diff --git a/readme.md b/readme.md index 883bd05adece41c666fc576b025c62edd6799deb..f3f5de88d62c70df4c05b4f706d2610278eb5dd3 100644 --- a/readme.md +++ b/readme.md @@ -1,17 +1,31 @@ +# Python - Intro to Tensorflow -# CS5300 a2 +## Description +Neural Network Tensorflow examination. Used the [CIFAR-10](http://www.cs.toronto.edu/~kriz/cifar.html) image dataset to train a network. The end goal is to have a +network that can look at a given image and accurately give it a label (ex: "bird", "dog", "train", "truck", etc). -## Author -Brandon Rodriguez +Project actually used two different neural network types. +The first one was a "deep layer" network, which is essentially what the previous assignments were. This network examined +pixels one by one, and made decisions from that. + +The second network was a "convolutional" network, which is designed specifically for examining groups of pixels +together, as a single unit. Both grouping and meaning of group is determined by network, in the learning stages. + +Both networks had high accuracy, but the Convolutional network was noticably higher across the board. -## Description -Neural Network Tensorflow examination. +## TensorFlow +TensorFlow is a Neural Net library for Python. It essentially acts as an easy interface directly to C++ Neural Network +code. For the most part, this allows for the ease of syntax that Python brings, but the speed and efficiency of C++ code. +Due to the complexity (both logically and computationally) of Neural Nets, it's rare to implement one's own Neural Net +logic over using a prebuilt library. + + +## References Tutorials from: * https://www.tensorflow.org/versions/r1.0/get_started/mnist/beginners * https://www.tensorflow.org/versions/r1.0/get_started/mnist/pros * https://www.tensorflow.org/tutorials/deep_cnn -