PrincetonUniversity / PsyNeuLink

84%

master: 85%

LAST BUILD ON BRANCH feat/pec-hierarchical

branch: SELECT

CHANGE BRANCH

x

• No branch selected
• 1d31a1dfb80
• 3382-driftonasphere-implementation
• 46afdeb6615
• allclose-clean
• build-ad9a15238a
• build-b48c2b6928
• build-f1294bdcc9
• cleanup
• cleanup_ego_scripts
• condition-names
• dependabot/pip/devel/beartype-lt-0.22.0
• dependabot/pip/devel/fastkde-gte-1.0.24-and-lt-2.1.4
• dependabot/pip/devel/grpcio-lt-1.74.0
• dependabot/pip/devel/matplotlib-gte-3.7.0-and-lt-3.10.6
• dependabot/pip/devel/matplotlib-gte-3.7.0-lt-3.10.4
• dependabot/pip/devel/matplotlib-lt-3.10.1
• dependabot/pip/devel/networkx-lt-3.6
• dependabot/pip/devel/numpy-gte-1.21.0-and-lt-2.2.6
• dependabot/pip/devel/numpy-gte-1.21.0-and-lt-2.3.3
• dependabot/pip/devel/numpy-gte-1.21.0-lt-2.2.6
• dependabot/pip/devel/numpy-gte-1.21.0-lt-2.3.1
• dependabot/pip/devel/numpy-gte-1.21.0-lt-2.3.2
• dependabot/pip/devel/rich-gte-10.7-and-lt-14.4
• dependabot/pip/devel/scipy-gte-1.7.3-lt-1.17
• dependabot/pip/devel/sphinx-autodoc-typehints-lt-3.10.0
• dependabot/pip/devel/sphinx-autodoc-typehints-lt-3.11.0
• dependabot/pip/devel/torch-gte-1.10.0-and-lt-2.7.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.10.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.11.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.12.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.13.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.8.0
• dependabot/pip/devel/torch-gte-1.13.0-and-lt-2.9.0
• devel
• doc-simple-integrator-function-escape-underscore
• doc/IntegratorFunctions/escape-math-underscores
• doc/additional_tutorials
• docs/BasicAndPrimer/rename-encoder-to-ffn
• docs/composition_randomization
• docs/examples/fix-EVC-OCM
• docs/spreadsheet_mods
• feat/add_bias_node
• feat/add_pathway_default_matrix
• feat/autodiff/copy_pytorch_param_to_matrix
• feat/autodiff/get_targets
• feat/autodiff/refactor_target_specs
• feat/autodiff/refactor_target_specs_codex
• feat/autodiff/refactor_target_specs_km
• feat/autodiff/torch_param_specific_lr
• feat/autodiff/torch_param_specific_lr_DEVEL
• feat/autodiff_add_support_for_parameter_learning_rates
• feat/autodiff_save
• feat/autodiff_various
• feat/backprop_fct_with_multi_args
• feat/batch_sequences
• feat/binomial_distort
• feat/collapsingbound
• feat/composition_reset_clear_results
• feat/compositon/additonal_pathway_syntax
• feat/em_composition
• feat/em_composition_EGO
• feat/em_composition_LEARNING
• feat/em_composition_NEW
• feat/em_composition_REFACTOR_LEARNING_MECH
• feat/em_composition_REFACTOR_MEMORY_TEMPLATE
• feat/emcomposition/add_soft_max_args
• feat/emcomposition/assign_field_weights
• feat/emcomposition/softmax_choice_prob
• feat/emcomposition/support_storage_function
• feat/function_kwargs
• feat/functions/normalization
• feat/gru
• feat/integrators/integrator_mech_reset_param
• feat/kaiming-matrix
• feat/lca/standard_output_ports_DEVEL_pec_fix
• feat/lca/standard_output_ports_DEVEL_pec_fix_final
• feat/learn_nested_direct
• feat/learning_nested
• feat/log/nparray_executions
• feat/models/EGO_MDP
• feat/models/EGO_MDP_REFACTOR_ATTENTION
• feat/onehot_tie_index
• feat/pec-dask-mle
• feat/pec-hierarchical
• feat/pec_cond_all
• feat/pec_hist
• feat/pec_optuna_study
• feat/pec_optuna_study_rebased
• feat/reset_functions
• feat/thread_config
• feat/timermechanism
• feat/torch_varseq
• fit2
• fit2-compiled
• fit2-compiled-bugs
• fix-softmax
• fix/_parse_port_spec_port_as_sender_or_recvr
• fix/add_obj_mech_warning
• fix/autodiff_mult_output_ports
• fix/bad_likelihood_warn
• fix/comp_warning_no_afferents
• fix/composition_existing_projections
• fix/contentaddressablememory_empty_entry
• fix/control_allocation_JDC
• fix/default_control_allocation
• fix/drift-on-a-sphere-scalar-noise-mechanism-wrap
• fix/ego_update_figs
• fix/emcomposition/field-index-alignment
• fix/emcomposition_fieldweights_assignment
• fix/emcomposition_learnable_projections
• fix/fhn-output-port-index
• fix/input_port_COMBINE
• fix/input_port_spec_w_mapping_projection
• fix/lccontrolmechanism_objectivemechanism
• fix/linearcomb_jacobian
• fix/mapping_projection_matrix_keyword_assignment
• fix/matrix_transform_l0
• fix/one_hot_max_indicator
• fix/onehot-prob-cumsum-edge
• fix/output_port_execution_count_v2
• fix/pec_exec_controller
• fix/pec_inputs_mapping_bug
• fix/pec_seed
• fix/projections/dual_projections_between_mechs
• fix/safe_equals
• fix/softmax_and_relu_derivatives
• fix/stab_flex_pec_bug
• fix/standard_output_ports_calculate
• fix_optuna_rng
• ipynp/502B
• ipynp/502B-correct-add-noise-function
• ipynp/502B-correct-use-summed-matrix
• ipynp/502B-fix-lab1-x-to-X
• ipynp/502B-use-summed-matrix
• jongkees2023
• jvesely/ci
• jvesely/ci-enabled
• jvesely/custom-runners
• jvesely/deps/torch
• jvesely/fix/docstyle
• jvesely/macos-self
• jvesely/test/worksteal
• master
• mode/add_botvinick_conflict_monitoring
• model/fix-botvinick-conflict-monitoring
• model/fix-cohen-huston
• models-update-docs
• models/ego/bandit
• models/ego/csw
• models/ego/use_emcomposition
• models/nback
• models/nback_CAM_FFN
• models/rumelhart-script-fix-targets
• nback
• origin/dependabot/pip/devel/matplotlib-lt-3.10.1
• parameter-bool-as-number
• patch/autodiff/docs_re_modules
• patch/autodiff/pytorch_custom_optimization
• patch/autodiff_pnl_showgraph
• patch/autodiffcomposition-show-graph-jupyter
• patch/bias_nodes
• patch/composition/duplicate_target_warnings
• patch/composition/objective_mech_as_output_node
• patch/convert_to_list
• patch/em_memory_template
• patch/emcomposition/field_memories
• patch/emcomposition/field_memory_indices
• patch/emcomposition/forbid_add_projection
• patch/fix_autodiff_pytorch_dependency
• patch/fix_show_graph_nested_node_to_nested_node
• patch/fixes_for_KADMLP
• patch/get_composition_for_node
• patch/get_composition_for_node_fix
• patch/grucomposition
• patch/lcamechanism/make_matrix_params_nonmodulable
• patch/nested_comps
• patch/port/enhance_error_for_bad_mod_param
• patch/transferfunctions/logistic_offset
• patch/transformfunctions/MatrixTransform
• pec_doc_fixes
• pec_optuna
• py310_fixes
• pytorch_batch
• refactor/EGO_move_figs
• refactor/autodiff/track_pnl
• refactor/autodiff_and_emcomposition
• refactor/composition/mech_constructor_callback
• refactor/emcomposition/fields_class
• refactor/emcomposition_field_handling
• refactor/emcomposition_mechs_and_concatenate
• refactor/emcomposition_none_fieldweights
• refactor/emcomposition_softmax_after_dot
• refactor/emcomposition_tranfser->processingmech
• refactor/exec_mode_pytorch
• refactor/exec_mode_pytorch_FULL
• refactor/execution_mode_COMPILED
• refactor/lccontrolmechanism_ctl_signals
• refactor/learning_pathways_using_ports
• refactor/linearmatrix_>matrixtransform
• refactor/linearmatrix_matrixtransform
• refactor/matrixtransform_pytorch
• refactor/onehot_options
• refactor/pytorchwrappers/execute_node
• refactor/transferfunction/scale_and_offset
• remove_autograd
• remove_typecheck
• revert-2533-codeql
• revert-2751-master-devel
• tentative-learning-branch
• test/autodiff_without_torch
• test_ci
• test_new_runners

Build # 28814065630

Build Type

push

github

Committed by aryan-pan

Commit Message

Prototype neural likelihood estimation against the KDE simulation likelihood

nle/: MNLE-style mixture density network over (parameters, trial
features) -> p(choice, rt), trained on simulator draws, plus an exact
Navarro-Fuss reference density. On the DDM benchmark, EM with the exact
or neural likelihood matches full-Bayes HSSM per-subject accuracy (RMSE
0.084 vs 0.086), attributing the KDE-EM gap to KDE noise rather than the
Laplace EM, at ~60x less wall time. stabflex_nle extends the test to the
stability-flexibility composition: training data generated by sweeping
theta through a worker-cached PEC (rebuilding a composition per config
leaks PNL/LLVM state until dask pauses the workers) with the sequential
gen -> train -> em pipeline in the SLURM script.

Coverage Stats

11060 of 14440 branches covered (76.59%)

Branch coverage included in aggregate %.

37807 of 43800 relevant lines covered (86.32%)

0.86 hits per line