Project Plan

ToDoList

1. Using GIT, Using Cassandra

• Getting started with GIT [DONE]

• Config Cassandra [DONE]

• Config & use Hector [DONE]

2. Reading ML algorithms for time series -> find an appropriate approach for it

• ANN -> Backpropagation & derivatives [DONE]

3. Read about simple backpropagation (graph & matrix form) [DONE]

2. Find a good framework for ANN [ex: Encog, Joone, Neuroph]

• Implement a basic backpropagation algorithm (for a better insight) [DONE]

• Getting started with Encog -> backropagation, rprop [DONE]

2. Finding and testing data-sets

• ECG data-sets (http://physionet.org/physiobank/database/)

• Arrhythmia heart-beat detection - data-set including ~48 patients (30' of data per each, 2ms measurements)

• Analyzing time & network-precision - 3 layers [DONE]

3. Merging data-sets to build a better training/testing set

1. 24 full-data patients for learning -> 24 full-data patients for testing [DONE - see Dropbox/datasets]

2. 48 half-data patients for learning -> 48 half-data patients for testing [CANCELED]

3. Analyzing time (forward feed & backprop time) + network precision on each 1. and 2. [DONE - see Dropbox/datasets/test_results]

4. Increase the number of layers and train/test the network on the big dataset [DONE]