NEWS

This file contains a summary of important changes from the previous
release of PDP++.  See the ChangeLog file for a comprehensive, though
less comprehensible, list.

===========================================
		Release 3.1
===========================================

This is primarily a bug fix release -- see the ChangeLog for details
(there were lots of them).  A few significant bugs were fixed with
leabra++, including fixing a problem that resulted in the accumulation
of roundoff errors in the weight values, to the point that it could
affect training results on some especially sensitive models (esp those
using shared weights).  Also, all new types defined in a css script
(e.g., new classes, enums, etc) are now local to a given script scope,
instead of being globally shared among all scripts, as was the case
previously.  This prevents numerous unintentional conflicts between
scripts, and sharing of types across scripts is such a tenuous and bad
idea that it was never really used anyway.

===========================================
		Release 3.0
===========================================

IMPORTANT NOTE FOR EXISTING PROJECTS:

- If you have a stopping criterion set on a MonitorStat (e.g.,
  stopping when MAX activation exceeds some threshold in the output
  layer), then this stopping criterion will be lost when you load the
  project.  If you look in the spew of messages generated during the
  load process, you'll see a message like this:

   *** Member: mon not found in type: MonitorStat (this is likely just
   harmless version skew)
   <<err_skp ->>{name="max_act_Output": disp_opts=" ": is_string=false:
   vec_n=0: val=0: str_val="": stopcrit={flag=true: rel=GREATERTHAN:
   val=.5: cnt=1: n_met=0: }: };<<- err_skp>>

  Just get the stopcrit value (val=.5 in this example) and
  relationship (GREATERTHAN) from this and enter it into the first
  element in the mon_vals list in the appropriate MonitorStat in the
  opened project to restore function (or replace this stat entirely
  with a new ActThreshRTStat).

- GraphLog FIXED ranges will be lost upon loading: there is now a
  fix_min and fix_max flag inline with the ranges that will need to be
  clicked as appropriate (the range values are preserved, just the
  flag to fix them is missing).  Look for FIXED messages in the spew
  as per above.

- SelectEdit (.edits) labels will be lost upon loading, and can be
  recovered in the same manner as above.

- NetView Layer fonts will be smaller than before -- use
  View:Actions/Set Layer Font to set a larger font.  The new default
  for new views is 18 point.

- For your additional C++ code: El() function on Lists/Groups/Arrays
  renamed to SafeEl() to better reflect its function (index range
  checking), and [] operator changed to FastEl() (no index range
  checking) instead of SafeEl(), to better reflect typical usage.

- Error functions in BP: if you've written your own, see comments in
  ChangeLog (search for date 2002-09-13).

NEW FEATURES:

- Wizard that automates the construction of simulation objects:
  creates commonly-used configurations of neworks, environments,
  processes, stats and logs.

- Distributed-memory parallel processing via MPI (instead of pthread):
  DMEM can be much more efficient than pthread, and is much more
  flexible in that it works in both shared and distributed memory
  architectures.  Support for distributing computation across both
  connections and across events is provided, by setting the number of
  processors in the Network and the EpochProcess.  In both cases, each
  processor runs everything redundantly except for a subset of events
  or connections -- this makes for relatively little extra code
  required to support dmem -- connections/events are divided across
  processes, and results are synchronized to keep everything consistent.

- Additional analysis functions: PCA (principal components analysis)
  and MDS (multidimensional scaling), and Cluster Plot efficiency
  vastly improved to handle large data sets.  Also added processes for
  automatically computing these functions over hidden layers, etc.
  Other new analysis routines include automatic generation of
  statistics on environment frequencies -- useful for validating
  environments.

- Much improved GraphLog, focusing on discriminating overlapping line
  traces (repeated passes through the same set of X axis values --
  eg. multiple settles of a network, multiple training runs, etc).
  Traces can be color-coded (line_type = TRACE_COLORS), incremented
  (producing a 3D-like effect) via trace_incr, and stacked (vertical =
  STACK_TRACES).  A spike-raster-like plot, or even a continuous
  color-coded version, can be achieved by not displaying any vertical
  axis values at all (vertical = NO_VERTICAL) and using either
  VALUE_COLORS (continuous color-coded) or THRESH_POINTS (thresholded
  spike raster).  Also, columns of data can be plotted row-wise
  instead (e.g., for monitored activations, or COPY aggregators), and
  the view_bufsz value is now actually respected, so you can scroll
  through large amounts of data instead of seeing it all.

- Log displays are now much more robust when you add or remove data to
  be logged -- you should now spend much less time reconfiguring the
  log views.

- Color-coded edit dialogs and view window backdrops based on an enhanced
  (and customizable) Project view color scheme.

- Ability to save view displays to a JPEG or TIFF file (in addition to
  existing Postscript), including automatic saving at each update for
  constructing animations.

- Incremental reading of events from a file during processing (FromFileEnv).

- Automatic SimLog creation whenever project is saved -- helps you
  keep track of what each project is.

- Added two new algorithms: Long Short Term Memory (Hochreiter,
  Schmidhuber et al) implemented as lstm++ (works really well on
  sequential learning problems in general), and RNS++, written by Josh
  Brown: http://iac.wustl.edu/~jwbrown/rns++/index.html

- Support for g++ 3.x compilers (default for CYGWIN, DARWIN, use
  config/Makefile.LINUX.3 for LINUX instead of LINUX.2 (see
  README.linux for important details on compiling under LINUX,
  including a sstream include file fix for gcc 2.9x (e.g., RedHat
  7.x)), and zlib now used instead of forking to call gzip for
  loading/saving compressed files: should be much faster and more
  reliable under CYGWIN (MS Windows).  Check your Makefile.in for
  $(X11_LIB) instead of -lX11 if you get link errors involving zlib or
  jpeg lib calls.  Also, the SIM_NONSHARED makefile variables have
  been eliminated -- these were included by default in mk_new_pdp
  Makefile.in files, and need to just be cut out of the makefile.

- Numerous small processes and statistics to facilitate auotomation of
  common tasks.  Also, a better interface for interactive environments
  where subsequent events depend on current network outputs (see
  demo/leabra/nav.proj.gz for an example).

- Mersenne Twister random number generator now used for all random
  number calls (by Makoto Matsumoto and Takuji Nishimura,
  http://www.math.keio.ac.jp/~matumoto/emt.html)

- Easier access to view configuration variables through view menu
  functions; support for window manager close button; SelectEdit can
  include Methods (Functions) in addition to member data; Net log view
  usability considerably improved; Added buttons for graphing
  activation functions, etc on relevant specs.

===========================================
		Release 2.2
===========================================

- IMPORTANT BUGS FIXED: 

  o If you've tried to use the layer lesion flag, it has been broken
    since the parallel-threading code was introduced in 2.1!  It
    failed to accurately record the layer indexes to run when skipping
    over lesioned layers.

  o Big memory leak in gcc version 2.96 shipped with red hat linux
    7.1/2 fixed: causes it to not call the appropriate Destroy() function
    for the array template destructor, so FreeArray_() is never called.

  o DARWIN: had the same problem that the CYGWIN code had with random
    seeds -- NewInit returns nearly the same value each time.. Fixed
    in the same way as CYGWIN. doh.

  o rbp: the real_time flag was not working properly, causing a BP to
    occur *every trial* instead of at the end of the sequence..

  o Several other smaller bugs fixed too -- see ChangeLog for details.

- Configuration files for both the "Settings" menu (taMisc) and the
  pdp root object are now savable and automatically loaded from users
  home directory as $HOME/.taconfig and $HOME/.pdpconfig.  This makes
  it much more automatic and easy to set up your own personal
  configuration options.

- Added window position offsets (in Settings/taMisc) to correct for
  differences between window managers in window placement (I find that
  KDE requires X offset of -4 and Y of 24 to match the original fvwm
  and other wm locations).

- Default number of colors in a color colorscale
  (taMisc::color_scale_size, in the Settings menu) increased from 32
  to 128 -- assumes people are probably using 16 bit color these days
  If you are running on an 8 bit color display, you can set it back to
  32 and save the configuration!  Also fixed the way that colorscales
  are constructed.

- Non-default values are now shown with a bright yellow highlight.
  This is useful esp. for algorithms such as Leabra that have a large
  number of parameters that are usually at their default values.
  Default values are shown in the help message obtained by clicking on
  field name -- these messages are now much more informative by
  including help messages for all sub-fields and enums where applicable. 

- Windows now automatically scroll while manipulating objects (e.g.,
  resizing/moving layers, event specs, grid log headers, etc), so you
  no longer have to drag outside the window!  Also, weird displacement
  of mouse relative to manipulated object bug is now fixed.

- TimeUsed information automatically tracked on processes and pthread
  code: easy to get detailed timing data.

- Added a number of useful buttons in various of the displays and
  other minor improvements -- see ChangeLog for details.


===========================================
		Release 2.1
===========================================

- Mac OSX (Darwin) fully supported, using XFree86 X windows server.
  See README.darwin for details.

- SMP parallel processing (pthread-based) for several algos.  This
  works by compiling a list of layers for each thread to process,
  based on the number of connections and units in the layers (separate
  list for each).  The allocation algorithm tries to distribute the
  computation as evenly as possible.  This is relatively fine-grained
  parallelism, and due to the overhead involved in managing the
  threads, it only speeds up relatively large networks (i.e., > 50
  units per layer, >= 4 layers).  Speedups can by upwards of 1.67
  times as fast on dual pentium III systems.

  Activate by editing the network and setting the n_threads value.
  You can also restrict parallelism to only connection-level
  computations --- unit level speedup is quite small (but
  measurable..)

Many of the other major improvements have to do with important (but
perhaps small) improvements in the usability of the interface:

- Added a mechanism for consolidating a variety of different
  parameters from different objects in to one edit dialog.  This is
  ideal for playing with various parameters that tend to be scattered
  across various different Specs in pdp++.  This was done by adding
  .edits to Project -- this contains new SelectEdit objects that can
  be configured to edit selected fields from any other objects.  Every
  object now contains a SelectForEdit menu option in their Objects
  menu, which asks for which member (field) of the object to stick
  onto a SelectEdit object.

- Added a way to make it clearer when there are parameters that are
  conditional on others --- if parameters are not currently relevant
  for the current setting of a flag variable, then they are not 
  editable.

- Direct analysis of log data via new Analyze menu: you can get a
  cluster plot or distance matrix of monitored state values, or copy
  the data over to an environment for further analysis.  This may be
  easier than using the CopyToEnvStat for ad-hoc analyses.

- Duplicate, ChangeType now works for everything: layers, units, sched
  procs.  Go crazy!

- Confirmation now obtained for most destructive acts.

- New buttons for easy control of process stepping level: Step Up
  moves the step process up one level, and, critically, removes the
  previous step process as a network updater.  Step Dn moves down one
  level, and adds this new level to the network updater.  Therefore,
  if you want to single-step through cycling for a specific trial, but
  skip over previous trials, you can Step at the trial level until you
  find the right trial, then Step Dn to the cycle level and step
  there, then Step Up to trial again and zip through other trials,
  then step down to cycle, etc.  Previously, this would have required
  coordinated setting of step proc via a menu and the AddUpdater and
  RemoveUpdater functions on the network.  Also, even when you
  manually set the step proc, it now autotmatically ensures that this
  proc is updating the network.  Basically, you shouldn't find the
  need to manually do AddUpdater, etc on the network any more.

- Enviro View GUI enhancements:

  o Event groups are now displayed with their own buttons in the view,
  and they can thus be manipulated, events transfered into them, etc.
  Clicking twice will select the sub-events within the group.  Also,
  new buttons were added for duplicating, transfering, changing the
  types and setting the specs for events/groups.  This makes it much
  easier to perform all common manipulations of events in the gui.
  Also, the Disp toggle now affects display of event patterns, so you
  can turn off Disp to operate on large numbers of events without
  having to actually display them all.

  o If there is only one EventSpec, then when you Edit Spec(s), it is
  automatically selected.  Also, if only one event spec is selected
  for editing, then all clicking actions skip the event spec level,
  saving one level of clicking to select patterns, etc.  This should
  all be very intuitive.

  o New buttons: Layout/Updt allows you to either do a linear layout of
  all the patterns (stacked one atop the other with one space between
  -- useful if you've got a messy layout), or to update the positions
  and geometries of all the patterns to correspond with the current
  configuration of their corresponding layers.  This latter function
  should prove extremely convenient. New Child, Copy From, Duplicate
  buttons also available.

  o middle mouse button now always sets pattern value to 0 (previously
  was an undo)

- Project View GUI enhancements:

  o New buttons to create top-level groups for procs/specs (for
  organizing them), and for creating procs/specs so that these can be
  created directly from the interface.

  o Transfer, ChangeType enabled for specs, other minor fixes.

- As always, many bug fixes, including window iconification bug with
  KDE window manager environment.

===========================================
		Release 2.0
===========================================

ISSUES WITH OLD PROJECTS:

- A number of object members have been removed or renamed, so you
  should expect a large number of "xxx not found" warnings when
  loading old projects -- these can all be ignored.

- Older projects didn't set the other_idx indexes in con groups, which
  point to the corresponding sending/receiving con group.  These are
  now checked for and if not set, nothing can be run.  So, you need to
  either reconnect your network (ConnectAll) or run FixPrjnIndexes on
  the network (avail in the Network Actions menu).  A warning will
  prompt you to do this.

NEW FEATURES:

- First and foremost, all known reliable means of crashing the
  software have been fixed!  Please report any further reliable
  crashes you find so they can also be fixed.

- Help option added to every object -- pulls up netscape by default
  and points it at relevant section of pdp manual.

- New Project View -- provides overall view of the principal objects
  in the project (Networks, Environments, Processes, Stats, Logs), or
  all of the specs, so you can see what is there at one glance, and
  everything can be manipulated just like in the NetView.  This is
  most important for processes -- you can view/modify what the
  processes are connected to, how all the stats are arranged, what
  logs they send to, etc.  Should make configuring processes much
  easier.  The spec editing is very useful for more complex setups.

- Environment View totally rewritten -- allows interactive editing of
  event specs that control layout of the events, much in the way
  networks can be constructed in the NetView.  Display of event
  information is now *much* more efficient and crashproof.  Use
  SetToLayer in spec editing to automatically configure a pattern to
  fit a given network layer.

- GridLog totally rewritten -- used to crash when resized, etc.  Now
  can arrange display in the header using left mouse button to move,
  middle to resize, and right to edit.

- "Grid View Weights" option in Network/Actions menu -- displays
  network weights for an entire layer of units in a grid log view.

- "Dist Matrix Grid" and "Env To Grid" options in Environment.  First
  displays distance matrix between all items in a grid log, second
  sends all patterns to a grid log for easier, more configurable
  viewing.

- Cluster Plot in Environment now displays in a Graph Log (used to
  require xgraph).  Graph Log can now display text data using Y axis
  as immediately preceding data column.

- Many improvements to the basic View technology used in the NetView,
  EnviroView, and GridLogView:
  - Selecting and manipulating (resizing, moving) objects is much more
    intuitive.  Whatever is current selected is manipulated.  Things
    that cannot be manipulated are automatically not selected for
    manipulation.  To manipulate a subobject, just click and release
    -- the release will select the next level down (e.g., units within
    a unit group), and then the next click will manipulate that
    object.  Previous system required major finger gymnastics to get
    the right thing selected.
  - Editing of textual stuff can be accomplished by clicking right
    mouse button on the object, which will pull up an edit dialog.
  - Images no longer migrate across to different views.
  - Display is automatically Init-ialized after moving, resizing, etc.

- NetView/Network improvements:
  - much better projection arrow positioning algorithm.
  - confusion between View and Select mode minimized because only
    units can be selected for Viewing (viewing weights).
  - listing of network variables now has weights first, then
    unit-level stuff, so it is easier to switch between weights and
    activation, esp for algoirthms that have a lot of unit-level
    variables.
  - New Disp Mode menu on right for selecting how to display values.
  - Height-field displays now actually work as expected -- nice for
    displaying continuously-varying variables (e.g., Kohonen-style
    activation blobs -- see demo/so/som.proj.gz).
  - Text display is no longer done with separate objects -- much more
    efficient and prevents problems View-ing unit weights with text
    displayed.
  - Netview is automatically updated when network objects are removed
    outside of the netview.
  - Left mouse now moves in the X (horiz) and Z (vert) planes, while
    right button does X and Y (depth).  X-Z is much more common.
  - Moved CopyWeights, WriteWeights, ReadWeights to the Network Object
    menu instead of the Actions menu -- all basic I/O should be in the
    Object menu.
  - Imported graphic displays (made by idraw) now work (can be
    manipulated, saved, etc w/out crashing)

- Read/WriteOldPDP on Environment changed to Read/WriteText, and
  rewritten so they actually work. # endgroup comment is now
  superfluous.

- New Object menu options for all objects: 
  - Duplicate will duplicate object (used to only be on groups)
  - CopyTo and CopyFrom used for copying, works even with subclasses
  and super-classes of given object
  - ChangeType will replace current object with one of a different
  (related) type, copying exisiting info.

- With too many items to fit in menus, used to get "Over Max-menu" and
  that was it, but now you can select an object from a file-chooser
  like object-chooser dialog that shows all the objects.  

- Object-chooser also available as a Browse function on root object --
  you can traverse the entire object heirarchy and edit anything.

- No more view:: submenu in lots of menus -- was never used anyway.

- Edit dialogs all have a Show menu that enables controlling of what
  level of detail is viewed.  A Detail level of viewing is now
  present, and these details are not visible by default, making it
  easier to find the high-priority stuff.  If you want to access
  something that is not visible, use the Show menu.

- EpochProcesses much more robust to changes in environments during an
  epoch -- no more crashing!