deforce: Derivative-Free Algorithms for Optimizing Cascade Forward Neural Networks
Project description
deforce: Derivative-Free Algorithms for Optimizing Cascade Forward Neural Networks
deforce (DErivative Free Optimization foR Cascade forward nEural networks) is a Python library that implements variants and the traditional version of Cascade Forward Neural Networks. These include Derivative Free-optimized CFN models (such as GA, PSO, WOA, TLO, DE, ...) and Gradient Descent-optimized CFN models (such as SGD, Adam, Adelta, Adagrad, ...). It provides a comprehensive list of optimizers for training CFN models and is also compatible with the Scikit-Learn library. With deforce, you can perform searches and hyperparameter tuning using the features provided by the Scikit-Learn library.
- Free software: GNU General Public License (GPL) V3 license
- Provided Estimator: CfnRegressor, CfnClassifier, DfoCfnRegressor, DfoCfnClassifier
- Total DFO-based CFN Regressor: > 200 Models
- Total DFO-based CFN Classifier: > 200 Models
- Total GD-based CFN Regressor: 12 Models
- Total GD-based CFN Classifier: 12 Models
- Supported performance metrics: >= 67 (47 regressions and 20 classifications)
- Supported objective functions (as fitness functions or loss functions): >= 67 (47 regressions and 20 classifications)
- Documentation: https://deforce.readthedocs.io
- Python versions: >= 3.8.x
- Dependencies: numpy, scipy, scikit-learn, pandas, mealpy, permetrics, torch, skorch
Citation Request
If you want to understand how Metaheuristic is applied to CFNN, you need to read the paper titled "Optimization of neural-network model using a meta-heuristic algorithm for the estimation of dynamic Poisson’s ratio of selected rock types". The paper can be accessed at the following link
Please include these citations if you plan to use this library:
@article{van2023mealpy,
title={MEALPY: An open-source library for latest meta-heuristic algorithms in Python},
author={Van Thieu, Nguyen and Mirjalili, Seyedali},
journal={Journal of Systems Architecture},
year={2023},
publisher={Elsevier},
doi={10.1016/j.sysarc.2023.102871}
}
@article{van2023groundwater,
title={Groundwater level modeling using Augmented Artificial Ecosystem Optimization},
author={Van Thieu, Nguyen and Barma, Surajit Deb and Van Lam, To and Kisi, Ozgur and Mahesha, Amai},
journal={Journal of Hydrology},
volume={617},
pages={129034},
year={2023},
publisher={Elsevier}
}
@article{thieu2019efficient,
title={Efficient time-series forecasting using neural network and opposition-based coral reefs optimization},
author={Thieu Nguyen, Tu Nguyen and Nguyen, Binh Minh and Nguyen, Giang},
journal={International Journal of Computational Intelligence Systems},
volume={12},
number={2},
pages={1144--1161},
year={2019}
}
Installation
- Install the current PyPI release:
$ pip install deforce==0.1.0
- Install directly from source code
$ git clone https://github.com/thieu1995/deforce.git
$ cd deforce
$ python setup.py install
- In case, you want to install the development version from Github:
$ pip install git+https://github.com/thieu1995/deforce
After installation, you can import deforce as any other Python module:
$ python
>>> import deforce
>>> deforce.__version__
Examples
Please check all use cases and examples in folder examples.
- deforce provides this useful classes
from deforce import DataTransformer, Data
from deforce import CfnRegressor, CfnClassifier
from deforce import DfoCfnRegressor, DfoCfnClassifier
- What you can do with
DataTransformer
class
We provide many scaler classes that you can select and make a combination of transforming your data via DataTransformer class. For example:
2.1) I want to scale data by Loge
and then Sqrt
and then MinMax
:
from deforce import DataTransformer
import pandas as pd
from sklearn.model_selection import train_test_split
dataset = pd.read_csv('Position_Salaries.csv')
X = dataset.iloc[:, 1:5].values
y = dataset.iloc[:, 5].values
X_train, y_train, X_test, y_test = train_test_split(X, y, test_size=0.2)
dt = DataTransformer(scaling_methods=("loge", "sqrt", "minmax"))
X_train_scaled = dt.fit_transform(X_train)
X_test_scaled = dt.transform(X_test)
2.2) I want to scale data by YeoJohnson
and then Standard
:
from deforce import DataTransformer
import pandas as pd
from sklearn.model_selection import train_test_split
dataset = pd.read_csv('Position_Salaries.csv')
X = dataset.iloc[:, 1:5].values
y = dataset.iloc[:, 5].values
X_train, y_train, X_test, y_test = train_test_split(X, y, test_size=0.2)
dt = DataTransformer(scaling_methods=("yeo-johnson", "standard"))
X_train_scaled = dt.fit_transform(X_train)
X_test_scaled = dt.transform(X_test)
- What can you do with
Data
class
- You can load your dataset into Data class
- You can split dataset to train and test set
- You can scale dataset without using DataTransformer class
- You can scale labels using LabelEncoder
from deforce import Data
import pandas as pd
dataset = pd.read_csv('Position_Salaries.csv')
X = dataset.iloc[:, 1:5].values
y = dataset.iloc[:, 5].values
data = Data(X, y, name="position_salaries")
#### Split dataset into train and test set
data.split_train_test(test_size=0.2, shuffle=True, random_state=100, inplace=True)
#### Feature Scaling
data.X_train, scaler_X = data.scale(data.X_train, scaling_methods=("standard", "sqrt", "minmax"))
data.X_test = scaler_X.transform(data.X_test)
data.y_train, scaler_y = data.encode_label(data.y_train) # This is for classification problem only
data.y_test = scaler_y.transform(data.y_test)
- What can you do with all model classes
- Define the model
- Use provides functions to train, predict, and evaluate model
from deforce import CfnRegressor, CfnClassifier, DfoCfnRegressor, DfoCfnClassifier
## Use standard CFN model for regression problem
regressor = CfnRegressor(hidden_size=50, act1_name="tanh", act2_name="sigmoid", obj_name="MSE",
max_epochs=1000, batch_size=32, optimizer="SGD", optimizer_paras=None, verbose=False)
## Use standard CFN model for classification problem
classifier = CfnClassifier(hidden_size=50, act1_name="tanh", act2_name="sigmoid", obj_name="NLLL",
max_epochs=1000, batch_size=32, optimizer="SGD", optimizer_paras=None, verbose=False)
## Use Metaheuristic-optimized CFN model for regression problem
print(DfoCfnClassifier.SUPPORTED_OPTIMIZERS)
print(DfoCfnClassifier.SUPPORTED_REG_OBJECTIVES)
opt_paras = {"name": "WOA", "epoch": 100, "pop_size": 30}
regressor = DfoCfnRegressor(hidden_size=50, act1_name="tanh", act2_name="sigmoid",
obj_name="MSE", optimizer="OriginalWOA", optimizer_paras=opt_paras, verbose=True)
## Use Metaheuristic-optimized CFN model for classification problem
print(DfoCfnClassifier.SUPPORTED_OPTIMIZERS)
print(DfoCfnClassifier.SUPPORTED_CLS_OBJECTIVES)
opt_paras = {"name": "WOA", "epoch": 100, "pop_size": 30}
classifier = DfoCfnClassifier(hidden_size=50, act1_name="tanh", act2_name="softmax",
obj_name="CEL", optimizer="OriginalWOA", optimizer_paras=opt_paras, verbose=True)
- What can you do with model object
from deforce import CfnRegressor, Data
data = Data() # Assumption that you have provide this object like above
model = CfnRegressor(hidden_size=50, act1_name="tanh", act2_name="sigmoid", obj_name="MSE",
max_epochs=1000, batch_size=32, optimizer="SGD", optimizer_paras=None, verbose=False)
## Train the model
model.fit(data.X_train, data.y_train)
## Predicting a new result
y_pred = model.predict(data.X_test)
## Calculate metrics using score or scores functions.
print(model.score(data.X_test, data.y_test, method="MAE"))
print(model.scores(data.X_test, data.y_test, list_methods=["MAPE", "NNSE", "KGE", "MASE", "R2", "R", "R2S"]))
## Calculate metrics using evaluate function
print(model.evaluate(data.y_test, y_pred, list_metrics=("MSE", "RMSE", "MAPE", "NSE")))
## Save performance metrics to csv file
model.save_evaluation_metrics(data.y_test, y_pred, list_metrics=("RMSE", "MAE"), save_path="history",
filename="metrics.csv")
## Save training loss to csv file
model.save_training_loss(save_path="history", filename="loss.csv")
## Save predicted label
model.save_y_predicted(X=data.X_test, y_true=data.y_test, save_path="history", filename="y_predicted.csv")
## Save model
model.save_model(save_path="history", filename="traditional_CFN.pkl")
## Load model
trained_model = CfnRegressor.load_model(load_path="history", filename="traditional_CFN.pkl")
Support (questions, problems)
Official Links
-
Official source code repo: https://github.com/thieu1995/deforce
-
Official document: https://metapeceptron.readthedocs.io/
-
Download releases: https://pypi.org/project/deforce/
-
Issue tracker: https://github.com/thieu1995/deforce/issues
-
Notable changes log: https://github.com/thieu1995/deforce/blob/master/ChangeLog.md
-
Official chat group: https://t.me/+fRVCJGuGJg1mNDg1
-
This project also related to our another projects which are "optimization" and "machine learning", check it here:
- https://github.com/thieu1995/mealpy
- https://github.com/thieu1995/metaheuristics
- https://github.com/thieu1995/opfunu
- https://github.com/thieu1995/enoppy
- https://github.com/thieu1995/permetrics
- https://github.com/thieu1995/MetaCluster
- https://github.com/thieu1995/pfevaluator
- https://github.com/thieu1995/IntelELM
- https://github.com/thieu1995/reflame
- https://github.com/aiir-team
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.