notebook.community

Edit and run

Experiment:

Evaluate pruning by magnitude weighted by coactivations (more thorough evaluation), compare it to baseline (SET).

Motivation.

Check if results are consistently above baseline.

Conclusion

  • No significant difference between both models
  • No support for early stopping




In [44]:



    


from IPython.display import Markdown, display
%load_ext autoreload
%autoreload 2



    


















    






The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

















In [45]:



    


from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import glob
import tabulate
import pprint
import click
import numpy as np
import pandas as pd
from ray.tune.commands import *
from nupic.research.frameworks.dynamic_sparse.common.browser import *

import matplotlib.pyplot as plt
from matplotlib import rcParams

%config InlineBackend.figure_format = 'retina'

import seaborn as sns
sns.set(style="whitegrid")
sns.set_palette("colorblind")

Load and check data





In [46]:



    


base = os.path.join('gsc-smalldense-2019-10-11-exp1')
exps = [
    os.path.join(base, exp) for exp in [
        'gsc-smalldense'
    ]
]
paths = [os.path.expanduser("~/nta/results/{}".format(e)) for e in exps]
df = load_many(paths)



    


















    






gsc-smalldense

















In [51]:



    


df



    


















    
Out[51]:











  
    

      
      Experiment Name
      train_acc_max
      train_acc_max_epoch
      train_acc_min
      train_acc_min_epoch
      train_acc_median
      train_acc_last
      val_acc_max
      val_acc_max_epoch
      val_acc_min
      ...
      lr_gamma
      lr_milestones
      lr_scheduler
      model
      momentum
      net_params
      network
      optim_alg
      test_noise
      weight_decay
    

  
  
    

      0
      0_equivalent_on_perc=0.02,model=BaseModel
      0.722293
      92
      0.256079
      0
      0.705961
      0.714383
      0.792712
      45
      0.286050
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      1
      1_equivalent_on_perc=0.04,model=BaseModel
      0.836149
      92
      0.224685
      0
      0.826238
      0.831169
      0.905172
      65
      0.305643
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      2
      2_equivalent_on_perc=0.06,model=BaseModel
      0.874426
      65
      0.288595
      0
      0.864466
      0.870813
      0.936912
      78
      0.447492
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      3
      3_equivalent_on_perc=0.08,model=BaseModel
      0.894200
      92
      0.300410
      0
      0.880163
      0.891710
      0.942006
      78
      0.298197
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      4
      4_equivalent_on_perc=0.1,model=BaseModel
      0.894737
      96
      0.327995
      0
      0.887194
      0.890831
      0.949451
      55
      0.486677
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      5
      5_equivalent_on_perc=0.02,model=BaseModel
      0.714481
      92
      0.169905
      0
      0.698662
      0.707450
      0.781740
      85
      0.137539
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      6
      6_equivalent_on_perc=0.04,model=BaseModel
      0.836588
      98
      0.231081
      0
      0.824187
      0.830876
      0.907132
      98
      0.296630
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      7
      7_equivalent_on_perc=0.06,model=BaseModel
      0.872913
      93
      0.251343
      0
      0.862465
      0.869544
      0.928292
      46
      0.378918
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      8
      8_equivalent_on_perc=0.08,model=BaseModel
      0.895030
      91
      0.317547
      0
      0.882873
      0.894444
      0.943182
      45
      0.528997
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      9
      9_equivalent_on_perc=0.1,model=BaseModel
      0.901767
      96
      0.305146
      0
      0.890562
      0.890196
      0.949451
      32
      0.391066
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      10
      10_equivalent_on_perc=0.02,model=BaseModel
      0.706425
      61
      0.174202
      0
      0.522776
      0.688849
      0.789577
      99
      0.087774
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      11
      11_equivalent_on_perc=0.04,model=BaseModel
      0.838688
      98
      0.252856
      0
      0.827361
      0.837076
      0.909875
      66
      0.375784
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      12
      12_equivalent_on_perc=0.06,model=BaseModel
      0.873499
      80
      0.214530
      1
      0.863563
      0.868519
      0.928292
      76
      0.126567
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      13
      13_equivalent_on_perc=0.08,model=BaseModel
      0.897422
      96
      0.295821
      0
      0.886437
      0.889806
      0.943966
      92
      0.464342
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      14
      14_equivalent_on_perc=0.1,model=BaseModel
      0.901572
      90
      0.302217
      0
      0.890831
      0.890587
      0.951019
      77
      0.505486
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      15
      15_equivalent_on_perc=0.02,model=BaseModel
      0.711015
      92
      0.175471
      0
      0.701299
      0.707206
      0.793103
      97
      0.157524
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      16
      16_equivalent_on_perc=0.04,model=BaseModel
      0.836784
      93
      0.233083
      0
      0.826018
      0.829948
      0.913793
      99
      0.321317
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      17
      17_equivalent_on_perc=0.06,model=BaseModel
      0.873694
      96
      0.244751
      0
      0.857802
      0.868909
      0.929075
      63
      0.359326
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      18
      18_equivalent_on_perc=0.08,model=BaseModel
      0.898399
      91
      0.236989
      0
      0.886144
      0.890440
      0.941614
      25
      0.346003
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      19
      19_equivalent_on_perc=0.1,model=BaseModel
      0.905869
      92
      0.266771
      0
      0.895127
      0.902402
      0.950235
      92
      0.366379
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      20
      20_equivalent_on_perc=0.02,model=BaseModel
      0.717117
      98
      0.156625
      0
      0.705180
      0.709403
      0.784875
      50
      0.152038
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      21
      21_equivalent_on_perc=0.04,model=BaseModel
      0.835075
      92
      0.260668
      0
      0.820940
      0.829704
      0.911834
      73
      0.404781
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      22
      22_equivalent_on_perc=0.06,model=BaseModel
      0.873303
      94
      0.263988
      0
      0.863099
      0.869007
      0.929075
      50
      0.370298
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      23
      23_equivalent_on_perc=0.08,model=BaseModel
      0.893419
      96
      0.250952
      0
      0.880871
      0.889220
      0.938871
      76
      0.344828
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      24
      24_equivalent_on_perc=0.1,model=BaseModel
      0.903086
      98
      0.331511
      0
      0.892344
      0.901621
      0.949843
      45
      0.429075
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      25
      25_equivalent_on_perc=0.02,model=BaseModel
      0.716922
      96
      0.172542
      0
      0.702544
      0.713505
      0.792712
      74
      0.145376
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      26
      26_equivalent_on_perc=0.04,model=BaseModel
      0.834977
      93
      0.228347
      0
      0.824382
      0.826433
      0.911442
      87
      0.272335
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      27
      27_equivalent_on_perc=0.06,model=BaseModel
      0.874133
      95
      0.291964
      0
      0.863588
      0.867493
      0.929859
      94
      0.439655
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      28
      28_equivalent_on_perc=0.08,model=BaseModel
      0.896250
      98
      0.306171
      0
      0.880798
      0.874573
      0.947884
      92
      0.463166
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      29
      29_equivalent_on_perc=0.1,model=BaseModel
      0.899814
      95
      0.327702
      0
      0.884923
      0.897422
      0.947100
      97
      0.473354
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      30
      30_equivalent_on_perc=0.02,model=BaseModel
      0.718826
      92
      0.174397
      0
      0.704790
      0.712772
      0.784875
      23
      0.130486
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      31
      31_equivalent_on_perc=0.04,model=BaseModel
      0.837858
      92
      0.171712
      0
      0.822478
      0.834098
      0.907132
      91
      0.141066
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      32
      32_equivalent_on_perc=0.06,model=BaseModel
      0.875256
      91
      0.272190
      0
      0.865223
      0.869007
      0.929859
      83
      0.399295
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      33
      33_equivalent_on_perc=0.08,model=BaseModel
      0.893467
      80
      0.309833
      0
      0.883068
      0.888195
      0.945533
      96
      0.458856
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      34
      34_equivalent_on_perc=0.1,model=BaseModel
      0.899473
      97
      0.287374
      0
      0.886827
      0.896494
      0.951411
      85
      0.429467
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      35
      35_equivalent_on_perc=0.02,model=BaseModel
      0.717410
      95
      0.253198
      0
      0.705546
      0.713456
      0.783699
      49
      0.318574
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      36
      36_equivalent_on_perc=0.04,model=BaseModel
      0.839274
      90
      0.170735
      0
      0.824358
      0.829509
      0.907915
      62
      0.138323
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      37
      37_equivalent_on_perc=0.06,model=BaseModel
      0.873059
      96
      0.285324
      0
      0.863270
      0.870228
      0.936129
      67
      0.465909
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      38
      38_equivalent_on_perc=0.08,model=BaseModel
      0.896446
      99
      0.303877
      0
      0.882092
      0.896446
      0.947884
      45
      0.466301
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

    

      39
      39_equivalent_on_perc=0.1,model=BaseModel
      0.900303
      93
      0.294991
      0
      0.890880
      0.897666
      0.948276
      96
      0.493339
      ...
      0.9
      60.0
      MultiStepLR
      BaseModel
      0
      {'boost_strength': 1.5, 'boost_strength_factor...
      small_dense_gsc
      SGD
      False
      0.01
    

  



40 rows × 40 columns



















In [52]:



    


df.columns



    


















    
Out[52]:








Index(['Experiment Name', 'train_acc_max', 'train_acc_max_epoch',
       'train_acc_min', 'train_acc_min_epoch', 'train_acc_median',
       'train_acc_last', 'val_acc_max', 'val_acc_max_epoch', 'val_acc_min',
       'val_acc_min_epoch', 'val_acc_median', 'val_acc_last', 'val_acc_all',
       'epochs', 'experiment_file_name', 'experiment_base_path', 'trial_time',
       'mean_epoch_time', 'scatter_plot_dicts', 'batch_size_test',
       'batch_size_train', 'data_dir', 'dataset_name', 'debug_small_dense',
       'debug_sparse', 'debug_weights', 'device', 'equivalent_on_perc',
       'learning_rate', 'lr_gamma', 'lr_milestones', 'lr_scheduler', 'model',
       'momentum', 'net_params', 'network', 'optim_alg', 'test_noise',
       'weight_decay'],
      dtype='object')

















In [53]:



    


df.shape



    


















    
Out[53]:








(40, 40)

















In [54]:



    


df.iloc[1]



    


















    
Out[54]:








Experiment Name                 1_equivalent_on_perc=0.04,model=BaseModel
train_acc_max                                                    0.836149
train_acc_max_epoch                                                    92
train_acc_min                                                    0.224685
train_acc_min_epoch                                                     0
train_acc_median                                                 0.826238
train_acc_last                                                   0.831169
val_acc_max                                                      0.905172
val_acc_max_epoch                                                      65
val_acc_min                                                      0.305643
val_acc_min_epoch                                                       0
val_acc_median                                                   0.881661
val_acc_last                                                      0.88127
val_acc_all             0     0.305643
1     0.498433
2     0.567790
3...
epochs                                                                100
experiment_file_name    /Users/mcaporale/nta/results/gsc-smalldense-20...
experiment_base_path                                       gsc-smalldense
trial_time                                                        13.2269
mean_epoch_time                                                  0.132269
scatter_plot_dicts                                                     {}
batch_size_test                                                      1000
batch_size_train                                                       10
data_dir                                               ~/nta/datasets/gsc
dataset_name                                              PreprocessedGSC
debug_small_dense                                                    True
debug_sparse                                                         True
debug_weights                                                        True
device                                                               cuda
equivalent_on_perc                                                   0.04
learning_rate                                                        0.01
lr_gamma                                                              0.9
lr_milestones                                                          60
lr_scheduler                                                  MultiStepLR
model                                                           BaseModel
momentum                                                                0
net_params              {'boost_strength': 1.5, 'boost_strength_factor...
network                                                   small_dense_gsc
optim_alg                                                             SGD
test_noise                                                          False
weight_decay                                                         0.01
Name: 1, dtype: object

















In [56]:



    


df.groupby('equivalent_on_perc')['equivalent_on_perc'].count()



    


















    
Out[56]:








equivalent_on_perc
0.02    8
0.04    8
0.06    8
0.08    8
0.10    8
Name: equivalent_on_perc, dtype: int64

Analysis

Experiment Details

# experiment configurations base_exp_config = dict( device=("cuda" if torch.cuda.device_count() > 0 else "cpu"), # dataset related dataset_name="PreprocessedGSC", data_dir="~/nta/datasets/gsc", # batch_size_train=(4, 16), # batch_size_test=(1000), # # ----- Optimizer Related ---- # optim_alg="SGD", # momentum=0.0, # learning_rate=0.01, # weight_decay=1e-2, # # ----- LR Scheduler Related ---- # lr_scheduler="StepLR", # lr_step_size=1, # lr_gamma=0.9, batch_size_train=(4, 16), batch_size_test=1000, optim_alg="SGD", momentum=0, # 0.9, learning_rate=0.01, # 0.1, weight_decay=0.01, # 1e-4, lr_scheduler="MultiStepLR", lr_milestones=[30, 60, 90], lr_gamma=0.9, # 0.1, # additional validation test_noise=False, # debugging debug_weights=True, debug_sparse=True, ) # ray configurations experiment_name = "gsc-smalldense-2019-10-11-exp1" tune_config = dict( name=experiment_name, num_samples=8, local_dir=os.path.expanduser(os.path.join("~/nta/results", experiment_name)), checkpoint_freq=0, checkpoint_at_end=False, stop={"training_iteration": 100}, resources_per_trial={ "cpu": os.cpu_count() / 8, "gpu": 0.5, }, loggers=DEFAULT_LOGGERS, verbose=1, config=base_exp_config, ) # define experiments net_params = dict( boost_strength=1.5, boost_strength_factor=0.9, k_inference_factor=1.5, duty_cycle_period=1000 ) experiments = { "gsc-smalldense": dict( model=ray.tune.grid_search(["BaseModel"]), network="small_dense_gsc", net_params=net_params, equivalent_on_perc=ray.tune.grid_search([ 0.02, 0.04, 0.06, 0.08, 0.10, ]), debug_small_dense=True, ), } 




In [57]:



    


# Did any  trials failed?
df[df["epochs"]<100]["epochs"].count()



    


















    
Out[57]:








0

















In [58]:



    


# Removing failed or incomplete trials
df_origin = df.copy()
df = df_origin[df_origin["epochs"]>=30]
df.shape



    


















    
Out[58]:








(40, 40)

















In [59]:



    


# which ones failed?
# failed, or still ongoing?
df_origin['failed'] = df_origin["epochs"]<30
df_origin[df_origin['failed']]['epochs']



    


















    
Out[59]:








Series([], Name: epochs, dtype: int64)

















In [60]:



    


# helper functions
def mean_and_std(s):
    return "{:.3f} ± {:.3f}".format(s.mean(), s.std())

def round_mean(s):
    return "{:.0f}".format(round(s.mean()))

stats = ['min', 'max', 'mean', 'std']

def agg(columns, filter=None, round=3):
    if filter is None:
        return (df.groupby(columns)
             .agg({'val_acc_max_epoch': round_mean,
                   'val_acc_max': stats,                
                   'model': ['count']})).round(round)
    else:
        return (df[filter].groupby(columns)
             .agg({'val_acc_max_epoch': round_mean,
                   'val_acc_max': stats,                
                   'model': ['count']})).round(round)
Does improved weight pruning outperforms regular SET




In [108]:



    


df_agg = agg(['equivalent_on_perc'])
df_agg



    


















    
Out[108]:











  
    

      
      val_acc_max_epoch
      val_acc_max
      model
    

    

      
      round_mean
      min
      max
      mean
      std
      count
    

    

      equivalent_on_perc
      
      
      
      
      
      
    

  
  
    

      0.02
      65
      0.782
      0.793
      0.788
      0.005
      8
    

    

      0.04
      80
      0.905
      0.914
      0.909
      0.003
      8
    

    

      0.06
      70
      0.928
      0.937
      0.931
      0.004
      8
    

    

      0.08
      69
      0.939
      0.948
      0.944
      0.003
      8
    

    

      0.10
      72
      0.947
      0.951
      0.950
      0.001
      8
    

  




















In [109]:



    


equivalent_on_percs = df_agg.index.values

val_means = df_agg['val_acc_max']['mean']
val_means = np.array(val_means)
val_stds = df_agg['val_acc_max']['std']
val_stds = np.array(val_stds)



    











In [110]:



    


print(equivalent_on_percs)
print(val_means)
print(val_stds)



    


















    






[0.02 0.04 0.06 0.08 0.1 ]
[0.788 0.909 0.931 0.944 0.95 ]
[0.005 0.003 0.004 0.003 0.001]

















In [39]:



    


# translate model names
rcParams['figure.figsize'] = 16, 8
# d = {
#     'DSNNWeightedMag': 'DSNN',
#     'DSNNMixedHeb': 'SET',
#     'SparseModel': 'Static',        
# }
# df_plot = df.copy()
# df_plot['model'] = df_plot['model'].apply(lambda x, i: model_name(x, i))



    











In [117]:



    


plt.errorbar(
    x=equivalent_on_percs,
    y=val_means,
    yerr=val_stds,
    color='k',
    marker='*',
    lw=0,
    elinewidth=2,
    capsize=2,
    markersize=10,
    label="Small-Dense Equivalents"
)
# sns.scatterplot(data=df_plot, x='on_perc', y='val_acc_max', hue='model')
# sns.lineplot(data=df, x='equivalent_on_perc', y='val_acc_max', hue='equivalent_on_perc')



    


















    
Out[117]:








<ErrorbarContainer object of 3 artists>

Content source: numenta/nupic.research

Similar notebooks:

notebook.community | gallery | about