aboutsummaryrefslogtreecommitdiff
path: root/nerv/layer/gate_fff.lua
diff options
context:
space:
mode:
Diffstat (limited to 'nerv/layer/gate_fff.lua')
-rw-r--r--nerv/layer/gate_fff.lua56
1 files changed, 29 insertions, 27 deletions
diff --git a/nerv/layer/gate_fff.lua b/nerv/layer/gate_fff.lua
index 751dde1..6082e27 100644
--- a/nerv/layer/gate_fff.lua
+++ b/nerv/layer/gate_fff.lua
@@ -1,36 +1,33 @@
-local GateFFFLayer = nerv.class('nerv.GateFFFLayer', 'nerv.Layer')
+local GateFFFLayer = nerv.class('nerv.GateFLayer', 'nerv.Layer') --Full matrix gate
function GateFFFLayer:__init(id, global_conf, layer_conf)
self.id = id
self.dim_in = layer_conf.dim_in
self.dim_out = layer_conf.dim_out
self.gconf = global_conf
-
- self.ltp1 = self:find_param("ltp1", layer_conf, global_conf, nerv.LinearTransParam, {self.dim_in[1], self.dim_out[1]}) --layer_conf.ltp
- self.ltp2 = self:find_param("ltp2", layer_conf, global_conf, nerv.LinearTransParam, {self.dim_in[2], self.dim_out[1]}) --layer_conf.ltp
- self.ltp3 = self:find_param("ltp3", layer_conf, global_conf, nerv.LinearTransParam, {self.dim_in[3], self.dim_out[1]}) --layer_conf.ltp
+
+ for i = 1, #self.dim_in do
+ self["ltp" .. i] = self:find_param("ltp" .. i, layer_conf, global_conf, nerv.LinearTransParam, {self.dim_in[i], self.dim_out[1]}) --layer_conf.ltp
+ end
self.bp = self:find_param("bp", layer_conf, global_conf, nerv.BiasParam, {1, self.dim_out[1]})--layer_conf.bp
- self:check_dim_len(3, 1) -- exactly one input and one output
+ self:check_dim_len(-1, 1) --accept multiple inputs
end
function GateFFFLayer:init(batch_size)
- if self.ltp1.trans:ncol() ~= self.bp.trans:ncol() or
- self.ltp2.trans:ncol() ~= self.bp.trans:ncol() or
- self.ltp3.trans:ncol() ~= self.bp.trans:ncol() then
- nerv.error("mismatching dimensions of linear transform and bias paramter")
- end
- if self.dim_in[1] ~= self.ltp1.trans:nrow() or
- self.dim_in[2] ~= self.ltp2.trans:nrow() or
- self.dim_in[3] ~= self.ltp3.trans:nrow() then
- nerv.error("mismatching dimensions of linear transform parameter and input")
+ for i = 1, #self.dim_in do
+ if self["ltp" .. i].trans:ncol() ~= self.bp.trans:ncol() then
+ nerv.error("mismatching dimensions of linear transform and bias paramter")
+ end
+ if self.dim_in[i] ~= self["ltp" .. i].trans:nrow() then
+ nerv.error("mismatching dimensions of linear transform parameter and input")
+ end
+ self["ltp"..i]:train_init()
end
+
if self.dim_out[1] ~= self.ltp1.trans:ncol() then
nerv.error("mismatching dimensions of linear transform parameter and output")
end
- self.ltp1:train_init()
- self.ltp2:train_init()
- self.ltp3:train_init()
self.bp:train_init()
self.err_bakm = self.gconf.cumat_type(batch_size, self.dim_out[1])
end
@@ -44,8 +41,9 @@ end
function GateFFFLayer:propagate(input, output)
-- apply linear transform
output[1]:mul(input[1], self.ltp1.trans, 1.0, 0.0, 'N', 'N')
- output[1]:mul(input[2], self.ltp2.trans, 1.0, 1.0, 'N', 'N')
- output[1]:mul(input[3], self.ltp3.trans, 1.0, 1.0, 'N', 'N')
+ for i = 2, #self.dim_in do
+ output[1]:mul(input[i], self["ltp" .. i].trans, 1.0, 1.0, 'N', 'N')
+ end
-- add bias
output[1]:add_row(self.bp.trans, 1.0)
output[1]:sigmoid(output[1])
@@ -53,19 +51,23 @@ end
function GateFFFLayer:back_propagate(bp_err, next_bp_err, input, output)
self.err_bakm:sigmoid_grad(bp_err[1], output[1])
- next_bp_err[1]:mul(self.err_bakm, self.ltp1.trans, 1.0, 0.0, 'N', 'T')
- next_bp_err[2]:mul(self.err_bakm, self.ltp2.trans, 1.0, 0.0, 'N', 'T')
- next_bp_err[3]:mul(self.err_bakm, self.ltp3.trans, 1.0, 0.0, 'N', 'T')
+ for i = 1, #self.dim_in do
+ next_bp_err[i]:mul(self.err_bakm, self["ltp" .. i].trans, 1.0, 0.0, 'N', 'T')
+ end
end
function GateFFFLayer:update(bp_err, input, output)
self.err_bakm:sigmoid_grad(bp_err[1], output[1])
- self.ltp1:update_by_err_input(self.err_bakm, input[1])
- self.ltp2:update_by_err_input(self.err_bakm, input[2])
- self.ltp3:update_by_err_input(self.err_bakm, input[3])
+ for i = 1, #self.dim_in do
+ self["ltp" .. i]:update_by_err_input(self.err_bakm, input[i])
+ end
self.bp:update_by_gradient(self.err_bakm:colsum())
end
function GateFFFLayer:get_params()
- return nerv.ParamRepo({self.ltp1, self.ltp2, self.ltp3, self.bp})
+ local pr = nerv.ParamRepo({self.bp})
+ for i = 1, #self.dim_in do
+ pr:add(self["ltp" .. i].id, self["ltp" .. i])
+ end
+ return pr
end