Skip to main content

Getting started with Tensorflow , keras and theano - Development setup with Anaconda Installation


Below are the steps to setup your development environment  for Deep learning :


1.) Download and Install Anaconda from here :https://www.continuum.io/downloads
2.) Create a conda environment for data science development so that it doesn't affect the other install components .
    conda create -n tensor_keras_py3.5 python=3.5 pandas scikit-learn jupyter matplotlib
3.) Activate the created environment
    source activate tensor_keras_py3.5
4.) Install tensorflow inside activated env.
    pip install tensorflow
5.) Install keras inside activated env.
   pip install keras
6.) Install opencv inside activated env.
   pip install opencv-python
7.) Install IMUTILS
 pip install imutils


Test your environment


1.) Type ipython in the shell , which should open ipython console .
2.) Type import tensorflow,keras  , it should reply using tensorflow backend 



Switching keras backend from Tensorflow to theano

keras backend is set in a hidden file stored in your home path . You can find it at $/.keras/keras.json . You can open it with a text editor and you should see something like this :


{ "image_dim_ordering": "tf", "epsilon": 1e-07, "floatx": "float32", "backend": "tensorflow" }



You can change backend to theano . save it and close it . Now if you open ipython and do import keras then it will return using theano backend . 






Comments

Post a Comment

Popular posts from this blog

MonoLithic Vs Microservice Architecture | which Architecture should i choose ?

From last few years ,microservices are an accelerating trend . Indeed, microservices approach offers tangible benefits including an increase in scalability, flexibility, agility, and other significant advantages. Netflix, Google, Amazon, and other tech leaders have successfully switched from monolithic architecture to microservices. Meanwhile, many companies consider following this example as the most efficient way for business growth. On the contrary, the monolithic approach is a default model for creating a software application. Still, its trend is going down because building a monolithic application poses a number of challenges associated with handling a huge code base, adopting a new technology, scaling, deployment, implementing new changes and others. So is the monolithic approach outdated and should be left in the past? And is it worth shifting the whole application from a monolith to microservices ? Will developing a microservices application help you reach you...
Post a Comment
Read more

C Program to sort an array using insertion sort

I have seen couple of students who use swap in insertion sort to sort an array.Insertion sort is an example of exchange sort ,we don't swap two elements to sort an array ,instead of this we shift elements if ith element is greater than i+1th element to sort an array.It works on simple concept:- Store the ith element in temporary variable and shift element until you find the element smaller than this element towards 0 index position.Here is the C Program to find insertion sort:- _________________________________________________________________________________ #include <stdio.h> #include <string.h> main() {     int arrSize,i;    printf("Enter Size of array:");    scanf("%d",&arrSize);    int array[arrSize];    for(i=0;i<arrSize;i++){        scanf("%d",&array[i]);    }    int temp,j;    for(i=1;i<arrSize;i++){        temp...
3 comments
Read more

Perfect Hashing and its implementation in C

Perfect hashing simply means hashing with no collision.But there is nothing like perfect hashing(no collision).These are just a work around to achieve perfect hashing.Using this we can able to search the large set of records in a constant time i.e O(1) time complexity.To achieve searching in strict O(1) time ,one needs to implement hash table and then search the given record in to hash table using hash function.         Here in perfect hashing ,one need to implement two level hash table , first level hash table consist a set of pointers that points to the second level hash table,we must select hash function that first maps the key in to first level hash table,so it simply means there must be collision in first level hash table ,thats  why i am saying its just a work around to achieve prefect  hashing.We must wisely choose hash function:                  h(x)=((a*x+b)mod p)mod m x is the key to be map, Here ...
Post a Comment
Read more