

quantdata.dysanos.grid.gendysanos.case_c_partition.project_rows_to_simplex_(value)[source]#

Euclidean-project every final-axis row onto the probability simplex.

quantdata.dysanos.grid.gendysanos.case_c_partition.unit_gross_long_short(logits, *, expiries, instruments_per_expiry)[source]#

Map unconstrained logits to separate nonnegative unit-gross positions.