# a quick little example using psuedo likelihood to
# estimate the parameters of an auto-logistic model...

x_matrix(scan("bintex.dat"),ncol=128)
par(mfrow=c(2,2),pty="s")
image(x)

round(binmple12(x),4)
 (Intercept)     x1     x2     x3      x4      x5     x6 
      0.1141 0.2977 0.0143 0.1776 -0.0911 -0.0291 0.0325

#source("bintexture.s")


# function that determines the maximum psuedo-likelihood
# values for a given matrix of binary values.  We'll assume 
# we're using the 12 nearest neighbors (3rd order).  This 
# means 7 coefficients, including the intercept. 

binmple12_function(x){
 n_dim(x)[1]; m_dim(x)[2]
 
  # now compute displaced matricies

  x0_x[c(-1,-2,-(n-1),-n),c(-1,-2,-(m-1),-m)]

  xr_x[c(-1,-2,-(n-1),-n),c(-1,-2,-3,-m)]
  xl_x[c(-1,-2,-(n-1),-n),c(-1,-(m-2),-(m-1),-m)]

  xu_x[c(-1,-(n-2),-(n-1),-n),c(-1,-2,-(m-1),-m)]
  xd_x[c(-1,-2,-3,-n),c(-1,-2,-(m-1),-m)]

  xud_x[c(-1,-(n-2),-(n-1),-n),c(-1,-2,-3,-m)]
  xdu_x[c(-1,-2,-3,-n),c(-1,-(m-2),-(m-1),-m)]

  xuu_x[c(-1,-(n-2),-(n-1),-n),c(-1,-(m-2),-(m-1),-m)]
  xdd_x[c(-1,-2,-3,-n),c(-1,-2,-3,-m)]

  xr2_x[c(-1,-2,-(n-1),-n),c(-1,-2,-3,-4)]
  xl2_x[c(-1,-2,-(n-1),-n),c(-(m-3),-(m-2),-(m-1),-m)]

  xu2_x[c(-(n-3),-(n-2),-(n-1),-n),c(-1,-2,-(m-1),-m)]
  xd2_x[c(-1,-2,-3,-4),c(-1,-2,-(m-1),-m)]


  x1_as.vector(xr+xl)
  x2_as.vector(xu+xd)
  x3_as.vector(xdu+xud)
  x4_as.vector(xdd+xuu)
  x5_as.vector(xr2+xl2)
  x6_as.vector(xu2+xd2)
  x0_as.vector(x0)

  pldf_data.frame(x1,x2,x3,x4,x5,x6,x0)

  a_glm(x0~x1+x2+x3+x4+x5+x6,data=pldf)

  return(a$coef)
}