By @dnl0x00 Recently, Google announced the eager execution for TensorFlow. Eager execution has some advantages when doing quick prototyping. The code is easier to debug because operations are executed immediately and you can build models via Python control flow (including if statements and for and while loops). You don’t have... [Read More]

By @dnl0x00 Non-keyworded, variable length argument list. Here’s an example which shows how to pass variable length arguments and get the values. 1 2 3 4 5 6 7 8 def g(a, *args): b, c = args assert a == 1 def f(a, *args): g(a, *args) f(1, 2, 3) for... [Read More]

By @dnl0x00 In October this year researchers from Baidu and NVIDIA have published a paper in which they describe a technique to train deep neural networks using half precision floating point numbers (FP16) to decrease the memory consumption of the models and to benefit from speed up effects due to... [Read More]

By @dnl0x00 The MNIST handwritten digit database is a very popular data set for testing machine learning algorithms. It contains 60,000 labeled training examples and 10,000 examples for testing. The data set can be downloaded from here. On GitHub I have published a repository which contains a file mnist.mat created... [Read More]

By @dnl0x00 In this article I just summarize some thoughts while reading an article about machine learning doing on a CPU vs GPU. All of my statements are not well researched and I could be totally wrong. Keep that in mind when you’re reading this post. ;) Recently I have... [Read More]