oopd-iiitd-group10 0.0.1

A small example package

Two Layer Perceptron

A Two Layer Perceptron(specialized Multi Layer Perceptron) from scratch. The whole program is implemented on the principles of Object Oriented Programming Design. The program fetches mnist data, stores into a database then retrieves it using APIs. These APIs for creation, storing and retreiving from the database has also been implemented from scratch on the lines of OOPD.

Authors

• Giridhar S. [MT21026]
• Palani Vigneshwar [MT21062]
• Shashwat Vaibhav [Mt21082]
• Arpit Mathur [MT20328]

Necessary Modules

The modules and standard routines need to be pre-installed are as follows:-

• numpy
• pandas
• matplotlib
• sklearn
• sqlite3
• pycallgraph

Necessary Imports

• numpy
• pandas
• matplotlib.pyplot
• matplotlib.image
• sklearn.datasets.load_digits
• sqlite3
• sklearn.preprocessing.StandardScaler
• sklearn.preprocessing.MinMaxScaler
• sklearn.model_selection.train_test_split
• sklearn.metrics.classification_report
• sklearn.model_selection accuracy_score
• pycallgraph.PyCallGraph
• pycallgraph.output.GraphvizOutput
• pycallgraph.Config
• pycallgraph.GlobbingFilter

Classes Defined

DescribeNCreate:

• provides methods to get information, descriptive statistics and dataframe creation utility
• accepts bunch dataset during instantiation.

DataBASE:

• Inherits from DescribeNCreate class
• Provides method to create and store to a database from dataframe proovided to it.
• Implements all the Schema creation, insertion and exception handling for Database Error from scratch.

FetchFB:

• Inherits from DataBASE class
• provides utility to fetch database rows by implementing fetch query from scratch.
• performs exception handling.

newDataBase:

• instantiated by providing database name and table name of our choice.
• provide methods to store the results from 2 Layer Perceptron into a database.
• performs schema creation and Insertion queries from scratch.

Activation_Function:

• It acts as a generic class to implement several activation functions such as Sigmoid, ReLu, Tanh and Softmax.

propagation:

• inherits from Activation_Function class.
• implements methods necessary for forward and backward propagation.

tlp:

• most imporrtant of all, inherits from propagation class.
• provides methods to set weights, hyperparameters parameters, initialization and updation.
• provides the important fit and predict utility.
• provides accuracy_score utility as well.
• all the methods implemented from scratch.

Folders,Files wheel and installation info @Giridhar

oopd_iiitd_group9-0.0.1-py3-none-any.whl:

• Wheel file which can be installed via -> pip install oopd_iiitd_group9-0.0.1-py3-none-any.whl
• the classes in the wheel file can be accessed as follows -> from project.code import *
• the project above is a folder in the wheel file

main:

• This folder contains the files with running code
• This folder contains two files main_code.py and code_with_whl.py

main_code.py:

• This file contains the main code with all the classes.

code_with_whl.py:

• This file contains only the code in main(). The whl file installed is used here and the classes from the wheel file is used here
• Please look into oopd_iiitd_group9-0.0.1-py3-none-any.whl section

src:

• This folder is used to create the whl file i.e. for packaging
• It is not the main running folder.

dist:

• This also contains the .whl file and got why building the file

doxygen_html:

• This folder contains all the html files generated by doxygen.

profiling_pycallgraph:

• This is the profiling report created using pycallgraph and is in the format of .png

UML Class diagram:

• This is the UML class diagram visualising the classes and the relations in the main code.

setup.cfg

• Used in creating the wheel file (.whl file)

pyproject.toml

• Used in creating the wheel file (.whl file)