Updated readme

mackelab · Sep 13, 2015 · 3e52d3a · 3e52d3a
1 parent 94d1b5a
commit 3e52d3a
Show file tree

Hide file tree

Showing 5 changed files with 26 additions and 99 deletions.
diff --git a/.DS_Store b/.DS_Store
diff --git a/README.md b/README.md
@@ -1,99 +1,25 @@
-## Import modules
-```python
-import funs.util as util
-import funs.engine as engine
-import matplotlib.pyplot as plt
-import numpy as np
-```
-## Initialize random number generator
-```python
-np.random.seed(123)
-```
-## Specify dataset & fitting parameters
-```python
-xdim 	  = 2		
-ydim      = 30		
-numTrials = 50		
-trialDur  = 200	# in ms
-binSize   = 20	# in ms
-maxEMiter = 100		
-dOffset   = 1	# controls firing rate
-```
-## Sample from the model (simulate a dataset)
-```python
-training_set  = util.dataset(
-	seed	  = np.random.randint(10000),
-	xdim 	  = xdim,
-	ydim 	  = ydim,
-	numTrials = numTrials,
-	trialDur  = trialDur,
-	binSize   = binSize,
-	dOffset   = dOffset,
-	fixTau 	  = True, 
-	fixedTau  = np.linspace(0.1,0.5,xdim),
-	drawSameX = True)
-```
-__Output:__
-```
-+------------- Simulated Dataset Options -------------+
-| Dimensionality of Latent State:                   2 |
-| Dimensionality of Observed State (# neurons):    30 |
-| Duration of trials (ms):                        200 |
-| Size of bins (ms):                               20 |
-| Number of Trials:                                50 |
-+-----------------------------------------------------+
-Sampling trial 50 ...
-Average firing rate per neuron in this dataset: 65.003 Hz.
-```
-## Initialize parameters using Poisson-PCA
-```python
-initParams = util.initializeParams(xdim, ydim, training_set)
-```
-## Fit using vanilla (batch) EM
-```python
-fitBatch = engine.PPGPFAfit(
-	experiment 		= training_set,
-	initParams 		= initParams,
-	inferenceMethod = 'laplace',
-	EMmode 			= 'Batch',
-	maxEMiter 		= maxEMiter)
-```
-__Output:__
-```
-+-------------------- Fit Options --------------------+
-| Dimensionality of Latent State:                   2 |
-| Dimensionality of Observed State (# neurons):    30 |
-| EM mode:                                      Batch |
-| Max EM iterations:                              100 |
-| Inference Method:                           laplace |
-+-----------------------------------------------------+
-Iteration: 100 of 100, nPLL: = -224.6181
-This dataset is a simulated dataset.
-Processing performance against ground truth parameters...
-```
+# Poisson-GPFA
+
+This package is written by:
+ * Hooram Nam, `hooramnam@openmailbox.org`
+ * Jakob Macke, `jakob.macke@caesar.de`
+
+This repository contains different methods for the Gaussian process model withPoisson observations. It has been developed and implemented with the goal of modelling spike-train recordings from neural populations, but some of the methods will be applicable more generally.
+
+In particular, the repository includes methods for
+
+ * Laplace approximation for state-inference
+ * Variational method for state-inference
+ * Expectation maximisation for parameter learning, using Laplace or Variational inference
+     * Full EM, where all available trials are processed in each iteration
+     * Variants of stochastic EM, where a subset of avilable trials are processed in each iteration
+
+## Usage
+
+To get started the the example script either by `python example.py` in bash or`run example.py` inside iPython. The software is developed within the Anaconda python 3 environment. 
+
+If you notice a but, watnt to request a feature, or have a question or feedback, please make use of the issue-tracking capabilities of the repository. We love to hear from people using our code -- please send an email to info@mackelab.org.
+
+The code in this repository is a work in progress. This work is published under the GNU General Public License. The code is provided "as is" and has not warranty whatsoever.
+
 
-## Fit using online EM
-```python
-fitOnline = engine.PPGPFAfit(
- 	experiment 		= training_set,
- 	initParams 		= initParams,
- 	EMmode 			= 'Online',
- 	maxEMiter 		= maxEMiter,
- 	inferenceMethod = 'laplace',
-	batchSize 		= 5)
-```
-__Output:__
-```
-+-------------------- Fit Options --------------------+
-| Dimensionality of Latent State:                   2 |
-| Dimensionality of Observed State (# neurons):    30 |
-| EM mode:                                     Online |
-| Max EM iterations:                              100 |
-| Inference Method:                           laplace |
-| Online Param Update Method:                  `diag` |
-| Batch size (trials):                              5 |
-+-----------------------------------------------------+
-Iteration: 100 of 100, nPLL: = -226.2654
-This dataset is a simulated dataset.
-Processing performance against ground truth parameters...
-```
diff --git a/data/.DS_Store b/data/.DS_Store
diff --git a/example.py b/example.py
@@ -48,9 +48,10 @@
  	inferenceMethod = 'laplace',
 	batchSize 		= 5)
 
+# Make plots
 training_set.plotTrajectory();plt.show()
 fitBatch.plotParamSeq()
 fitOnline.plotParamSeq();plt.show()
 
 fitBatch.plotTrajectory()
-fitOnline.plotTrajectory();plt.show()
+fitOnline.plotTrajectory();plt.show()
diff --git a/funs/.DS_Store b/funs/.DS_Store