Loading ...
Sorry, an error occurred while loading the content.
 

ANJI and starting a net with disconnected inputs...

Expand Messages
  • bmiller3230
    Does anyone know if its possible to start a net in ANJI with some of the inputs disconnected per FS-NEAT and
    Message 1 of 5 , May 5, 2005
      Does anyone know if its possible to start a net in ANJI with some of
      the inputs disconnected per FS-NEAT and
      http://www.cs.utexas.edu/users/kstanley/neat.html#disconnected ?

      Thanks.
    • Philip Tucker
      Yes, it is possible. In fact we ve done XOR runs before where the starting topology is completely disconnected (i.e., 0 connections) and it is able to find a
      Message 2 of 5 , May 6, 2005
        Yes, it is possible. In fact we've done XOR runs before where the
        starting topology is completely disconnected (i.e., 0 connections) and
        it is able to find a solution.

        - Philip

        > Does anyone know if its possible to start a net in ANJI with some of
        > the inputs disconnected per FS-NEAT and
        > http://www.cs.utexas.edu/users/kstanley/neat.html#disconnected ?
        >
        > Thanks.
      • bmiller3230
        ... and ... of ... Thanks for your reply, Phillip. As a practical matter, is setting all connections to zero (including input connections) somthing that I set
        Message 3 of 5 , May 9, 2005
          --- In neat@yahoogroups.com, Philip Tucker <ptucker@g...> wrote:
          > Yes, it is possible. In fact we've done XOR runs before where the
          > starting topology is completely disconnected (i.e., 0 connections)
          and
          > it is able to find a solution.
          >
          > - Philip
          >
          > > Does anyone know if its possible to start a net in ANJI with some
          of
          > > the inputs disconnected per FS-NEAT and
          > > http://www.cs.utexas.edu/users/kstanley/neat.html#disconnected ?
          > >
          > > Thanks.

          Thanks for your reply, Phillip.

          As a practical matter, is setting all connections to zero (including
          input connections) somthing that I set through the properties file or
          is it something I need to code? Thanks again.
        • Derek James
          Hi, This is the other half of the ANJI team. I believe if you just set the following two properties the following ways: initial.topology.fully.connected=false
          Message 4 of 5 , May 9, 2005
            Hi,

            This is the other half of the ANJI team.

            I believe if you just set the following two properties the following ways:

            initial.topology.fully.connected=false
            initial.topology.num.hidden.neurons=0

            That should start with an intial topology with only input and output
            neurons, with no connections. Philip can correct me if that isn't
            right, but go ahead and give it a try. :)

            Derek

            On 5/9/05, bmiller3230 <bmiller3230@...> wrote:
            > --- In neat@yahoogroups.com, Philip Tucker <ptucker@g...> wrote:
            > > Yes, it is possible. In fact we've done XOR runs before where the
            > > starting topology is completely disconnected (i.e., 0 connections)
            > and
            > > it is able to find a solution.
            > >
            > > - Philip
            > >
            > > > Does anyone know if its possible to start a net in ANJI with some
            > of
            > > > the inputs disconnected per FS-NEAT and
            > > > http://www.cs.utexas.edu/users/kstanley/neat.html#disconnected ?
            > > >
            > > > Thanks.
            >
            > Thanks for your reply, Phillip.
            >
            > As a practical matter, is setting all connections to zero (including
            > input connections) somthing that I set through the properties file or
            > is it something I need to code? Thanks again.
            >
            >
            >
            > Yahoo! Groups Links
            >
            >
            >
            >
            >
          • bmiller3230
            ... Thanks, Derek. I ll give that a try and let you know how it works out. Brett
            Message 5 of 5 , May 10, 2005
              --- In neat@yahoogroups.com, Derek James <djames@g...> wrote:
              > Hi,
              >
              > This is the other half of the ANJI team.
              >
              > I believe if you just set the following two properties the
              following ways:
              >
              > initial.topology.fully.connected=false
              > initial.topology.num.hidden.neurons=0
              >
              > That should start with an intial topology with only input and output
              > neurons, with no connections. Philip can correct me if that isn't
              > right, but go ahead and give it a try. :)
              >
              > Derek
              >


              Thanks, Derek. I'll give that a try and let you know how it works
              out.

              Brett
            Your message has been successfully submitted and would be delivered to recipients shortly.