Skip to content

Data-Science-and-Analytics-Club/ANNs-using-Tensorflow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Introduction to TensorFlow and Keras

TensorFlow: A Powerful Machine Learning Framework

What is TensorFlow?

TensorFlow stands as a robust open-source machine learning framework, initially developed by the Google Brain team. Launched in 2015, TensorFlow has become a cornerstone for building and training intricate deep learning models across a diverse range of applications.

Key Concepts in TensorFlow:

  • Tensor:

At the core of TensorFlow lies the concept of a tensor—a multi-dimensional array that serves as the fundamental data structure. Tensors can represent scalar values, vectors, matrices, or higher-dimensional structures.

  • Graph:

TensorFlow adopts a computational graph paradigm to express and organize operations. In this graph, nodes symbolize mathematical operations, while edges denote the flow of data (tensors) between these operations, facilitating parallelism and distributed computing.

  • Session:

A TensorFlow session provides the environment for executing computational graphs. It manages resources and orchestrates the execution of operations, ensuring efficiency and optimization during model training. Keras: A High-Level Neural Networks API

What is Keras?

Keras complements TensorFlow as an open-source, high-level neural networks API written in Python. Renowned for its simplicity and versatility, Keras operates as an interface for building, configuring, and training neural network models, running seamlessly atop lower-level frameworks such as TensorFlow, Theano, and CNTK.

Key Features of Keras:

  • User-Friendly:

Keras distinguishes itself through its user-friendly interface, striking a balance between accessibility for beginners and robustness for advanced users. The API simplifies the intricate process of neural network development.

  • Modularity:

The modular architecture of Keras enables the construction of models as a sequence of layers, each serving a distinct purpose. This modular design facilitates transparency and ease in model construction and understanding.

  • Compatibility:

Keras exhibits compatibility with various backends, offering flexibility and adaptability to different computational engines. Notably, it is widely used with TensorFlow, the default backend.

  • Extensibility:

The extensible nature of Keras allows users to seamlessly incorporate additional functionalities or introduce custom components, contributing to the versatility of the API. Building Blocks in Keras:

  • Layers:

In Keras, layers represent the basic building blocks of neural networks. These layers process input data, transforming it through a series of operations to yield meaningful representations.

  • Models:

Keras introduces the concept of models—a linear stack of layers. The most common model type is the sequential model, where layers are sequentially added, forming the neural network architecture.

  • Compile:

The compilation phase configures the learning process. Users specify essential elements such as the optimizer, loss function, and metrics, tailoring the model to the specifics of the learning task.

  • Fit:

The fitting process involves training the model on input data. Parameters like the number of epochs (iterations) and batch size are defined, guiding the model through the learning process. Evaluate and Predict:

Post-training, models are evaluated on test data to assess performance metrics. Additionally, trained models can be deployed to make predictions on new, unseen data.

About

Car engine type prediction using Artificial Neural Networks and Used-cars-catalog dataset

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages