diff --git a/examples/power/.dvc/.gitignore b/examples/power/.dvc/.gitignore
deleted file mode 100644
index 528f30c7..00000000
--- a/examples/power/.dvc/.gitignore
+++ /dev/null
@@ -1,3 +0,0 @@
-/config.local
-/tmp
-/cache
diff --git a/examples/power/conf/clean.yaml b/examples/power/conf/clean.yaml
new file mode 100644
index 00000000..47de8dad
--- /dev/null
+++ b/examples/power/conf/clean.yaml
@@ -0,0 +1,13 @@
+attacks:
+  FastGradientMethod: FGM
+defences:
+  Control: Control
+  FeatureSqueezing: FSQ
+params:
+  FGM: attack.init.eps
+  Control: model.trainer.nb_epoch
+  FSQ: model.art.pipeline.preprocessor.bit_depth
+fillna:
+  FGM: 0.0
+  Control: 20
+  FSQ: 32
diff --git a/examples/power/conf/compile.yaml b/examples/power/conf/compile.yaml
deleted file mode 100644
index 8514e6df..00000000
--- a/examples/power/conf/compile.yaml
+++ /dev/null
@@ -1,8 +0,0 @@
-attacks:
-  FastGradientMethod: FGM
-defences:
-  Control: Control
-  # FeatureSqueezing: FSQ
-params:
-  FGM: attack.init.eps
-  Control: epochs
diff --git a/examples/power/params.yaml b/examples/power/params.yaml
new file mode 100644
index 00000000..81856502
--- /dev/null
+++ b/examples/power/params.yaml
@@ -0,0 +1,241 @@
+_target_: deckard.base.experiment.Experiment
+attack:
+  _target_: deckard.base.attack.Attack
+  attack_size: 10
+  data:
+    _target_: deckard.base.data.Data
+    generate:
+      _target_: deckard.base.data.generator.DataGenerator
+      name: torch_mnist
+    sample:
+      _target_: deckard.base.data.sampler.SklearnDataSampler
+      random_state: 0
+      stratify: true
+    sklearn_pipeline:
+      _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+      preprocessor:
+        name: sklearn.preprocessing.StandardScaler
+        with_mean: true
+        with_std: true
+  init:
+    _target_: deckard.base.attack.AttackInitializer
+    batch_size: 1024
+    eps: 0.99
+    minimal: true
+    model:
+      _target_: deckard.base.model.Model
+      art:
+        _target_: deckard.base.model.art_pipeline.ArtPipeline
+        data:
+          _target_: deckard.base.data.Data
+          generate:
+            _target_: deckard.base.data.generator.DataGenerator
+            name: torch_mnist
+          sample:
+            _target_: deckard.base.data.sampler.SklearnDataSampler
+            random_state: 0
+            stratify: true
+          sklearn_pipeline:
+            _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+            preprocessor:
+              name: sklearn.preprocessing.StandardScaler
+              with_mean: true
+              with_std: true
+        initialize:
+          clip_values:
+          - 0
+          - 255
+          criterion:
+            name: torch.nn.CrossEntropyLoss
+          optimizer:
+            lr: 0.01
+            momentum: 0.9
+            name: torch.optim.SGD
+        library: pytorch
+      data:
+        _target_: deckard.base.data.Data
+        generate:
+          _target_: deckard.base.data.generator.DataGenerator
+          name: torch_mnist
+        sample:
+          _target_: deckard.base.data.sampler.SklearnDataSampler
+          random_state: 0
+          stratify: true
+        sklearn_pipeline:
+          _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+          preprocessor:
+            name: sklearn.preprocessing.StandardScaler
+            with_mean: true
+            with_std: true
+      init:
+        _target_: deckard.base.model.ModelInitializer
+        name: torch_example.ResNet18
+        num_channels: 1
+      library: pytorch
+      trainer:
+        batch_size: 1024
+        nb_epoch: 1
+    name: art.attacks.evasion.FastGradientMethod
+    targeted: false
+  method: evasion
+  model:
+    _target_: deckard.base.model.Model
+    art:
+      _target_: deckard.base.model.art_pipeline.ArtPipeline
+      data:
+        _target_: deckard.base.data.Data
+        generate:
+          _target_: deckard.base.data.generator.DataGenerator
+          name: torch_mnist
+        sample:
+          _target_: deckard.base.data.sampler.SklearnDataSampler
+          random_state: 0
+          stratify: true
+        sklearn_pipeline:
+          _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+          preprocessor:
+            name: sklearn.preprocessing.StandardScaler
+            with_mean: true
+            with_std: true
+      initialize:
+        clip_values:
+        - 0
+        - 255
+        criterion:
+          name: torch.nn.CrossEntropyLoss
+        optimizer:
+          lr: 0.01
+          momentum: 0.9
+          name: torch.optim.SGD
+      library: pytorch
+    data:
+      _target_: deckard.base.data.Data
+      generate:
+        _target_: deckard.base.data.generator.DataGenerator
+        name: torch_mnist
+      sample:
+        _target_: deckard.base.data.sampler.SklearnDataSampler
+        random_state: 0
+        stratify: true
+      sklearn_pipeline:
+        _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+        preprocessor:
+          name: sklearn.preprocessing.StandardScaler
+          with_mean: true
+          with_std: true
+    init:
+      _target_: deckard.base.model.ModelInitializer
+      name: torch_example.ResNet18
+      num_channels: 1
+    library: pytorch
+    trainer:
+      batch_size: 1024
+      nb_epoch: 1
+data:
+  _target_: deckard.base.data.Data
+  generate:
+    _target_: deckard.base.data.generator.DataGenerator
+    name: torch_mnist
+  sample:
+    _target_: deckard.base.data.sampler.SklearnDataSampler
+    random_state: 0
+    stratify: true
+  sklearn_pipeline:
+    _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+    preprocessor:
+      name: sklearn.preprocessing.StandardScaler
+      with_mean: true
+      with_std: true
+device_id: cpu
+direction:
+- maximize
+- minimize
+- minimize
+- minimize
+files:
+  _target_: deckard.base.files.FileConfig
+  adv_predictions_file: adv_predictions.json
+  attack_dir: attacks
+  attack_file: attack
+  attack_type: .pkl
+  data_dir: data
+  data_file: data
+  data_type: .pkl
+  directory: /result/mnist/
+  model_dir: null
+  model_file: null
+  model_type: null
+  name: default
+  params_file: params.yaml
+  predictions_file: predictions.json
+  reports: reports
+  score_dict_file: score_dict.json
+model:
+  _target_: deckard.base.model.Model
+  art:
+    _target_: deckard.base.model.art_pipeline.ArtPipeline
+    data:
+      _target_: deckard.base.data.Data
+      generate:
+        _target_: deckard.base.data.generator.DataGenerator
+        name: torch_mnist
+      sample:
+        _target_: deckard.base.data.sampler.SklearnDataSampler
+        random_state: 0
+        stratify: true
+      sklearn_pipeline:
+        _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+        preprocessor:
+          name: sklearn.preprocessing.StandardScaler
+          with_mean: true
+          with_std: true
+    initialize:
+      clip_values:
+      - 0
+      - 255
+      criterion:
+        name: torch.nn.CrossEntropyLoss
+      optimizer:
+        lr: 0.01
+        momentum: 0.9
+        name: torch.optim.SGD
+    library: pytorch
+  data:
+    _target_: deckard.base.data.Data
+    generate:
+      _target_: deckard.base.data.generator.DataGenerator
+      name: torch_mnist
+    sample:
+      _target_: deckard.base.data.sampler.SklearnDataSampler
+      random_state: 0
+      stratify: true
+    sklearn_pipeline:
+      _target_: deckard.base.data.sklearn_pipeline.SklearnDataPipeline
+      preprocessor:
+        name: sklearn.preprocessing.StandardScaler
+        with_mean: true
+        with_std: true
+  init:
+    _target_: deckard.base.model.ModelInitializer
+    name: torch_example.ResNet18
+    num_channels: 1
+  library: pytorch
+  trainer:
+    batch_size: 1024
+    nb_epoch: 1
+optimizers:
+- accuracy
+- train_time
+- adv_accuracy
+- adv_fit_time
+scorers:
+  _target_: deckard.base.scorer.ScorerDict
+  accuracy:
+    _target_: deckard.base.scorer.ScorerConfig
+    direction: maximize
+    name: sklearn.metrics.accuracy_score
+  log_loss:
+    _target_: deckard.base.scorer.ScorerConfig
+    direction: minimize
+    name: sklearn.metrics.log_loss
+stage: ???
diff --git a/examples/power/plots/dvc.lock b/examples/power/plots/dvc.lock
deleted file mode 100644
index ad996cd0..00000000
--- a/examples/power/plots/dvc.lock
+++ /dev/null
@@ -1,502 +0,0 @@
-schema: '2.0'
-stages:
-  compile@mnist:
-    cmd: python -m deckard.layers.compile --report_folder /result/mnist/reports/attack/
-      --results_file attack.csv --results_folder mnist --config "../conf/compile.yaml"
-    deps:
-    - path: ../conf/compile.yaml
-      hash: md5
-      md5: 61054aeaa58015252df89c20db9ec595
-      size: 139
-    - path: /result/mnist/reports/attack/
-      hash: md5
-      md5: 2fd9e6eb909c65078ded91f85c5938d1.dir
-      size: 5846593866
-      nfiles: 8024
-    outs:
-    - path: mnist/attack.csv
-      hash: md5
-      md5: e58a980dc2e0da18144d8a7d55e41ac4
-      size: 9759136
-  compile@cifar:
-    cmd: python -m deckard.layers.compile --report_folder /result/cifar/reports/attack/
-      --results_file attack.csv --results_folder cifar --config "../conf/compile.yaml"
-    deps:
-    - path: ../conf/compile.yaml
-      hash: md5
-      md5: 61054aeaa58015252df89c20db9ec595
-      size: 139
-    - path: /result/cifar/reports/attack/
-      hash: md5
-      md5: ed8b007df2788a65643c2c0b99a87a4d.dir
-      size: 5035287644
-      nfiles: 8031
-    outs:
-    - path: cifar/attack.csv
-      hash: md5
-      md5: 1adead42bdfa4a1cf7721c563f08fb17
-      size: 9787000
-  compile@cifar100:
-    cmd: python -m deckard.layers.compile --report_folder /result/cifar100/reports/attack/
-      --results_file attack.csv --results_folder cifar100 --config "../conf/compile.yaml"
-    deps:
-    - path: ../conf/compile.yaml
-      hash: md5
-      md5: 61054aeaa58015252df89c20db9ec595
-      size: 139
-    - path: /result/cifar100/reports/attack/
-      hash: md5
-      md5: d70914292c9d7d2d6703ce8c5179f334.dir
-      size: 66900144277
-      nfiles: 11419
-    outs:
-    - path: cifar100/attack.csv
-      hash: md5
-      md5: d710860394e7af9976b96f399dcc9e8a
-      size: 14439946
-  plot@mnist:
-    cmd: python -m deckard.layers.plots --path mnist --file mnist/attack.csv -o plot_data.csv
-      -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
-    deps:
-    - path: ../conf/plots.yaml
-      hash: md5
-      md5: 6b372706e37d0d371d245d9d927e7577
-      size: 1321
-    - path: mnist/attack.csv
-      hash: md5
-      md5: e58a980dc2e0da18144d8a7d55e41ac4
-      size: 9759136
-    outs:
-    - path: mnist/adv_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 4d67d03038f6910890a9b3de6c388a5d
-      size: 32913
-    - path: mnist/ben_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: cf4423c2e57d02639ac7ad3f73e7923f
-      size: 34114
-    - path: mnist/plot_data.csv
-      hash: md5
-      md5: ef9fa5261715b08c90deec69c6e6831b
-      size: 4773500
-    - path: mnist/train_time_vs_accuracy.pdf
-      hash: md5
-      md5: 15597e7f04132220b4263a5c0d92625e
-      size: 21898
-  plot@cifar:
-    cmd: python -m deckard.layers.plots --path cifar --file cifar/attack.csv -o plot_data.csv
-      -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
-    deps:
-    - path: ../conf/plots.yaml
-      hash: md5
-      md5: 6b372706e37d0d371d245d9d927e7577
-      size: 1321
-    - path: cifar/attack.csv
-      hash: md5
-      md5: 1adead42bdfa4a1cf7721c563f08fb17
-      size: 9787000
-    outs:
-    - path: cifar/adv_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 68daee0b3952983c42c57e8ec1b52e73
-      size: 28362
-    - path: cifar/ben_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 01496b89613804e691a7ed15a375d302
-      size: 29341
-    - path: cifar/plot_data.csv
-      hash: md5
-      md5: 71f94277b1c3e8f50136a11a3d563f8d
-      size: 3311418
-    - path: cifar/train_time_vs_accuracy.pdf
-      hash: md5
-      md5: fa158ac5977c5acca89eefcde397ae11
-      size: 22760
-  plot@cifar100:
-    cmd: python -m deckard.layers.plots --path cifar100 --file cifar100/attack.csv
-      -o plot_data.csv -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
-    deps:
-    - path: ../conf/plots.yaml
-      hash: md5
-      md5: 6b372706e37d0d371d245d9d927e7577
-      size: 1321
-    - path: cifar100/attack.csv
-      hash: md5
-      md5: d710860394e7af9976b96f399dcc9e8a
-      size: 14439946
-    outs:
-    - path: cifar100/adv_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: e63a303612b5c5e5b71bdb8efc4fead9
-      size: 30309
-    - path: cifar100/ben_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 2a1e64977699d108f9a6ceb14a44fc89
-      size: 32121
-    - path: cifar100/plot_data.csv
-      hash: md5
-      md5: a8a813a9d3a8a4171730c3ef09ece6ee
-      size: 4016181
-    - path: cifar100/train_time_vs_accuracy.pdf
-      hash: md5
-      md5: df6751797da220c51e14f1424522dcd5
-      size: 23971
-  afr@mnist:
-    cmd: python -m deckard.layers.afr --dataset mnist --data_file mnist/plot_data.csv
-      --config_file "../conf/afr.yaml"
-    deps:
-    - path: ../conf/afr.yaml
-      hash: md5
-      md5: 2f7ff61ce1517a4320ab990cbf5f2b16
-      size: 8084
-    - path: mnist/plot_data.csv
-      hash: md5
-      md5: ef9fa5261715b08c90deec69c6e6831b
-      size: 4773500
-    outs:
-    - path: mnist/aft_comparison.csv
-      hash: md5
-      md5: 17d4ef4926c0725a02c16d16a83098fc
-      size: 213
-    - path: mnist/aft_comparison.tex
-      hash: md5
-      md5: 106a151a290681609de5d9073ad83cd7
-      size: 451
-    - path: mnist/cox_aft.pdf
-      hash: md5
-      md5: 738d989e4e913ec2a4a0aa770108ecf9
-      size: 20071
-    - path: mnist/cox_epochs_partial_effect.pdf
-      hash: md5
-      md5: 7b2e73f00570d8ea48993e85aae4a462
-      size: 27938
-    - path: mnist/log_logistic_aft.pdf
-      hash: md5
-      md5: f8ec551175e747f38da9fae23a76250d
-      size: 22634
-    - path: mnist/log_logistic_epochs_partial_effect.pdf
-      hash: md5
-      md5: b1f1c283e0ccb15c6a0946cd5bcad1c6
-      size: 28146
-    - path: mnist/log_normal_aft.pdf
-      hash: md5
-      md5: c8f0312afa35ff2a92958666581831a1
-      size: 23628
-    - path: mnist/log_normal_epochs_partial_effect.pdf
-      hash: md5
-      md5: 3b4dedcbb7c7b41b362c4f6b41648a2f
-      size: 28548
-    - path: mnist/weibull_aft.pdf
-      hash: md5
-      md5: 8538a93dcc9267ec07bf554a7d703d02
-      size: 31596
-    - path: mnist/weibull_epochs_partial_effect.pdf
-      hash: md5
-      md5: f6d46d7e580e3834abd9cef71f0f03b3
-      size: 28446
-  afr@cifar:
-    cmd: python -m deckard.layers.afr --dataset cifar --data_file cifar/plot_data.csv
-      --config_file "../conf/afr.yaml"
-    deps:
-    - path: ../conf/afr.yaml
-      hash: md5
-      md5: 2f7ff61ce1517a4320ab990cbf5f2b16
-      size: 8084
-    - path: cifar/plot_data.csv
-      hash: md5
-      md5: 71f94277b1c3e8f50136a11a3d563f8d
-      size: 3311418
-    outs:
-    - path: cifar/aft_comparison.csv
-      hash: md5
-      md5: 7ecfb10f7474b4fbebb530fb50fe164e
-      size: 207
-    - path: cifar/aft_comparison.tex
-      hash: md5
-      md5: 8904c5c2bc9a04ec544cede940da5819
-      size: 449
-    - path: cifar/cox_aft.pdf
-      hash: md5
-      md5: 268bc658b99003fc31bc2f0cbc474c21
-      size: 19797
-    - path: cifar/cox_epochs_partial_effect.pdf
-      hash: md5
-      md5: 0ff0825bae278d55cb35e795b7daa236
-      size: 27647
-    - path: cifar/log_logistic_aft.pdf
-      hash: md5
-      md5: ff26a0a9eea8d3ae73a671c058aa477b
-      size: 22632
-    - path: cifar/log_logistic_epochs_partial_effect.pdf
-      hash: md5
-      md5: 0fd26ac4269fc6a36179a7724aa42664
-      size: 27417
-    - path: cifar/log_normal_aft.pdf
-      hash: md5
-      md5: 0c8ad4f39d0c1df902b4764c667e9006
-      size: 23630
-    - path: cifar/log_normal_epochs_partial_effect.pdf
-      hash: md5
-      md5: f46d273316733b83d06639fa017f53c6
-      size: 28173
-    - path: cifar/weibull_aft.pdf
-      hash: md5
-      md5: bea79bd48e84d4f314852e108d3f460a
-      size: 31594
-    - path: cifar/weibull_epochs_partial_effect.pdf
-      hash: md5
-      md5: 9d8a75464641a821cbac19af2df585c6
-      size: 27820
-  afr@cifar100:
-    cmd: python -m deckard.layers.afr --dataset cifar100 --data_file cifar100/plot_data.csv
-      --config_file "../conf/afr.yaml"
-    deps:
-    - path: ../conf/afr.yaml
-      hash: md5
-      md5: 2f7ff61ce1517a4320ab990cbf5f2b16
-      size: 8084
-    - path: cifar100/plot_data.csv
-      hash: md5
-      md5: a8a813a9d3a8a4171730c3ef09ece6ee
-      size: 4016181
-    outs:
-    - path: cifar100/aft_comparison.csv
-      hash: md5
-      md5: 398a9aad117e230afa2ca484581b37bc
-      size: 212
-    - path: cifar100/aft_comparison.tex
-      hash: md5
-      md5: ab03a9867122740b0409c6de0da49ef3
-      size: 457
-    - path: cifar100/cox_aft.pdf
-      hash: md5
-      md5: b648636cea335b8b003ff9aae39c265b
-      size: 19562
-    - path: cifar100/cox_epochs_partial_effect.pdf
-      hash: md5
-      md5: e62c782ffcbec52ff8f1378bbb85ddf6
-      size: 27029
-    - path: cifar100/log_logistic_aft.pdf
-      hash: md5
-      md5: 6ee44957cd08349936dffacafe8bd7c9
-      size: 22629
-    - path: cifar100/log_logistic_epochs_partial_effect.pdf
-      hash: md5
-      md5: a8ed3b1d0a9a804220ef7e6e0c7839bc
-      size: 27149
-    - path: cifar100/log_normal_aft.pdf
-      hash: md5
-      md5: f6cddf8b1053801067bdb5dba610fd32
-      size: 23633
-    - path: cifar100/log_normal_epochs_partial_effect.pdf
-      hash: md5
-      md5: 34ee6b72c4570faf0da1e0023146b16b
-      size: 27783
-    - path: cifar100/weibull_aft.pdf
-      hash: md5
-      md5: 444e08b12dd3332a2dc3c90a6458a527
-      size: 32232
-    - path: cifar100/weibull_epochs_partial_effect.pdf
-      hash: md5
-      md5: 4f34911f42f42218ff08c13c0e35ba18
-      size: 26687
-  compile@bit_depth/mnist/:
-    cmd: python -m deckard.layers.compile --report_folder /result/bit_depth/mnist//reports/attack/
-      --results_file attack.csv --results_folder bit_depth/mnist/ --config "../conf/compile.yaml"
-    deps:
-    - path: ../conf/compile.yaml
-      hash: md5
-      md5: 61054aeaa58015252df89c20db9ec595
-      size: 139
-    - path: /result/bit_depth/mnist//reports/attack/
-      hash: md5
-      md5: 0e98e6497ec11efe7d15d6d26d48e373.dir
-      size: 3045516686
-      nfiles: 4182
-    outs:
-    - path: bit_depth/mnist//attack.csv
-      hash: md5
-      md5: 0a82d226a00f0e72266122a5dfcc063e
-      size: 5312357
-  compile@bit_depth/cifar/:
-    cmd: python -m deckard.layers.compile --report_folder /result/bit_depth/cifar//reports/attack/
-      --results_file attack.csv --results_folder bit_depth/cifar/ --config "../conf/compile.yaml"
-    deps:
-    - path: ../conf/compile.yaml
-      hash: md5
-      md5: 61054aeaa58015252df89c20db9ec595
-      size: 139
-    - path: /result/bit_depth/cifar//reports/attack/
-      hash: md5
-      md5: 2ccefb9124bfd7288cb342cfa615b814.dir
-      size: 1706177501
-      nfiles: 2719
-    outs:
-    - path: bit_depth/cifar//attack.csv
-      hash: md5
-      md5: d31c178c71be151864b2867130b25834
-      size: 3470593
-  plot@bit_depth/mnist/:
-    cmd: python -m deckard.layers.plots --path bit_depth/mnist/ --file bit_depth/mnist//attack.csv
-      -o plot_data.csv -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
-    deps:
-    - path: ../conf/plots.yaml
-      hash: md5
-      md5: 6b372706e37d0d371d245d9d927e7577
-      size: 1321
-    - path: bit_depth/mnist//attack.csv
-      hash: md5
-      md5: 0a82d226a00f0e72266122a5dfcc063e
-      size: 5312357
-    outs:
-    - path: bit_depth/mnist//adv_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 1a12d0b27a77579d682185449bad34ba
-      size: 26881
-    - path: bit_depth/mnist//ben_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: dc2a0e22088ecdf8da8a416c3c9983b0
-      size: 27978
-    - path: bit_depth/mnist//plot_data.csv
-      hash: md5
-      md5: 33ba83a5d6f73875448ab74a2f18916d
-      size: 2982894
-    - path: bit_depth/mnist//train_time_vs_accuracy.pdf
-      hash: md5
-      md5: 1c220c2e8566b0a63846b1868c4d1f20
-      size: 20902
-  plot@bit_depth/cifar/:
-    cmd: python -m deckard.layers.plots --path bit_depth/cifar/ --file bit_depth/cifar//attack.csv
-      -o plot_data.csv -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
-    deps:
-    - path: ../conf/plots.yaml
-      hash: md5
-      md5: 6b372706e37d0d371d245d9d927e7577
-      size: 1321
-    - path: bit_depth/cifar//attack.csv
-      hash: md5
-      md5: d31c178c71be151864b2867130b25834
-      size: 3470593
-    outs:
-    - path: bit_depth/cifar//adv_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: 6ff4890ddbfeb078d62dfb5707fe82e2
-      size: 22234
-    - path: bit_depth/cifar//ben_failure_rate_vs_train_time.pdf
-      hash: md5
-      md5: c9f9e2bab55d1b0683de55a3217cadbf
-      size: 23164
-    - path: bit_depth/cifar//plot_data.csv
-      hash: md5
-      md5: 08ca910134d6e1806f05c1d298b0f664
-      size: 1561678
-    - path: bit_depth/cifar//train_time_vs_accuracy.pdf
-      hash: md5
-      md5: 82c370b0db144053df5370102026435d
-      size: 20902
-  afr@bit_depth/mnist/:
-    cmd: python -m deckard.layers.afr --dataset bit_depth/mnist/ --data_file bit_depth/mnist//plot_data.csv
-      --config_file "../conf/afr.yaml"
-    deps:
-    - path: ../conf/afr.yaml
-      hash: md5
-      md5: 2f7ff61ce1517a4320ab990cbf5f2b16
-      size: 8084
-    - path: bit_depth/mnist//plot_data.csv
-      hash: md5
-      md5: 33ba83a5d6f73875448ab74a2f18916d
-      size: 2982894
-    outs:
-    - path: bit_depth/mnist//aft_comparison.csv
-      hash: md5
-      md5: 0985db77aaa78ff65efb3d981ac19a7d
-      size: 207
-    - path: bit_depth/mnist//aft_comparison.tex
-      hash: md5
-      md5: b66410cbc16c7287655066e624cc4a64
-      size: 467
-    - path: bit_depth/mnist//cox_aft.pdf
-      hash: md5
-      md5: 81ab402581dfeb80094b02e96ccbebd6
-      size: 19564
-    - path: bit_depth/mnist//cox_epochs_partial_effect.pdf
-      hash: md5
-      md5: c8bf7109789922c17a281a87e7115c05
-      size: 28240
-    - path: bit_depth/mnist//log_logistic_aft.pdf
-      hash: md5
-      md5: 981244878b659ce0978bf393da0e6ef0
-      size: 22635
-    - path: bit_depth/mnist//log_logistic_epochs_partial_effect.pdf
-      hash: md5
-      md5: 9293c5362f3e21d03723a8a9fb979efb
-      size: 27817
-    - path: bit_depth/mnist//log_normal_aft.pdf
-      hash: md5
-      md5: 33fcc4485293fa4704c033191f910578
-      size: 23621
-    - path: bit_depth/mnist//log_normal_epochs_partial_effect.pdf
-      hash: md5
-      md5: 6977db1ba0125ef9d4151ddafdd8529e
-      size: 28250
-    - path: bit_depth/mnist//weibull_aft.pdf
-      hash: md5
-      md5: 261fb580d43eaffff310d6e3c6dde80b
-      size: 32224
-    - path: bit_depth/mnist//weibull_epochs_partial_effect.pdf
-      hash: md5
-      md5: b4780c1de1b658896c1044183f1b6fb6
-      size: 28039
-  afr@bit_depth/cifar/:
-    cmd: python -m deckard.layers.afr --dataset bit_depth/cifar/ --data_file bit_depth/cifar//plot_data.csv
-      --config_file "../conf/afr.yaml"
-    deps:
-    - path: ../conf/afr.yaml
-      hash: md5
-      md5: 2f7ff61ce1517a4320ab990cbf5f2b16
-      size: 8084
-    - path: bit_depth/cifar//plot_data.csv
-      hash: md5
-      md5: 08ca910134d6e1806f05c1d298b0f664
-      size: 1561678
-    outs:
-    - path: bit_depth/cifar//aft_comparison.csv
-      hash: md5
-      md5: e2ad4bbeaceffac83f2669d1fc7aaf6d
-      size: 207
-    - path: bit_depth/cifar//aft_comparison.tex
-      hash: md5
-      md5: b84efba153648578fba3167e642c566f
-      size: 467
-    - path: bit_depth/cifar//cox_aft.pdf
-      hash: md5
-      md5: 4c361cb01e604398080ac26b712ec931
-      size: 19972
-    - path: bit_depth/cifar//cox_epochs_partial_effect.pdf
-      hash: md5
-      md5: 8e150547d47eb745d4be47974cb1029c
-      size: 27127
-    - path: bit_depth/cifar//log_logistic_aft.pdf
-      hash: md5
-      md5: 10cc72619e730733e3e975834aaec15e
-      size: 22642
-    - path: bit_depth/cifar//log_logistic_epochs_partial_effect.pdf
-      hash: md5
-      md5: de33aa8a707b76b647f1eac3f71ed444
-      size: 28988
-    - path: bit_depth/cifar//log_normal_aft.pdf
-      hash: md5
-      md5: 6ef63cefaea6cbb2bdf6d3d92635d99e
-      size: 23631
-    - path: bit_depth/cifar//log_normal_epochs_partial_effect.pdf
-      hash: md5
-      md5: 19153bae3e9b778fccee3cda459832c5
-      size: 29285
-    - path: bit_depth/cifar//weibull_aft.pdf
-      hash: md5
-      md5: 8518326aecf02b25f3b69d5d2e2a582e
-      size: 32225
-    - path: bit_depth/cifar//weibull_epochs_partial_effect.pdf
-      hash: md5
-      md5: 8ccc2f82015de53960a131d92b8d2653
-      size: 29270
diff --git a/examples/power/plots/dvc.yaml b/examples/power/plots/dvc.yaml
index 7a8f988a..02e75d1d 100644
--- a/examples/power/plots/dvc.yaml
+++ b/examples/power/plots/dvc.yaml
@@ -1,6 +1,10 @@
 vars:
   - ../conf/plots.yaml:line_plot
   - ../conf/plots.yaml:scatter_plot
+  - ../conf/clean.yaml:attacks
+  - ../conf/clean.yaml:defences
+  - ../conf/clean.yaml:params
+  - ../conf/clean.yaml:fillna
 stages:
   compile:
     foreach: # iterates through each stage
@@ -9,53 +13,74 @@ stages:
       - cifar100
       - bit_depth/mnist/
       - bit_depth/cifar/
+      - bit_depth/cifar100/
     do:
       cmd: python -m deckard.layers.compile --report_folder /result/${item}/reports/attack/ --results_file attack.csv --results_folder ${item} --config "../conf/compile.yaml"
-      deps:
-       - "../conf/compile.yaml"
-       - /result/${item}/reports/attack/
+      # deps:
+      #  - /result/${item}/reports/attack/
       outs:
       - ${item}/attack.csv
-  plot:
+  clean:
     foreach:
-       - mnist
-       - cifar
-       - cifar100
-       - bit_depth/mnist/
-       - bit_depth/cifar/
+      - mnist
+      - cifar
+      - cifar100
+      - bit_depth/mnist/
+      - bit_depth/cifar/
+      - bit_depth/cifar100/
     do:
-      cmd: python -m deckard.layers.plots --path ${item} --file ${item}/attack.csv -o plot_data.csv -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
+      cmd: python -m deckard.layers.clean_data --path ${item} --file attack.csv --config "../conf/clean.yaml"
       deps:
         - ${item}/attack.csv
-        - "../conf/plots.yaml"
-      outs:
-        - ${item}/plot_data.csv
-      plots:
-        - ${item}/${scatter_plot[0].file} # You can use the file specified in each list entry from the parameters specified above in vars
-        - ${item}/${scatter_plot[1].file}
-        - ${item}/${line_plot[0].file}
-  afr:
-    foreach:
-       - mnist
-       - cifar
-       - cifar100
-       - bit_depth/mnist/
-       - bit_depth/cifar/
-    do:
-      cmd: python -m deckard.layers.afr --dataset ${item} --data_file ${item}/plot_data.csv --config_file "../conf/afr.yaml"
-      deps:
-        - ${item}/plot_data.csv
-        - ../conf/afr.yaml
-      plots:
-        - ${item}/weibull_aft.pdf # You can also define them manually
-        - ${item}/weibull_epochs_partial_effect.pdf
-        - ${item}/cox_epochs_partial_effect.pdf
-        - ${item}/cox_aft.pdf
-        - ${item}/log_logistic_aft.pdf
-        - ${item}/log_logistic_epochs_partial_effect.pdf
-        - ${item}/log_normal_aft.pdf
-        - ${item}/log_normal_epochs_partial_effect.pdf
-      metrics:
-        - ${item}/aft_comparison.csv
+        - ../conf/clean.yaml
       outs:
-        - ${item}/aft_comparison.tex
+        - ${item}/clean.csv
+      params:
+        - ../conf/clean.yaml:
+          - attacks
+          - defences
+          - params
+          - fillna
+  # plot:
+  #   foreach:
+  #      - mnist
+  #      - cifar
+  #      - cifar100
+  #      - bit_depth/mnist/
+  #      - bit_depth/cifar/
+  #   do:
+  #     cmd: python -m deckard.layers.plots --path ${item} --file ${item}/attack.csv -o plot_data.csv -c "../conf/plots.yaml" --subset train_time,predict_proba_time,accuracy,adv_accuracy,adv_fit_time
+  #     deps:
+  #       # - ${item}/attack.csv  # This should be uncommented, but I don't have enough disk space to cache the hashes of the results.
+  #       - "../conf/plots.yaml"
+  #     outs:
+  #       - ${item}/plot_data.csv
+  #     plots:
+  #       - ${item}/${scatter_plot[0].file} # You can use the file specified in each list entry from the parameters specified above in vars
+  #       - ${item}/${scatter_plot[1].file}
+  #       - ${item}/${line_plot[0].file}
+  # afr:
+  #   foreach:
+  #      - mnist
+  #      - cifar
+  #      - cifar100
+  #      - bit_depth/mnist/
+  #      - bit_depth/cifar/
+  #   do:
+  #     cmd: python -m deckard.layers.afr --dataset ${item} --data_file ${item}/plot_data.csv --config_file "../conf/afr.yaml"
+  #     deps:
+  #       - ${item}/plot_data.csv
+  #       - ../conf/afr.yaml
+  #     plots:
+  #       - ${item}/weibull_aft.pdf # You can also define them manually
+  #       - ${item}/weibull_epochs_partial_effect.pdf
+  #       - ${item}/cox_epochs_partial_effect.pdf
+  #       - ${item}/cox_aft.pdf
+  #       - ${item}/log_logistic_aft.pdf
+  #       - ${item}/log_logistic_epochs_partial_effect.pdf
+  #       - ${item}/log_normal_aft.pdf
+  #       - ${item}/log_normal_epochs_partial_effect.pdf
+  #     metrics:
+  #       - ${item}/aft_comparison.csv
+  #     outs:
+  #       - ${item}/aft_comparison.tex