This course provides you to be able to build Deep Neural Networks models for different business domains with one of the most common machine learning library TensorFlow provided by Google AI team. The both concept of deep learning and its applications will be mentioned in this course. Also, we will focus on Keras. This course appeals to ones who interested in Machine Learning, Data Science and AI. Also, you don't have to be attend any ML course before.
This video includes installation of Deep Learning Framework Tensorflow and its prerequisites. Python 3.5.3, Anaconda 4.4.0 and Tensorflow 1.2.0 respectively on Windows 7 64-bit OS.
In previous post, we've gotten TensorFlow up. In this video, we are going to mention how to build deep neural networks classifier with TensorFlow. Classification is applied on Exclusive OR (XOR) gate dataset.
Actually, XOR gate solution is hello world program for machine learning studies. We will also focus the reason of it.
In this video, we have mentioned how to re-use already trained neural networks in TensorFlow. Thus, we can make predictions fast even though long learning time required systems.
In previous lecture, we've mentioned how to re-use trained neural networks in TensorFlow.
In this video, we've mentioned how to load and re-use trained TensorFlow neural networks in external higher level systems such as Java. Learning is implemented in TensorFlow whereas predictions are made in Java. TensorFlow also supports to be used in C++.
In this lecture, we will mention how to evaluate a machine learning model and commonly used metrics in ML studies. We will also monitor the change of these metrics over learning. And finally, we will focus how to use TensorBoard to monitor these metrics easily.
Until now, we have built deep neural networks classifiers. Neural networks can also build models for regression studies. Today, we will focus on how to build a deep neural networks regressor in TensorFlow. Sine wave non-linear time series dataset will be used in the study. Finally, we will mention to monitor time series forecasts in TensorBoard.
Until now, we have mentioned the out of the box drawing capabilities of TensorFlow and TensorBoard for monitoring. We can also consume python matplotlib library to monitor results of machine learning studies.
Even though TensorFlow is developed as a Deep Learning Framework, it is also powerful about other ML algorithms. Today, we will mention how to handle unsupervised learning with TensorFlow. And we will apply k-means clustering algorithm a dataset. Also, we will use matplotlib to visualize clusters.
In this video, we'll apply k-means clustering algorithm to n-dimensional wine data set in TensorFlow and visualize it in 3D.
In this video, we will apply different optimization algorithms which are Gradient Descent, Adaptive Learning, Momentum and Adam (Adaptive Momentum) in TensorFlow and monitor loss changes and converge speed in TensorBoard.
In this video, we will mention activation functions in deep neural networks. Also, we will focus on that what makes these function common. Finally, we will monitor loss change (mean squared error) for these functions in TensorBoard.
After all, softplus funtion would be winner for xor gate classification among sigmoid, tanh and relu.
Deep neural networks is modeled for detecting handwritten digits
In this video, we will install Keras. TensorFlow installation is expected before installing Keras.
We will build a keras deep neural networks classifier. Classifier still runs on TensorFlow in background.
This is Sefik.
I received my MSc in Computer Science from Galatasaray University in 2011.
I have been working as a software developer for a Finance IT company since 2010. Currently, I am member of AI team as a Data Scientist.
My current research interests are Machine Learning and Cryptography. I’ve published several research papers about these motivations. Also, I enjoy speaking to communities about these disciplines.
BTW, I am blogging and creating online courses related to my research interest.