Data Reporting and Post Training Data Evaluation

This notebook contains information on how to evaluate results generated by HARDy. The sample outcomes used in this notebook were the result of transformation configuration and classifier configuration files available in the repository.

The data on which classification model were tuned was Small-angle Scattering (SAXS) data generated through procedure mentioned in this guide. Both the Cartesian Coordinate representation and RGB representation were evaluated.

This notebook is binder ready. Please uncomment the next cell to code and run it to use this notebook in binder.

 #!: $(pip install ../../)

Step 1: Extraction of results

Extracting the results obtained from cartesian representation

!tar -xzf './scat_cart_pp.tar.gz' -C '.'

Extracting the results obtained from RGB representation

!tar -xzf './scat_rgb_pp.tar.gz' -C '.'

Importing the required libraries for setting environment parameters to visualize the dataframes properly

import pandas as pd
pd.set_option('max_rows', 9999)

Step 2: Import HARDy data reporting

from hardy import reporting

Using TensorFlow backend.

Step 3: Running data reporting module

For Cartesian representation

loss_accuracy, parallel = reporting.summary_report_plots('./scat_rgb_cart_1/')

Visualizing results

loss_accuracy.show()

parallel.show()

For RGB representation

loss_accuracy_rgb, parallel_rgb = reporting.summary_report_plots('./scat_rgb_pp/')

loss_accuracy_rgb.show()

parallel_rgb.show()

Using further tabulation modules available

hyperparameter_df, history_df, tform_rankdf = reporting.report_dataframes('./scat_rgb_pp/')

hyperparameter_df

	report_name	layers	kernel_size	activation_function	optimizer	pooling	test_accuracy
0	lin_q_lin_I	4	3	relu	adam	max	0.803
1	log_q_der_I	3	5	relu	adam	max	0.901
2	lin_q_rec_I	4	4	relu	SGD	max	0.769
3	multi_transform	3	3	sigmoid	adam	max	0.839
4	der_q_log_I	5	3	relu	adam	max	0.805
5	sqr_q_lin_I	5	3	sigmoid	adam	max	0.250
6	log_q_lin_I	5	3	relu	adam	avg	0.831
7	log_q_sqr_I	4	4	relu	adam	avg	0.604
8	lin_q_log_I	4	4	relu	SGD	max	0.365
9	rec_q_rec_I	4	3	relu	adam	max	0.804
10	rec_q_lin_I	3	5	relu	SGD	max	0.250
11	sqr_q_log_I	3	5	relu	adam	max	0.618
12	sqr_q_sqr_I	5	3	relu	adam	max	0.691
13	der_q_der_I	3	5	sigmoid	adam	avg	0.626
14	lin_q_sqr_I	5	3	relu	adam	avg	0.703
15	log_q_log_I	4	4	relu	adam	avg	0.459

history_df

	report_name	epochs	train_loss	val_loss	test_loss	train_accuracy	val_accuracy	test_accuracy
0	lin_q_lin_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[0.8081760375564163, 0.6698134615471413, 0.643...	[0.6929170756504454, 0.6729079228023003, 0.605...	0.424934	[0.5958858728408813, 0.6530330181121826, 0.665...	[0.6251351237297058, 0.6432432532310486, 0.706...	0.803000
1	log_q_der_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.3147423730168615, 0.9628631132429426, 0.620...	[1.0729141317564865, 0.7326729235977962, 0.490...	0.253629	[0.3543243110179901, 0.570480465888977, 0.7474...	[0.5170270204544067, 0.7113513350486755, 0.812...	0.901000
2	lin_q_rec_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[0.9746777676104067, 0.774592170994561, 0.7245...	[0.9451659054591738, 0.7242444527560267, 0.687...	0.474972	[0.5410210490226746, 0.614954948425293, 0.6388...	[0.5735135078430176, 0.6283783912658691, 0.654...	0.768667
3	multi_transform	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[0.9804753610822889, 0.7650474348154154, 0.684...	[0.8447682929450068, 0.7578490273705845, 0.733...	0.370734	[0.541501522064209, 0.6331531405448914, 0.6908...	[0.5764864683151245, 0.6618918776512146, 0.654...	0.839333
4	der_q_log_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.1135094868456636, 0.7191920271721688, 0.585...	[0.7560349147895287, 0.6721863993282976, 0.587...	0.393638	[0.5552252531051636, 0.6741141080856323, 0.730...	[0.6781080961227417, 0.6951351165771484, 0.750...	0.804667
5	sqr_q_lin_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.4043014319880947, 1.4011263673226755, 1.398...	[1.4084424068187844, 1.3933849293610145, 1.396...	1.386304	[0.25387388467788696, 0.25153154134750366, 0.2...	[0.2405405342578888, 0.25567567348480225, 0.25...	0.250000
6	log_q_lin_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.3667990244902648, 1.3610757085296126, 1.356...	[1.3047192507776721, 1.351464189332107, 1.3487...	0.423266	[0.3101801872253418, 0.30681681632995605, 0.30...	[0.44351351261138916, 0.29540541768074036, 0.3...	0.830667
7	log_q_sqr_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.378919125934979, 1.365878099378523, 1.36232...	[1.3658991764331687, 1.3594434281875347, 1.355...	0.977161	[0.2797597646713257, 0.30243241786956787, 0.30...	[0.3097297251224518, 0.29702702164649963, 0.30...	0.604000
8	lin_q_log_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]	[1.210034712714118, 1.0089355350256681, 0.9367...	[0.9949895297658855, 0.8996418241796822, 0.873...	1.320202	[0.41822823882102966, 0.5132432579994202, 0.54...	[0.523783802986145, 0.5743243098258972, 0.5505...	0.364667
9	rec_q_rec_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.1603663299892757, 0.9183210932957876, 0.813...	[0.9965362641318091, 0.819087128187048, 0.7786...	0.497543	[0.4788588583469391, 0.6230930685997009, 0.671...	[0.5867567658424377, 0.6740540266036987, 0.697...	0.804333
10	rec_q_lin_I	[1, 2, 3, 4, 5, 6, 7, 8, 9]	[1.38856608064325, 1.3867126006765051, 1.38661...	[1.386404105301561, 1.3862949856396378, 1.3868...	1.386523	[0.2456456422805786, 0.2475675642490387, 0.246...	[0.2405405342578888, 0.25567567348480225, 0.24...	0.250000
11	sqr_q_log_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.2729175752920432, 0.9154544173203432, 0.845...	[1.0506971036565715, 0.8518056088480456, 0.847...	0.757464	[0.3636035919189453, 0.5509609580039978, 0.577...	[0.4881080985069275, 0.5737837553024292, 0.579...	0.618333
12	sqr_q_sqr_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[0.9430359243129467, 0.7854952984672409, 0.762...	[0.7861571579143919, 0.7752227721543148, 0.746...	0.637161	[0.5370270013809204, 0.6112312078475952, 0.622...	[0.6067567467689514, 0.6108108162879944, 0.617...	0.691000
13	der_q_der_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[2.0273581344157727, 1.3065366548962063, 1.240...	[1.4033411453510154, 1.2285571612160782, 1.100...	0.758060	[0.3035435378551483, 0.3927627503871918, 0.429...	[0.25297296047210693, 0.3813513517379761, 0.50...	0.626333
14	lin_q_sqr_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.054254633914959, 0.8097827472199907, 0.7764...	[0.823198978243203, 0.7806079634304705, 0.7823...	0.589434	[0.44318318367004395, 0.5674474239349365, 0.60...	[0.550540566444397, 0.6202702522277832, 0.6343...	0.702667
15	log_q_log_I	[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...	[1.379027757043237, 1.3666725359736263, 1.3624...	[1.3648796985889304, 1.3617369388711864, 1.355...	1.207866	[0.29105105996131897, 0.3072372376918793, 0.31...	[0.3218919038772583, 0.28837838768959045, 0.31...	0.459333

tform_rankdf

	report_name	test_accuracy
0	lin_q_lin_I	0.803
1	log_q_der_I	0.901
2	lin_q_rec_I	0.769
3	multi_transform	0.839
4	der_q_log_I	0.805
5	sqr_q_lin_I	0.250
6	log_q_lin_I	0.831
7	log_q_sqr_I	0.604
8	lin_q_log_I	0.365
9	rec_q_rec_I	0.804
10	rec_q_lin_I	0.250
11	sqr_q_log_I	0.618
12	sqr_q_sqr_I	0.691
13	der_q_der_I	0.626
14	lin_q_sqr_I	0.703
15	log_q_log_I	0.459

---