Theano Python Library

pip install theano

Description

Theano is a python library and an optimizing compiler for manipulating and evaluating mathematical expressions involving multi-dimensional arrays. It is built on Numpy and named after a Greek mathematician and philosopher.
In Theano, computations are expressed using a syntax very similar to NumPy's and compiled to run efficiently on either CPU or GPU architectures.

Theano was developed by the Montreal Institute for Learning Algorithms (MILA), University of Montreal in 2007 and It has been powering large-scale computationally intensive scientific research since then. It is one of the most mature deep learning frameworks in existence, and is open sourced under the BSD license for all to use.

Developed specifically to handle the types of computation required for large neural network algorithms used in Deep Learning.

 

Features of Theano

1. It has a tight integraion with Numpy.

2. Performs efficient symbolic differentiation.

3. Fast to write and execute.

4. Parallelism on GPU. (performs computation much faster on a GPU)

5. Includes tools for extensive unit-testing and self-verification.

6. Evaluates expressions faster by dynamic generation of C code.

7. Stability optimizations

 

Uses of Theano

Here are some uses of theano:

1. Theano is used to implement deep leanring models.

2. Used for regular mathematical computation and research.

 

Installation

You can easily install Theano by running

pip install Theano

 

Comparison

Theano and Tensorflow

Theano
Tensorflow
1.
Not very known.
 Tensorflow is the most famous tool for deep learning
2.
Performs calculations faster than tensorflow.
Slower compared to Theano.
3.
Written in Python
Written in C++ and Python
4.
Cross platform
Cross-platform
5. 
Large Community due to its maturity.
Faster growing community.
6.
Extensive documentation
Good Documentation
7.
Comes with pre-existing trained models.
Also has pre-trained models
8.
Open-source
Also open-source

 

Theano and Scikit Learn

Theano
Scikit Learn
1.
Used for scientific computing and deep learning.
Used for general purpose machine learning
2.
Open source
Open source

 


Subscribe to receive updates!