# Setting up Tensorflow with Keras

Following up the deep learning workshop presented in SIS (slides), this note introduces how to setup TensorFlow and Keras in your own computer without affecting your existing installations of Python.

## Setup

2. Create a sperated python environment by issuing the following command:

3. To activate this environment:

4. You need to activate this environment everytime you want to play with TensorFlow+Keras

5. Install TensorFlow by invoking:

6. Install Keras by invoking:

7. If you encounter scipy installation erros here (look for red words), you can invoke the following command and install keras again.

8. Now you can call your python script:

9. Or you can play with a python notebook:

# Compile LibVex on Windows

In order to use LibVex in Kam1n0, we need to compile libVex from valgrind on windows.
Dependencies:

• mingw64 with msys tool installed
• add mingw64/bin and msys/bin to environment variable

Clone libvex source (from angr repo).

• git clone git@github.com:angr/vex

We need to update the Makefile-gcc. Specifically we need to define cc and ar.

Also we need to re-define HWord in libvex to long long int (64bit)
Then just hit make; and we can find the libvex.a file in the vex-master directory.

# Sklean+Xgboost Cross Validation GridSearch Tuning

This note illustrates an example using Xgboost with Sklean to tune the parameter using cross-validation. The example is based on our recent task of age regression on personal information management data. The code covers: