require 'nn'

local nfeats = 500
local nstates = {}
local filtsize = 10
local poolsize = 10

-- a typical modern convolution network (conv+relu+pool)
model = nn.Sequential()

-- stage 1 : filter bank -> squashing -> L2 pooling -> normalization
model:add(nn.SpatialConvolutionMM(nfeats, nstates[1], filtsize, filtsize))
model:add(nn.ReLU())
model:add(nn.SpatialMaxPooling(poolsize,poolsize,poolsize,poolsize))

-- stage 2 : filter bank -> squashing -> L2 pooling -> normalization
model:add(nn.SpatialConvolutionMM(nstates[1], nstates[2], filtsize, filtsize))
model:add(nn.ReLU())
model:add(nn.SpatialMaxPooling(poolsize,poolsize,poolsize,poolsize))

-- stage 3 : standard 2-layer neural network
model:add(nn.View(nstates[2]*filtsize*filtsize))
model:add(nn.Dropout(0.5))
model:add(nn.Linear(nstates[2]*filtsize*filtsize, nstates[3]))
model:add(nn.ReLU())
model:add(nn.Linear(nstates[3], noutputs))