pyNNRW

pyNNRW: A Python library for NNRW (neural network with random weights).

Basic functions:

Implements 2 fundamental NNRW flavors, i.e., ELM and RVFL.
Performance comparison with main machine learning models, e.g., SVM, decision tree, MLP.
NNRW-based ensembles (in progress).

Publication

Spectroscopic Profiling-based Geographic Herb Identification by Neural Network with Random Weights [J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2022, doi: 10.1016/j.saa.2022.121348

Installation

pip install pyNNRW

How to use

Download the sample dataset from the /data folder. There are two data files: 1. 7044X_RAMAN.csv 2. 7143X_UV.csv.
The two files are the Raman and UV (ultra-violet) spectroscopic profiling data of herb samples from 4 different regions.

Use the following sample code to use the package:

<1> Use low-level classes, e.g., ELM, RVFL. The following code trains and tests an ELM model.

# ===============================
# Import library
# ===============================
# import the library
from pyNNRW import elm, rvfl

# ===============================
# Load dataset
# ===============================
df = pd.read_csv('7044X_RAMAN.csv')
X = np.array(df.iloc[:,1:])
y = np.array(df.iloc[:,0]) # 1st col is the label
n_classes = len(set(y))
x_train, x_test, t_train, t_test = train_test_split(X, y, test_size=0.2)
t_train = to_categorical(t_train, n_classes).astype(np.float32)
t_test = to_categorical(t_test, n_classes).astype(np.float32)
# print(x_train.shape, x_test.shape, t_train.shape, t_test.shape)

# ===============================
# set ELM parameters
# ===============================
n_hidden_nodes = L #x_train.shape[1]
loss = 'mean_squared_error' # 'mean_absolute_error'
activation = 'sigmoid' # 'identity'

# ===============================
# Instantiate ELM
# ===============================
model = elm.ELM( # or rvfl.RVFL
    n_input_nodes = x_train.shape[1],
    n_hidden_nodes = n_hidden_nodes,
    n_output_nodes = n_classes,
    loss = loss,
    activation=activation,
    name='elm'
)

# ===============================
# Training
# ===============================
    
train_loss, train_acc, train_precision, train_recall = model.evaluate(x_train, t_train, metrics=['loss', 'accuracy', 'precision', 'recall'])

# ===============================
# Validation
# ===============================
val_loss, val_acc, val_precision, val_recall = model.evaluate(x_test, t_test, metrics=['loss', 'accuracy', 'precision', 'recall'])

<2> You may also use high-level APIs, as follows.

from pyNNRW import nnrw

# train and test an ELM model
train_acc, val_acc, t = nnrw.ELMClf(X, y, L = 20, verbose = False) # L is hidden layer nodes

# train and test a RVFL model
train_acc, val_acc, t = nnrw.RVFLClf(X, y, L = 20, verbose = False) # L is hidden layer nodes

# Conduct a performance test for ELM at varied L hyper-parameters (1~60). Each iteration is averaged on 20 rounds.
train_accs, val_accs, ts = nnrw.PerformenceTests(ELMClf, X, y, Ls = list(range(1, 60)), N = 20)

New function in v0.2.0

We added Kernel-NNRW, which provides a series of kernels combined with NNRW.

Metadata

Release files for pyNNRW 0.3.5

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for pyNNRW 0.3.5
File	Size	Uploaded
pyNNRW-0.3.5.tar.gz	30.1 kB	Jul 14, 2023	Details

Built distribution (wheel)

Table of built distributions (wheels) for pyNNRW 0.3.5
File	Interpreter	ABI	Platform	Reset
pyNNRW-0.3.5-py3-none-any.whl 32.8 kB Jul 14, 2023	Python 3	none	any	Details

Total release size: 63.0 kB

Release files / pyNNRW-0.3.5.tar.gz

Download URL	pyNNRW-0.3.5.tar.gz
Size	30.1 kB
Tags	Source
SHA-256 checksum How to use checksums	`5ed56817b747cd7106ea10f342faf97bbfb8eabd69c3aa74b5791deee34cf9e8`
BLAKE2b-256 checksum How to use checksums	`e96872cd8535d75d69b150a153495e9e9ab608c08e70e4d96009c39d23b57754`
Upload date	Jul 14, 2023
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/4.0.1 CPython/3.9.12`

Release files / pyNNRW-0.3.5-py3-none-any.whl

Download URL	pyNNRW-0.3.5-py3-none-any.whl
Size	32.8 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`ab0b52c17d44486eb5a1cc338d9bef3922161b0b50d95255af4cbfd999e65ae7`
BLAKE2b-256 checksum How to use checksums	`6bf92e77be271b0a671afafbe3cfe8689a81c03a9c3e940f1f86a78e25ddc941`
Upload date	Jul 14, 2023
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/4.0.1 CPython/3.9.12`