Question about how to maximize the search space

[christofer-f]

I am trying to understand how to get the optimal neural network for a given pipeline.
I assume that one step is to maximize the search space.

In example_custom_configuration_space.py row 59:

configuration = estimator.get_search_space(dataset).get_default_configuration()

The call get_default_configuration() only gets a small subset of the complete search space.
It would be convenient with a function get_full_configuration().

In file: example_custom_configuration_space.py, I can see how to modify the search space.

   def get_search_space_updates():
        updates = HyperparameterSearchSpaceUpdates()
        updates.append(
    
        node_name="data_loader",
            hyperparameter="batch_size",
            value_range=[16, 512],
            default_value=32,
        )

Question:
Should I iterate over the config space to generate python-code that I can paste in a function as above? Is there an easier way?

Like this row:
network_backbone:MLPBackbone:dropout_12 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 11

How can I add this into the HyperparameterSearchSpaceUpdates() function?

If I add hyperparameters about a specific network_backbone type. Do I need to include all the components like this?

api = TabularRegressionTask(
        search_space_updates=get_search_space_updates(),
        include_components={
            "network_backbone": [
                "MLPBackbone",
                "ResNetBackbone",
                "ShapedMLPBackbone",
                "ShapedResNetBackbone",
            ],
        },
    )

The full config space is attached below in pcs_new format.

coalescer:__choice__ categorical {MinorityCoalescer, NoCoalescer} [NoCoalescer]
data_loader:batch_size integer [32, 320] [64]
encoder:__choice__ categorical {NoEncoder, OneHotEncoder} [OneHotEncoder]
feature_preprocessor:__choice__ categorical {NoFeaturePreprocessor} [NoFeaturePreprocessor]
lr_scheduler:__choice__ categorical {CosineAnnealingLR, CosineAnnealingWarmRestarts, CyclicLR, ExponentialLR, NoScheduler, ReduceLROnPlateau, StepLR} [ReduceLROnPlateau]
network_backbone:__choice__ categorical {MLPBackbone, ResNetBackbone, ShapedMLPBackbone, ShapedResNetBackbone} [ShapedMLPBackbone]
network_embedding:__choice__ categorical {LearnedEntityEmbedding, NoEmbedding} [NoEmbedding]
network_head:__choice__ categorical {fully_connected} [fully_connected]
network_init:__choice__ categorical {KaimingInit, NoInit, OrthogonalInit, SparseInit, XavierInit} [XavierInit]
optimizer:__choice__ categorical {AdamOptimizer, AdamWOptimizer, RMSpropOptimizer, SGDOptimizer} [AdamOptimizer]
scaler:__choice__ categorical {NoScaler} [NoScaler]
trainer:__choice__ categorical {MixUpTrainer, StandardTrainer} [StandardTrainer]
coalescer:MinorityCoalescer:min_frac real [0.0001, 0.5] [0.01]
lr_scheduler:CosineAnnealingLR:T_max integer [10, 500] [200]
lr_scheduler:CosineAnnealingWarmRestarts:T_0 integer [1, 20] [1]
lr_scheduler:CosineAnnealingWarmRestarts:T_mult real [1.0, 2.0] [1.0]
lr_scheduler:CyclicLR:base_lr real [1e-06, 0.1] [0.01]
lr_scheduler:CyclicLR:max_lr real [0.001, 0.1] [0.1]
lr_scheduler:CyclicLR:mode categorical {triangular, triangular2, exp_range} [triangular]
lr_scheduler:CyclicLR:step_size_up integer [1000, 4000] [2000]
lr_scheduler:ExponentialLR:gamma real [0.7, 0.9999] [0.9]
lr_scheduler:ReduceLROnPlateau:factor real [0.01, 0.9] [0.1]
lr_scheduler:ReduceLROnPlateau:mode categorical {min, max} [min]
lr_scheduler:ReduceLROnPlateau:patience integer [5, 20] [10]
lr_scheduler:StepLR:gamma real [0.001, 0.9] [0.1]
lr_scheduler:StepLR:step_size integer [1, 10] [5]
network_backbone:MLPBackbone:activation categorical {relu, tanh, sigmoid} [relu]
network_backbone:MLPBackbone:num_groups integer [1, 15] [5]
network_backbone:MLPBackbone:num_units_1 integer [10, 1024] [200]
network_backbone:MLPBackbone:use_dropout categorical {True, False} [False]
network_backbone:ResNetBackbone:activation categorical {relu, tanh, sigmoid} [relu]
network_backbone:ResNetBackbone:blocks_per_group_0 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_1 integer [1, 4] [2]
network_backbone:ResNetBackbone:num_groups integer [1, 15] [5]
network_backbone:ResNetBackbone:num_units_0 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_1 integer [10, 1024] [200]
network_backbone:ResNetBackbone:use_dropout categorical {True, False} [False]
network_backbone:ResNetBackbone:use_shake_drop categorical {True, False} [True]
network_backbone:ResNetBackbone:use_shake_shake categorical {True, False} [True]
network_backbone:ShapedMLPBackbone:activation categorical {relu, tanh, sigmoid} [relu]
network_backbone:ShapedMLPBackbone:max_units integer [10, 1024] [200]
network_backbone:ShapedMLPBackbone:mlp_shape categorical {funnel, long_funnel, diamond, hexagon, brick, triangle, stairs} [funnel]
network_backbone:ShapedMLPBackbone:num_groups integer [1, 15] [5]
network_backbone:ShapedMLPBackbone:output_dim integer [10, 1024] [200]
network_backbone:ShapedMLPBackbone:use_dropout categorical {True, False} [False]
network_backbone:ShapedResNetBackbone:activation categorical {relu, tanh, sigmoid} [relu]
network_backbone:ShapedResNetBackbone:blocks_per_group integer [1, 4] [2]
network_backbone:ShapedResNetBackbone:max_units integer [10, 1024] [200]
network_backbone:ShapedResNetBackbone:num_groups integer [1, 15] [5]
network_backbone:ShapedResNetBackbone:output_dim integer [10, 1024] [200]
network_backbone:ShapedResNetBackbone:resnet_shape categorical {funnel, long_funnel, diamond, hexagon, brick, triangle, stairs} [funnel]
network_backbone:ShapedResNetBackbone:use_dropout categorical {True, False} [False]
network_backbone:ShapedResNetBackbone:use_shake_drop categorical {True, False} [True]
network_backbone:ShapedResNetBackbone:use_shake_shake categorical {True, False} [True]
network_embedding:LearnedEntityEmbedding:dimension_reduction_0 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_1 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_10 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_11 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_12 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_13 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_14 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_15 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_16 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_17 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_18 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_19 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_2 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_20 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_21 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_22 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_23 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_24 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_25 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_26 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_27 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_28 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_29 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_3 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_30 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_31 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_32 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_33 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_34 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_35 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_4 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_5 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_6 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_7 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_8 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:dimension_reduction_9 real [0.0, 1.0] [0.5]
network_embedding:LearnedEntityEmbedding:min_unique_values_for_embedding integer [3, 7] [5]log
network_head:fully_connected:num_layers integer [1, 4] [2]
network_init:KaimingInit:bias_strategy categorical {Zero, Normal} [Normal]
network_init:NoInit:bias_strategy categorical {Zero, Normal} [Normal]
network_init:OrthogonalInit:bias_strategy categorical {Zero, Normal} [Normal]
network_init:SparseInit:bias_strategy categorical {Zero, Normal} [Normal]
network_init:XavierInit:bias_strategy categorical {Zero, Normal} [Normal]
optimizer:AdamOptimizer:beta1 real [0.85, 0.999] [0.9]
optimizer:AdamOptimizer:beta2 real [0.9, 0.9999] [0.9]
optimizer:AdamOptimizer:lr real [1e-05, 0.1] [0.01]log
optimizer:AdamOptimizer:weight_decay real [0.0, 0.1] [0.0]
optimizer:AdamWOptimizer:beta1 real [0.85, 0.999] [0.9]
optimizer:AdamWOptimizer:beta2 real [0.9, 0.9999] [0.9]
optimizer:AdamWOptimizer:lr real [1e-05, 0.1] [0.01]log
optimizer:AdamWOptimizer:weight_decay real [0.0, 0.1] [0.0]
optimizer:RMSpropOptimizer:alpha real [0.1, 0.99] [0.99]
optimizer:RMSpropOptimizer:lr real [1e-05, 0.1] [0.01]log
optimizer:RMSpropOptimizer:momentum real [0.0, 0.99] [0.0]
optimizer:RMSpropOptimizer:weight_decay real [0.0, 0.1] [0.0]
optimizer:SGDOptimizer:lr real [1e-05, 0.1] [0.01]log
optimizer:SGDOptimizer:momentum real [0.0, 0.99] [0.0]
optimizer:SGDOptimizer:weight_decay real [0.0, 0.1] [0.0]
trainer:MixUpTrainer:alpha real [0.0, 1.0] [0.2]
trainer:MixUpTrainer:weighted_loss categorical {True, False} [True]
trainer:StandardTrainer:weighted_loss categorical {True, False} [True]
network_backbone:MLPBackbone:dropout_1 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_10 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_11 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_12 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_13 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_14 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_15 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_2 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_3 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_4 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_5 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_6 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_7 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_8 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:dropout_9 real [0.0, 0.8] [0.5]
network_backbone:MLPBackbone:num_units_10 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_11 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_12 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_13 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_14 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_15 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_2 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_3 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_4 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_5 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_6 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_7 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_8 integer [10, 1024] [200]
network_backbone:MLPBackbone:num_units_9 integer [10, 1024] [200]
network_backbone:ResNetBackbone:blocks_per_group_10 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_11 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_12 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_13 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_14 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_15 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_2 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_3 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_4 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_5 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_6 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_7 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_8 integer [1, 4] [2]
network_backbone:ResNetBackbone:blocks_per_group_9 integer [1, 4] [2]
network_backbone:ResNetBackbone:dropout_0 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_1 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_10 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_11 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_12 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_13 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_14 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_15 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_2 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_3 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_4 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_5 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_6 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_7 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_8 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:dropout_9 real [0.0, 0.8] [0.5]
network_backbone:ResNetBackbone:max_shake_drop_probability real [0.0, 1.0] [0.5]
network_backbone:ResNetBackbone:num_units_10 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_11 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_12 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_13 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_14 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_15 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_2 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_3 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_4 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_5 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_6 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_7 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_8 integer [10, 1024] [200]
network_backbone:ResNetBackbone:num_units_9 integer [10, 1024] [200]
network_backbone:ShapedMLPBackbone:max_dropout real [0.0, 1.0] [0.5]
network_backbone:ShapedResNetBackbone:max_dropout real [0.0, 0.8] [0.5]
network_backbone:ShapedResNetBackbone:max_shake_drop_probability real [0.0, 1.0] [0.5]
network_head:fully_connected:activation categorical {relu, tanh, sigmoid} [relu]
network_head:fully_connected:units_layer_1 integer [64, 512] [128]
network_head:fully_connected:units_layer_2 integer [64, 512] [128]
network_head:fully_connected:units_layer_3 integer [64, 512] [128]
network_head:fully_connected:units_layer_4 integer [64, 512] [128]

coalescer:MinorityCoalescer:min_frac | coalescer:__choice__ == MinorityCoalescer
lr_scheduler:CosineAnnealingLR:T_max | lr_scheduler:__choice__ == CosineAnnealingLR
lr_scheduler:CosineAnnealingWarmRestarts:T_0 | lr_scheduler:__choice__ == CosineAnnealingWarmRestarts
lr_scheduler:CosineAnnealingWarmRestarts:T_mult | lr_scheduler:__choice__ == CosineAnnealingWarmRestarts
lr_scheduler:CyclicLR:base_lr | lr_scheduler:__choice__ == CyclicLR
lr_scheduler:CyclicLR:max_lr | lr_scheduler:__choice__ == CyclicLR
lr_scheduler:CyclicLR:mode | lr_scheduler:__choice__ == CyclicLR
lr_scheduler:CyclicLR:step_size_up | lr_scheduler:__choice__ == CyclicLR
lr_scheduler:ExponentialLR:gamma | lr_scheduler:__choice__ == ExponentialLR
lr_scheduler:ReduceLROnPlateau:factor | lr_scheduler:__choice__ == ReduceLROnPlateau
lr_scheduler:ReduceLROnPlateau:mode | lr_scheduler:__choice__ == ReduceLROnPlateau
lr_scheduler:ReduceLROnPlateau:patience | lr_scheduler:__choice__ == ReduceLROnPlateau
lr_scheduler:StepLR:gamma | lr_scheduler:__choice__ == StepLR
lr_scheduler:StepLR:step_size | lr_scheduler:__choice__ == StepLR
network_backbone:MLPBackbone:activation | network_backbone:__choice__ == MLPBackbone
network_backbone:MLPBackbone:num_groups | network_backbone:__choice__ == MLPBackbone
network_backbone:MLPBackbone:num_units_1 | network_backbone:__choice__ == MLPBackbone
network_backbone:MLPBackbone:use_dropout | network_backbone:__choice__ == MLPBackbone
network_backbone:ResNetBackbone:activation | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:blocks_per_group_0 | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:blocks_per_group_1 | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:num_groups | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:num_units_0 | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:num_units_1 | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:use_dropout | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:use_shake_drop | network_backbone:__choice__ == ResNetBackbone
network_backbone:ResNetBackbone:use_shake_shake | network_backbone:__choice__ == ResNetBackbone
network_backbone:ShapedMLPBackbone:activation | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedMLPBackbone:max_units | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedMLPBackbone:mlp_shape | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedMLPBackbone:num_groups | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedMLPBackbone:output_dim | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedMLPBackbone:use_dropout | network_backbone:__choice__ == ShapedMLPBackbone
network_backbone:ShapedResNetBackbone:activation | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:blocks_per_group | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:max_units | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:num_groups | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:output_dim | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:resnet_shape | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:use_dropout | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:use_shake_drop | network_backbone:__choice__ == ShapedResNetBackbone
network_backbone:ShapedResNetBackbone:use_shake_shake | network_backbone:__choice__ == ShapedResNetBackbone
network_embedding:LearnedEntityEmbedding:dimension_reduction_0 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_1 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_10 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_11 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_12 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_13 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_14 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_15 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_16 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_17 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_18 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_19 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_2 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_20 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_21 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_22 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_23 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_24 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_25 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_26 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_27 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_28 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_29 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_3 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_30 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_31 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_32 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_33 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_34 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_35 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_4 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_5 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_6 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_7 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_8 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:dimension_reduction_9 | network_embedding:__choice__ == LearnedEntityEmbedding
network_embedding:LearnedEntityEmbedding:min_unique_values_for_embedding | network_embedding:__choice__ == LearnedEntityEmbedding
network_head:fully_connected:num_layers | network_head:__choice__ == fully_connected
network_init:KaimingInit:bias_strategy | network_init:__choice__ == KaimingInit
network_init:NoInit:bias_strategy | network_init:__choice__ == NoInit
network_init:OrthogonalInit:bias_strategy | network_init:__choice__ == OrthogonalInit
network_init:SparseInit:bias_strategy | network_init:__choice__ == SparseInit
network_init:XavierInit:bias_strategy | network_init:__choice__ == XavierInit
optimizer:AdamOptimizer:beta1 | optimizer:__choice__ == AdamOptimizer
optimizer:AdamOptimizer:beta2 | optimizer:__choice__ == AdamOptimizer
optimizer:AdamOptimizer:lr | optimizer:__choice__ == AdamOptimizer
optimizer:AdamOptimizer:weight_decay | optimizer:__choice__ == AdamOptimizer
optimizer:AdamWOptimizer:beta1 | optimizer:__choice__ == AdamWOptimizer
optimizer:AdamWOptimizer:beta2 | optimizer:__choice__ == AdamWOptimizer
optimizer:AdamWOptimizer:lr | optimizer:__choice__ == AdamWOptimizer
optimizer:AdamWOptimizer:weight_decay | optimizer:__choice__ == AdamWOptimizer
optimizer:RMSpropOptimizer:alpha | optimizer:__choice__ == RMSpropOptimizer
optimizer:RMSpropOptimizer:lr | optimizer:__choice__ == RMSpropOptimizer
optimizer:RMSpropOptimizer:momentum | optimizer:__choice__ == RMSpropOptimizer
optimizer:RMSpropOptimizer:weight_decay | optimizer:__choice__ == RMSpropOptimizer
optimizer:SGDOptimizer:lr | optimizer:__choice__ == SGDOptimizer
optimizer:SGDOptimizer:momentum | optimizer:__choice__ == SGDOptimizer
optimizer:SGDOptimizer:weight_decay | optimizer:__choice__ == SGDOptimizer
trainer:MixUpTrainer:alpha | trainer:__choice__ == MixUpTrainer
trainer:MixUpTrainer:weighted_loss | trainer:__choice__ == MixUpTrainer
trainer:StandardTrainer:weighted_loss | trainer:__choice__ == StandardTrainer
network_backbone:MLPBackbone:dropout_10 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 9
network_backbone:MLPBackbone:dropout_11 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 10
network_backbone:MLPBackbone:dropout_12 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 11
network_backbone:MLPBackbone:dropout_13 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 12
network_backbone:MLPBackbone:dropout_14 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 13
network_backbone:MLPBackbone:dropout_15 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 14
network_backbone:MLPBackbone:dropout_2 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 1
network_backbone:MLPBackbone:dropout_3 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 2
network_backbone:MLPBackbone:dropout_4 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 3
network_backbone:MLPBackbone:dropout_5 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 4
network_backbone:MLPBackbone:dropout_6 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 5
network_backbone:MLPBackbone:dropout_7 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 6
network_backbone:MLPBackbone:dropout_8 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 7
network_backbone:MLPBackbone:dropout_9 | network_backbone:MLPBackbone:use_dropout == True && network_backbone:MLPBackbone:num_groups > 8
network_backbone:MLPBackbone:num_units_10 | network_backbone:MLPBackbone:num_groups > 9
network_backbone:MLPBackbone:num_units_11 | network_backbone:MLPBackbone:num_groups > 10
network_backbone:MLPBackbone:num_units_12 | network_backbone:MLPBackbone:num_groups > 11
network_backbone:MLPBackbone:num_units_13 | network_backbone:MLPBackbone:num_groups > 12
network_backbone:MLPBackbone:num_units_14 | network_backbone:MLPBackbone:num_groups > 13
network_backbone:MLPBackbone:num_units_15 | network_backbone:MLPBackbone:num_groups > 14
network_backbone:MLPBackbone:num_units_2 | network_backbone:MLPBackbone:num_groups > 1
network_backbone:MLPBackbone:num_units_3 | network_backbone:MLPBackbone:num_groups > 2
network_backbone:MLPBackbone:num_units_4 | network_backbone:MLPBackbone:num_groups > 3
network_backbone:MLPBackbone:num_units_5 | network_backbone:MLPBackbone:num_groups > 4
network_backbone:MLPBackbone:num_units_6 | network_backbone:MLPBackbone:num_groups > 5
network_backbone:MLPBackbone:num_units_7 | network_backbone:MLPBackbone:num_groups > 6
network_backbone:MLPBackbone:num_units_8 | network_backbone:MLPBackbone:num_groups > 7
network_backbone:MLPBackbone:num_units_9 | network_backbone:MLPBackbone:num_groups > 8
network_backbone:MLPBackbone:dropout_1 | network_backbone:MLPBackbone:use_dropout == True
network_backbone:ResNetBackbone:blocks_per_group_10 | network_backbone:ResNetBackbone:num_groups > 9
network_backbone:ResNetBackbone:blocks_per_group_11 | network_backbone:ResNetBackbone:num_groups > 10
network_backbone:ResNetBackbone:blocks_per_group_12 | network_backbone:ResNetBackbone:num_groups > 11
network_backbone:ResNetBackbone:blocks_per_group_13 | network_backbone:ResNetBackbone:num_groups > 12
network_backbone:ResNetBackbone:blocks_per_group_14 | network_backbone:ResNetBackbone:num_groups > 13
network_backbone:ResNetBackbone:blocks_per_group_15 | network_backbone:ResNetBackbone:num_groups > 14
network_backbone:ResNetBackbone:blocks_per_group_2 | network_backbone:ResNetBackbone:num_groups > 1
network_backbone:ResNetBackbone:blocks_per_group_3 | network_backbone:ResNetBackbone:num_groups > 2
network_backbone:ResNetBackbone:blocks_per_group_4 | network_backbone:ResNetBackbone:num_groups > 3
network_backbone:ResNetBackbone:blocks_per_group_5 | network_backbone:ResNetBackbone:num_groups > 4
network_backbone:ResNetBackbone:blocks_per_group_6 | network_backbone:ResNetBackbone:num_groups > 5
network_backbone:ResNetBackbone:blocks_per_group_7 | network_backbone:ResNetBackbone:num_groups > 6
network_backbone:ResNetBackbone:blocks_per_group_8 | network_backbone:ResNetBackbone:num_groups > 7
network_backbone:ResNetBackbone:blocks_per_group_9 | network_backbone:ResNetBackbone:num_groups > 8
network_backbone:ResNetBackbone:dropout_10 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 9
network_backbone:ResNetBackbone:dropout_11 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 10
network_backbone:ResNetBackbone:dropout_12 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 11
network_backbone:ResNetBackbone:dropout_13 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 12
network_backbone:ResNetBackbone:dropout_14 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 13
network_backbone:ResNetBackbone:dropout_15 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 14
network_backbone:ResNetBackbone:dropout_2 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 1
network_backbone:ResNetBackbone:dropout_3 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 2
network_backbone:ResNetBackbone:dropout_4 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 3
network_backbone:ResNetBackbone:dropout_5 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 4
network_backbone:ResNetBackbone:dropout_6 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 5
network_backbone:ResNetBackbone:dropout_7 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 6
network_backbone:ResNetBackbone:dropout_8 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 7
network_backbone:ResNetBackbone:dropout_9 | network_backbone:ResNetBackbone:use_dropout == True && network_backbone:ResNetBackbone:num_groups > 8
network_backbone:ResNetBackbone:num_units_10 | network_backbone:ResNetBackbone:num_groups > 9
network_backbone:ResNetBackbone:num_units_11 | network_backbone:ResNetBackbone:num_groups > 10
network_backbone:ResNetBackbone:num_units_12 | network_backbone:ResNetBackbone:num_groups > 11
network_backbone:ResNetBackbone:num_units_13 | network_backbone:ResNetBackbone:num_groups > 12
network_backbone:ResNetBackbone:num_units_14 | network_backbone:ResNetBackbone:num_groups > 13
network_backbone:ResNetBackbone:num_units_15 | network_backbone:ResNetBackbone:num_groups > 14
network_backbone:ResNetBackbone:num_units_2 | network_backbone:ResNetBackbone:num_groups > 1
network_backbone:ResNetBackbone:num_units_3 | network_backbone:ResNetBackbone:num_groups > 2
network_backbone:ResNetBackbone:num_units_4 | network_backbone:ResNetBackbone:num_groups > 3
network_backbone:ResNetBackbone:num_units_5 | network_backbone:ResNetBackbone:num_groups > 4
network_backbone:ResNetBackbone:num_units_6 | network_backbone:ResNetBackbone:num_groups > 5
network_backbone:ResNetBackbone:num_units_7 | network_backbone:ResNetBackbone:num_groups > 6
network_backbone:ResNetBackbone:num_units_8 | network_backbone:ResNetBackbone:num_groups > 7
network_backbone:ResNetBackbone:num_units_9 | network_backbone:ResNetBackbone:num_groups > 8
network_backbone:ResNetBackbone:dropout_0 | network_backbone:ResNetBackbone:use_dropout == True
network_backbone:ResNetBackbone:dropout_1 | network_backbone:ResNetBackbone:use_dropout == True
network_backbone:ResNetBackbone:max_shake_drop_probability | network_backbone:ResNetBackbone:use_shake_drop == True
network_backbone:ShapedMLPBackbone:max_dropout | network_backbone:ShapedMLPBackbone:use_dropout == True
network_backbone:ShapedResNetBackbone:max_dropout | network_backbone:ShapedResNetBackbone:use_dropout == True
network_backbone:ShapedResNetBackbone:max_shake_drop_probability | network_backbone:ShapedResNetBackbone:use_shake_drop == True
network_head:fully_connected:activation | network_head:fully_connected:num_layers > 1
network_head:fully_connected:units_layer_1 | network_head:fully_connected:num_layers > 1
network_head:fully_connected:units_layer_2 | network_head:fully_connected:num_layers > 2
network_head:fully_connected:units_layer_3 | network_head:fully_connected:num_layers > 3
network_head:fully_connected:units_layer_4 | network_head:fully_connected:num_layers > 4

{network_embedding:__choice__=LearnedEntityEmbedding, encoder:__choice__=NoEncoder}

[ravinkohli]

Hi,

I am not sure I understand what you are saying. estimator.get_search_space(dataset).get_default_configuration() samples a single configuration from the full search space. But its a single configuration and not necessarily the best configuration. You can get the best configuration by searching over the full search space which you can check with estimator.get_search_space(dataset). AutoPyTorch will by default search over this search space. You dont have to pass any search space updates, unless you want to change specific hyperparameters. Let me know if you still need help with this.