-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Add more talos testing for classification
- Loading branch information
1 parent
877040a
commit 9b4a7fa
Showing
3 changed files
with
325 additions
and
12 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,139 @@ | ||
| import talos as ta | ||
| from keras.activations import relu, elu, softmax, hard_sigmoid, tanh | ||
| from keras.layers import Flatten, ConvLSTM2D, Dense, Conv3D, MaxPooling3D, Dropout | ||
| from keras.losses import mean_squared_error, mean_absolute_error | ||
| from keras.models import Sequential | ||
| from keras.optimizers import adam, nadam, rmsprop | ||
| from talos.model.early_stopper import early_stopper | ||
| from talos.model.layers import hidden_layers | ||
| from talos.model.normalizers import lr_normalizer | ||
| from talos.metrics.keras_metrics import root_mean_squared_error, fmeasure_acc, matthews_correlation_acc, precision_acc, recall_acc | ||
|
|
||
| from factnn.utils.cross_validate import get_chunk_of_data | ||
|
|
||
| # Parameter Dictionary for talos | ||
|
|
||
| params = {'lr': (1, 10, 10), | ||
| 'first_neuron': [16, 32, 64], | ||
| 'last_neuron': [8, 16, 32], | ||
| 'hidden_layers': [2, 3, 6], | ||
| 'batch_size': [2, 8, 16], | ||
| 'epochs': [500], | ||
| 'dropout': (0, 0.80, 4), | ||
| 'weight_regulizer': [None], | ||
| 'emb_output_dims': [None], | ||
| 'optimizer': [adam, nadam, rmsprop], | ||
| 'losses': [mean_squared_error], | ||
| 'activation': [relu, elu, hard_sigmoid], | ||
| 'last_activation': [softmax], | ||
|
|
||
| 'neuron_1': [8, 16, 64], | ||
| 'kernel_1': [1, 3, 5], | ||
| 'stride_1': [1, 2, 3], | ||
| 'layer_drop': [0.0, 0.8, 4], | ||
| 'layers': [2,3,4], | ||
| 'pool': [0,1] | ||
|
|
||
| } | ||
| ''' | ||
| 'rec_dropout': [0.0, 0.4, 5], | ||
| 'rec_act': [hard_sigmoid, tanh], | ||
| 'pool': [0, 1], | ||
| 'neuron_2': [4, 8, 16, 32, 64], | ||
| 'kernel_2': [1, 2, 3], | ||
| 'stride_2': [1, 2], | ||
| 'three': [0, 1], | ||
| 'neuron_3': [4, 8, 16, 32, 64], | ||
| 'kernel_3': [1, 2, 3], | ||
| 'stride_3': [1, 2], | ||
| ''' | ||
|
|
||
| def input_model(x_train, y_train, x_val, y_val, params): | ||
| model = Sequential() | ||
|
|
||
| model.add(Conv3D(params['neuron_1'], kernel_size=params['kernel_1'], strides=params['stride_1'], | ||
| padding='same', | ||
| input_shape=(100, 100, 10, 1), | ||
| activation=params['activation'])) | ||
| if params['pool']: | ||
| model.add(MaxPooling3D()) | ||
| if params['layer_drop'] > 0.001: | ||
| model.add(Dropout(params['layer_drop'])) | ||
| model.add(Conv3D(params['neuron_1'], kernel_size=params['kernel_1'], strides=params['stride_1'], | ||
| padding='same', activation=params['activation'])) | ||
| #if params['pool']: | ||
| # model.add(MaxPooling2D()) | ||
| if params['layer_drop'] > 0.001: | ||
| model.add(Dropout(params['layer_drop'])) | ||
|
|
||
| if params['layers'] >= 3: | ||
| model.add(Conv3D(params['neuron_1'], kernel_size=params['kernel_1'], strides=params['stride_1'], | ||
| padding='same', activation=params['activation'])) | ||
| if params['pool']: | ||
| model.add(MaxPooling3D()) | ||
| if params['layer_drop'] > 0.001: | ||
| model.add(Dropout(params['layer_drop'])) | ||
|
|
||
| if params['layers'] >= 4: | ||
| model.add(Conv3D(params['neuron_1'], kernel_size=params['kernel_1'], strides=params['stride_1'], | ||
| padding='same', activation=params['activation'])) | ||
| # if params['pool']: | ||
| # model.add(MaxPooling2D()) | ||
| if params['layer_drop'] > 0.001: | ||
| model.add(Dropout(params['layer_drop'])) | ||
| model.add(Flatten()) | ||
|
|
||
| hidden_layers(model, params, params['last_neuron']) | ||
|
|
||
| model.add(Dense(2, activation='softmax')) | ||
|
|
||
| model.compile(loss=params['losses'], | ||
| optimizer=params['optimizer'](lr_normalizer(params['lr'], params['optimizer'])), | ||
| metrics=['acc', | ||
| fmeasure_acc, | ||
| matthews_correlation_acc, | ||
| precision_acc, | ||
| recall_acc]) | ||
|
|
||
| out = model.fit(x_train, y_train, | ||
| batch_size=params['batch_size'], | ||
| epochs=params['epochs'], verbose=0, | ||
| validation_data=[x_val, y_val], | ||
| callbacks=[early_stopper(params['epochs'], | ||
| mode='moderate')]) | ||
|
|
||
| return out, model | ||
|
|
||
|
|
||
| import argparse | ||
|
|
||
| # construct the argument parse and parse the arguments | ||
| ap = argparse.ArgumentParser() | ||
| ap.add_argument("-d", "--dir", required=True, | ||
| help="Directory for Files") | ||
| ap.add_argument("-s", "--size", type=int, required=True, | ||
| help="Chunk Size") | ||
| ap.add_argument("-g", "--grid", type=float, required=True, | ||
| help="Downsample Amount") | ||
|
|
||
| args = vars(ap.parse_args()) | ||
|
|
||
| directory = args['dir'] # "/media/jacob/WDRed8Tb1/Insync/iact_events/" | ||
| gamma_dir = [directory + "gammaFeature/no_clean/"] | ||
| proton_dir = [directory + "protonFeature/no_clean/"] | ||
|
|
||
| x, y = get_chunk_of_data(directory=gamma_dir, proton_directory=proton_dir, indicies=(30, 129, 10), rebin=100, | ||
| chunk_size=args['size'], as_channels=True) | ||
|
|
||
| x = x.reshape(100,100,10,1) | ||
|
|
||
| print("Got data") | ||
| print("X Shape", x.shape) | ||
| print("Y Shape", y.shape) | ||
| history = ta.Scan(x, y, | ||
| params=params, | ||
| dataset_name='3d_separation_test', | ||
| experiment_no='1', | ||
| model=input_model, | ||
| search_method='random', | ||
| grid_downsample=args['grid']) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.