clean-up of NEWS_DEV

NEWS_DEV.txt

 .. _NEWS:
-up to #4129
 
 ===================
 DRAFT Release Notes
@@ -19,23 +18,20 @@ NEWS.txt:
 We recommend that everybody update to this version.
 
 Highlights:
-- Multi-GPU for data parallism via Platoon (https://github.com/mila-udem/platoon/)
-- cnmem
-- optimizer=fast_compile move computation to the GPU.
-- BreakpointOp
-- Faster optimization
-- Better lock
-- d3viz
-- CuDNN
-- Many Scan update (execution speed up, ...)
 - Python 2 and 3 support with the same code base
-- COp
-- __props__
-- Better convolution on CPU and GPU. (CorrMM, cudnn, 3d conv, more parameter)
+- Faster optimization
 - New GPU back-end
    - Float16 new back-end (need cuda 7.5)
    - Multi dtypes
    - Multi-GPU support in the same process
+- Integration of CuDNN for better GPU performance
+- Many Scan improvements (execution speed up, ...)
+- optimizer=fast_compile moves computation to the GPU.
+- Better convolution on CPU and GPU. (CorrMM, cudnn, 3d conv, more parameter)
+- Interactive visualization of graphs with d3viz
+- cnmem (better memory management on GPU)
+- BreakpointOp
+- Multi-GPU for data parallism via Platoon (https://github.com/mila-udem/platoon/)
 
 
 A total of 135 people contributed to this release, see the list at the bottom.
@@ -43,20 +39,20 @@ A total of 135 people contributed to this release, see the list at the bottom.
 
 Installation:
 - Better blas detection
-- Fix to more recent software and OS.
+- Fixes for more recent software and OS versions
 - Support Anaconda on Windows
 
 Bug fixes:
-- GpuJoin support neg axis
+- GpuJoin now supports negative axis
 - Fix GpuCumsum for negative axis
 
 Interface Deprecation (a warning is printed):
-- Deprecate Param class.
+- Deprecate Param class, use In instead
 
 Interface Changes:
 - Rename DownsampleFactorMax to Pool.
-- Stack now use numpy interface
-- optimizer=fast_compile move to the GPU
+- tensor.stack now uses the same interface as numpy.stack
+- optimizer=fast_compile moves computation to the GPU
 - Raise the user stack trace more frequently.
 - Change dev version numbering to follow the PEP 404
 
@@ -68,9 +64,11 @@ New Interface (reuses existing functionality):
 - MaxAndArgmax support axis=None
 - Add theano.tensor.compress (equivalent of numpy.compress)
 - theano.tensor.signal.downsamples.max_pool_2d_same_size
+- COp
+- __props__
 
 New features
-- Unique (Iban Harlouchet)
+- tensor.unique
 - map_variables
 - erfcx
 - mgrid, ogrid
@@ -78,18 +76,17 @@ New features
 - BreakpointOp
 - Make bincount work on GPU
 - SolveOp on GPU
-- Optinal optimization remove_all_assert
+- Optional optimization remove_all_assert
 - AllocEmpty
-- LogSoftmax, for stability optimization when the crossentropy optimization don't apply.
-- theano.tensor.repeat work on GPU
+- LogSoftmax, for stability optimization when the crossentropy optimization does not apply.
+- theano.tensor.repeat works on GPU
 - BatchedDot on the GPU and faster on the CPU.
 - Faster batched_tensordot and make it work on GPU.
 - SoftmaxGrad grad
 - 3d conv via CorrMM on the GPU
 - CPU Max Pool support of padding and strides!=windows size
-- Make stability optimization was disabled in a corner cases
-- theano.function() now accept a dict for the outputs. When doing this, the function will return a dict. Helpful to keep trac of which output is what.
-- Warn Theano flags defined by user, but don't exist.
+- theano.function() now accepts a dict for the outputs. When doing this, the function will return a dict. Helpful to keep track of which output is what.
+- Warn for unknown or misspelled theano config variables
 - theano.tensor.tile update (accept symbolic reps, work on GPU)
 - scan how have a strict flag. If set to True, this make scan building faster and could make execution faster.
 - theano.tensor.signal.conv2d(2d,2d) output 2d answer
@@ -122,8 +119,9 @@ Others:
 - function_dump to help us reproduce user error during compilation
 - assert_no_cpu_op
 - pep8, flake8
-- Better error message
-- On not default mode, redure the number of allocation  when allow_gc=False
+- Better error messages
+- On non-default modes, reduce the number of allocation when allow_gc=False
+- Better lock
 
 
 Committers for this dev version only:
@@ -259,7 +257,3 @@ Committers for this dev version only:
 - jotterbach
 - rolf
 - wuaalb
-
-
-Todo for the final release:
- * update the NEWS.txt file.