ekloberdanz / neuralnet-experiments Goto Github PK
View Code? Open in Web Editor NEWFully connected feed forward and convolutional neural nets - hyper-parameters tuning
Fully connected feed forward and convolutional neural nets - hyper-parameters tuning
In this lab I experimented with Neural Network (NN)classifiers for handwritten digit optical recognition (a multi-class problem) using Keras with TensorFlow. I created 3 groups of fully connected feed-forward (FNN) and 1 class of Convolutional Neural Net (CNN). All NNs use the softmax function for the output layer, which converts numbers (logits) into probabilities that sum up to one. Additionally, all NNs use 20% of training data as the validation set, 1-of-c output encoding (keras.utils.to_categorical) and early stopping to decide when to stop training to prevent overfitting (keras.callbacks.EarlyStopping). Also, I created with 20 different models for every FNN (60 in total) and 10 different models for the CNN and recorded the best model for the FNNs and the CNN including the overall classification accuracy, class accuracy, and confusion matrix for both training and testing data. All 4 programs mse_relu.py (Task 1.a), cross_entropy_relu.py (Task 1.a), cross_entropy_tanh.py (Task 1. b), and cnn_relu.py (Task 2) must be run with python3 (it was made with and tested with Python 3.5.2 on ubuntu). Install dependencies: pip install -r requirements.txt usage:type into terminal: python3 mse_relu.py training.csv testing.csv python3 cross_entropy_relu.py training.csv testing.csv python3 cross_entropy_tanh.py training.csv testing.csv python3 cnn_relu.py training.csv testing.csv (Each of the first three programs trains 20 models and the last one trains 10 models - takes about a minute to run a program.) the csv files that contain the data are provided in the submission folder
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.