<?xml version="1.0"?><?xml-stylesheet type="text/xsl" href="/rss.xsl"?><rss version="2.0"><channel><title>autodiff Issue Tracker Rss Feed</title><link>http://autodiff.codeplex.com/workitem/list/basic</link><description>autodiff Issue Tracker Rss Description</description><item><title>Commented Feature: Try to accelerate computation using IL generation [1574]</title><link>http://autodiff.codeplex.com/workitem/1574</link><description>Use Reflection.Emit API to construct a specific set of instructions for a given function such that future differentiations will be faster than with current Visitor based implementations&lt;br /&gt;Comments: Hm, interesting, and a bit surprising. I&amp;#39;ve worked quite a bit on dynamically generating code lately, if you want I could have a look at this as well, maybe you could send me the code of your experiment&amp;#63;</description><author>davidacoder</author><pubDate>Wed, 26 Sep 2012 18:42:55 GMT</pubDate><guid isPermaLink="false">Commented Feature: Try to accelerate computation using IL generation [1574] 20120926064255P</guid></item><item><title>Closed Feature: Try to accelerate computation using IL generation [1574]</title><link>http://autodiff.codeplex.com/workitem/1574</link><description>Use Reflection.Emit API to construct a specific set of instructions for a given function such that future differentiations will be faster than with current Visitor based implementations&lt;br /&gt;Comments: Closed because manually generating IL does not accelerate computation.</description><author>alexshtf</author><pubDate>Wed, 26 Sep 2012 14:34:40 GMT</pubDate><guid isPermaLink="false">Closed Feature: Try to accelerate computation using IL generation [1574] 20120926023440P</guid></item><item><title>Commented Feature: Try to accelerate computation using IL generation [1574]</title><link>http://autodiff.codeplex.com/workitem/1574</link><description>Use Reflection.Emit API to construct a specific set of instructions for a given function such that future differentiations will be faster than with current Visitor based implementations&lt;br /&gt;Comments: An experiment shows that custom IL generation actually works SLOWER both when using DynamicMethod or Expression.Compile. &amp;#10;&amp;#10;Unless proven otherwise, it seems that manually generating IL does not accelerate the gradient computation but slows it down.</description><author>alexshtf</author><pubDate>Wed, 26 Sep 2012 14:34:15 GMT</pubDate><guid isPermaLink="false">Commented Feature: Try to accelerate computation using IL generation [1574] 20120926023415P</guid></item><item><title>Created Feature: Try to accelerate computation using IL generation [1574]</title><link>http://autodiff.codeplex.com/workitem/1574</link><description>Use Reflection.Emit API to construct a specific set of instructions for a given function such that future differentiations will be faster than with current Visitor based implementations&lt;br /&gt;</description><author>alexshtf</author><pubDate>Wed, 26 Sep 2012 14:33:08 GMT</pubDate><guid isPermaLink="false">Created Feature: Try to accelerate computation using IL generation [1574] 20120926023308P</guid></item><item><title>Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366]</title><link>http://autodiff.codeplex.com/workitem/1366</link><description>The current Differentiate&amp;#47;Evaluate APIs can handle an array of doubles. However, some external libraries use different data structures that are not simple arrays but do implement the IList&amp;#60;double&amp;#62; interface.&lt;br /&gt;Comments: Done for IList&amp;#60;T&amp;#62; objects. IEnumerable will shortly follow.</description><author>alexshtf</author><pubDate>Sun, 22 Jul 2012 22:08:22 GMT</pubDate><guid isPermaLink="false">Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366] 20120722100822P</guid></item><item><title>Commented Issue: Hessian [1264]</title><link>http://autodiff.codeplex.com/workitem/1264</link><description>Not sure whether I will really need this at some point, but was just wondering whether this is something that might be added in the future to the library&amp;#63; I am mainly opening this issue so that people can vote on this.&lt;br /&gt;Comments: ** Comment from web user: davidacoder ** &lt;p&gt;Sounds good&amp;#33;&lt;/p&gt;</description><author>davidacoder</author><pubDate>Tue, 29 May 2012 16:56:52 GMT</pubDate><guid isPermaLink="false">Commented Issue: Hessian [1264] 20120529045652P</guid></item><item><title>Commented Issue: Hessian [1264]</title><link>http://autodiff.codeplex.com/workitem/1264</link><description>Not sure whether I will really need this at some point, but was just wondering whether this is something that might be added in the future to the library&amp;#63; I am mainly opening this issue so that people can vote on this.&lt;br /&gt;Comments: ** Comment from web user: alexshtf ** &lt;p&gt;Scheduled for the next release in the form of &amp;#34;Hessian-Vector computation&amp;#34;. The library will be able, given a vector v, to compute H&amp;#42;v where H is the hessian.&lt;/p&gt;&lt;p&gt;This design will allow the library to use the sparsity of the function, without directly computing the Hessian matrix itself. It will also allow me to add a utility method to extract the Hessian directly by multiplying by the standard basis vectors &amp;#40;setting v &amp;#61; &amp;#40;1, 0, ... 0&amp;#41;, then v &amp;#61; &amp;#40;0, 1, 0, ..., 0&amp;#41; and so on&amp;#41;.&lt;/p&gt;</description><author>alexshtf</author><pubDate>Tue, 29 May 2012 06:56:49 GMT</pubDate><guid isPermaLink="false">Commented Issue: Hessian [1264] 20120529065649A</guid></item><item><title>Closed Issue: Performance improvement [1573]</title><link>http://autodiff.codeplex.com/workitem/1573</link><description>Changed the structure of the evaluation&amp;#47;differentiation tape.&lt;br /&gt;Comments: &lt;p&gt;Fixed&lt;/p&gt;</description><author>alexshtf</author><pubDate>Mon, 07 May 2012 11:55:10 GMT</pubDate><guid isPermaLink="false">Closed Issue: Performance improvement [1573] 20120507115510A</guid></item><item><title>Commented Feature: Experiment with parallel computation of value + gradient [683]</title><link>http://autodiff.codeplex.com/workitem/683</link><description>Try using Parallel extensions for .NET to allow developers to compute gradients and function values faster on multi-code machines.&lt;br /&gt;Comments: ** Comment from web user: alexshtf ** &lt;p&gt;The dependency analysis doesn&amp;#39;t seem simple. If you know of a good algorithm to divide the computation graph in advance to parallel computation sequences &amp;#40;during the compilation stage&amp;#41; it might work.&lt;br /&gt;Right now I have no such good idea. Need to do more reading.&lt;/p&gt;</description><author>alexshtf</author><pubDate>Mon, 07 May 2012 11:54:47 GMT</pubDate><guid isPermaLink="false">Commented Feature: Experiment with parallel computation of value + gradient [683] 20120507115447A</guid></item><item><title>Commented Issue: Performance improvement [1573]</title><link>http://autodiff.codeplex.com/workitem/1573</link><description>&lt;br /&gt;Comments: Associated with changeset 19334.</description><author>alexshtf</author><pubDate>Thu, 05 Apr 2012 00:16:01 GMT</pubDate><guid isPermaLink="false">Commented Issue: Performance improvement [1573] 20120405121601A</guid></item><item><title>Reopened Issue: Performance improvement [1573]</title><link>http://autodiff.codeplex.com/workitem/1573</link><description>&lt;br /&gt;</description><author>alexshtf</author><pubDate>Wed, 04 Apr 2012 22:03:07 GMT</pubDate><guid isPermaLink="false">Reopened Issue: Performance improvement [1573] 20120404100307P</guid></item><item><title>Closed Issue: Performance improvement [1573]</title><link>http://autodiff.codeplex.com/workitem/1573</link><description>&lt;br /&gt;Comments: Resolved with changeset 19330.</description><author>alexshtf</author><pubDate>Wed, 04 Apr 2012 22:02:48 GMT</pubDate><guid isPermaLink="false">Closed Issue: Performance improvement [1573] 20120404100248P</guid></item><item><title>Created Issue: Performance improvement [1573]</title><link>http://autodiff.codeplex.com/workitem/1573</link><description>&lt;br /&gt;</description><author>alexshtf</author><pubDate>Wed, 04 Apr 2012 22:01:30 GMT</pubDate><guid isPermaLink="false">Created Issue: Performance improvement [1573] 20120404100130P</guid></item><item><title>Closed Feature: Expression-system based evaluation and differentiation [756]</title><link>http://autodiff.codeplex.com/workitem/756</link><description>Allow evaluating function using the .NET 4 expression system. This allows us to compile the expression into code for really fast evaluation.&lt;br /&gt;Comments: Resolved with changeset 19328.</description><author>alexshtf</author><pubDate>Wed, 04 Apr 2012 21:41:29 GMT</pubDate><guid isPermaLink="false">Closed Feature: Expression-system based evaluation and differentiation [756] 20120404094129P</guid></item><item><title>Commented Feature: Expression-system based evaluation and differentiation [756]</title><link>http://autodiff.codeplex.com/workitem/756</link><description>Allow evaluating function using the .NET 4 expression system. This allows us to compile the expression into code for really fast evaluation.&lt;br /&gt;Comments: Associated with changeset 16387.</description><author>alexshtf</author><pubDate>Tue, 17 Jan 2012 14:54:00 GMT</pubDate><guid isPermaLink="false">Commented Feature: Expression-system based evaluation and differentiation [756] 20120117025400P</guid></item><item><title>Commented Feature: Expression-system based evaluation and differentiation [756]</title><link>http://autodiff.codeplex.com/workitem/756</link><description>Allow evaluating function using the .NET 4 expression system. This allows us to compile the expression into code for really fast evaluation.&lt;br /&gt;Comments: Associated with changeset 16378.</description><author>alexshtf</author><pubDate>Tue, 17 Jan 2012 12:07:26 GMT</pubDate><guid isPermaLink="false">Commented Feature: Expression-system based evaluation and differentiation [756] 20120117120726P</guid></item><item><title>Commented Feature: Expression-system based evaluation and differentiation [756]</title><link>http://autodiff.codeplex.com/workitem/756</link><description>Allow evaluating function using the .NET 4 expression system. This allows us to compile the expression into code for really fast evaluation.&lt;br /&gt;Comments: Associated with changeset 16377.</description><author>alexshtf</author><pubDate>Tue, 17 Jan 2012 11:00:18 GMT</pubDate><guid isPermaLink="false">Commented Feature: Expression-system based evaluation and differentiation [756] 20120117110018A</guid></item><item><title>Closed Feature: Expression-system based differentiation [757]</title><link>http://autodiff.codeplex.com/workitem/757</link><description>Use the .NET 4 expression system to compile differentiation into code for very fast differentiation.&lt;br /&gt;</description><author>alexshtf</author><pubDate>Tue, 17 Jan 2012 10:49:16 GMT</pubDate><guid isPermaLink="false">Closed Feature: Expression-system based differentiation [757] 20120117104916A</guid></item><item><title>Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366]</title><link>http://autodiff.codeplex.com/workitem/1366</link><description>The current Differentiate&amp;#47;Evaluate APIs can handle an array of doubles. However, some external libraries use different data structures that are not simple arrays but do implement the IList&amp;#60;double&amp;#62; interface.&lt;br /&gt;Comments: ** Comment from web user: alexshtf ** &lt;p&gt;Accepted&lt;/p&gt;</description><author>alexshtf</author><pubDate>Sun, 27 Nov 2011 19:00:18 GMT</pubDate><guid isPermaLink="false">Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366] 20111127070018P</guid></item><item><title>Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366]</title><link>http://autodiff.codeplex.com/workitem/1366</link><description>The current Differentiate&amp;#47;Evaluate APIs can handle an array of doubles. However, some external libraries use different data structures that are not simple arrays but do implement the IList&amp;#60;double&amp;#62; interface.&lt;br /&gt;Comments: ** Comment from web user: alexshtf ** &lt;p&gt;IEnumerable&amp;#60;double&amp;#62; is nice, but I &lt;/p&gt;</description><author>alexshtf</author><pubDate>Sun, 27 Nov 2011 18:59:22 GMT</pubDate><guid isPermaLink="false">Commented Feature: Allow the evaluation/differentiation API to accept any IList&lt;double&gt; and IEnumerable&lt;double&gt; instead of an array [1366] 20111127065922P</guid></item></channel></rss>