Iblis Lin
|
6c7613e02b
|
batchnorm: leverage TrackedArray mean
|
2017-11-02 14:20:34 +08:00 |
|
Iblis Lin
|
88bd8a8fbd
|
batchnorm: make CuArrays happy
|
2017-11-02 14:02:41 +08:00 |
|
Iblis Lin
|
477da75428
|
batchnorm: fix mapchildren
|
2017-11-02 13:32:12 +08:00 |
|
Iblis Lin
|
5253841acc
|
batchnorm: update docs
|
2017-11-02 13:32:12 +08:00 |
|
Iblis Lin
|
b3356cc6bb
|
batchnorm: batch σ correct coefficient
|
2017-11-02 13:32:12 +08:00 |
|
Iblis Lin
|
e0201be770
|
batchnorm: parameterize momentum and epsilon
|
2017-11-02 13:32:12 +08:00 |
|
Iblis Lin
|
669273b008
|
layer: implement BatchNorm layer
See [Batch Normalization: Accelerating Deep Network Training by Reducing
Internal Covariate Shift](https://arxiv.org/pdf/1502.03167.pdf)
|
2017-11-02 13:32:12 +08:00 |
|
Mike J Innes
|
e943a39ee7
|
combine special cases
|
2017-10-31 16:37:33 +00:00 |
|
Mike J Innes
|
cf6b930f63
|
reorganise
|
2017-10-26 11:46:12 +01:00 |
|
CarloLucibello
|
711ea09d99
|
address comments
|
2017-10-25 02:35:27 +02:00 |
|
CarloLucibello
|
536ab3861d
|
setmode! -> testmode!
|
2017-10-23 16:23:29 +02:00 |
|
CarloLucibello
|
86c7c9246e
|
add == and < for tracked arrays
|
2017-10-23 11:41:08 +02:00 |
|
CarloLucibello
|
2e1ed4c3fc
|
add dropout
|
2017-10-23 10:12:53 +02:00 |
|
Mike J Innes
|
2a66545ef8
|
rnn state reset
|
2017-10-19 17:21:08 +01:00 |
|
Mike J Innes
|
b817ce632c
|
syntax highlighting
|
2017-10-18 15:44:06 +01:00 |
|
Mike J Innes
|
fd249b773e
|
rnn docs
|
2017-10-18 15:30:05 +01:00 |
|
Mike J Innes
|
190f48a709
|
nnlib docs
|
2017-10-18 14:40:58 +01:00 |
|
Mike J Innes
|
0fbc8dff61
|
typoe
|
2017-10-18 12:48:58 +01:00 |
|
Mike J Innes
|
d6dd27dae5
|
dense layer example
|
2017-10-18 12:47:45 +01:00 |
|
Mike J Innes
|
23674b2555
|
logitcrossentropy tweaks
|
2017-10-17 17:58:32 +01:00 |
|
pevnak
|
4aa7741ba9
|
logit cross entropy
|
2017-10-17 17:57:46 +01:00 |
|
Mike J Innes
|
6dff8ca8d3
|
rename crossentropy loss
|
2017-10-17 17:36:18 +01:00 |
|
Mike J Innes
|
949fd9ba97
|
loss function tweaks
|
2017-10-17 17:30:11 +01:00 |
|
Mike J Innes
|
d3db051ca0
|
flip
|
2017-10-16 08:53:39 +01:00 |
|
Dave Kleinschmidt
|
2b95aff158
|
actually use init argument in LSTMCell
|
2017-10-03 19:26:42 +01:00 |
|
Mike J Innes
|
a60a754d68
|
beginnings of gpu support
|
2017-09-27 21:58:34 +01:00 |
|
Mike J Innes
|
4bafa2b374
|
generic tree functions
|
2017-09-27 21:11:21 +01:00 |
|
Mike J Innes
|
2ec8401d2c
|
remove compiler
|
2017-09-27 20:48:39 +01:00 |
|
Mike J Innes
|
f2052739c1
|
tweaks
|
2017-09-12 14:11:03 +01:00 |
|
Mike J Innes
|
a3fe89e348
|
rnn tanh by default
|
2017-09-12 13:12:25 +01:00 |
|
Mike J Innes
|
6728295355
|
Merge pull request #63 from JobJob/rnncell-args
Enables passing an activation function to RNN/RNNCell
|
2017-09-12 13:10:43 +01:00 |
|
Mike J Innes
|
28bbef81b9
|
f
|
2017-09-12 13:06:32 +01:00 |
|
Mike J Innes
|
972ecab9f9
|
rm Over Seq
|
2017-09-12 13:03:16 +01:00 |
|
Joel Mason
|
00439555d1
|
Enables passing an activation function to RNN/RNNCell
Also, fixes it not using the `init` function provided
|
2017-09-12 20:54:56 +10:00 |
|
Mike J Innes
|
b9652f1812
|
typo, fixes #61
|
2017-09-12 10:45:07 +01:00 |
|
Mike J Innes
|
5f24d61ba3
|
important
|
2017-09-11 14:10:12 +01:00 |
|
Mike J Innes
|
7041ab9960
|
rm chainseq
|
2017-09-11 14:02:43 +01:00 |
|
Mike J Innes
|
33a5d26e57
|
chain utility note
|
2017-09-09 20:02:48 -04:00 |
|
Mike J Innes
|
fedee95b14
|
docs updates
|
2017-09-09 19:58:32 -04:00 |
|
Mike J Innes
|
a36d6d2af3
|
layer docs
|
2017-09-08 17:52:41 -04:00 |
|
Mike J Innes
|
f55b8cd20e
|
track -> param
|
2017-09-07 15:13:04 -04:00 |
|
Mike J Innes
|
e837bb0745
|
rnn stuff
|
2017-09-07 00:27:04 -04:00 |
|
Mike J Innes
|
cca4d25a10
|
efficient traversal
|
2017-09-06 23:09:32 -04:00 |
|
Mike J Innes
|
4083c34547
|
seq stuff
|
2017-09-06 18:59:36 -04:00 |
|
Mike J Innes
|
1946c46e29
|
basic seq functionality
|
2017-09-06 14:03:25 -04:00 |
|
Mike J Innes
|
c95e9376a5
|
constructors
|
2017-09-05 19:25:34 -04:00 |
|
Mike J Innes
|
b023da1b7d
|
lstm initialisation
|
2017-09-05 02:42:32 -04:00 |
|
Mike J Innes
|
61de692b50
|
lstm nonzero hidden state
|
2017-09-05 02:37:48 -04:00 |
|
Mike J Innes
|
ec02f1fabd
|
batching in rnns
|
2017-09-05 02:29:31 -04:00 |
|
Mike J Innes
|
bd5822fd71
|
cleaner lstm
|
2017-09-03 02:24:47 -04:00 |
|
Mike J Innes
|
9642ae8cd6
|
basic recurrence
|
2017-09-03 02:12:44 -04:00 |
|
Mike J Innes
|
f6771b98cd
|
clearer name for dense
|
2017-09-02 16:50:11 -04:00 |
|
Mike J Innes
|
9ce0439943
|
better mse
|
2017-08-24 11:40:51 +01:00 |
|
Mike J Innes
|
e4e9794f5e
|
loss function gradients
|
2017-08-23 17:50:43 +01:00 |
|
Mike J Innes
|
0ce8c0cee4
|
param collection
|
2017-08-22 17:13:03 +01:00 |
|
Mike J Innes
|
ef681f16ea
|
use nnlib for activations
|
2017-08-21 17:53:04 +01:00 |
|
Mike J Innes
|
3e0f45046c
|
nicer show
|
2017-08-21 17:20:09 +01:00 |
|
Mike J Innes
|
09974caba0
|
extend affine with activation
|
2017-08-20 13:35:35 +01:00 |
|
Mike J Innes
|
18e69b33c9
|
forwarddiff does these
|
2017-08-19 22:05:50 +01:00 |
|
Mike J Innes
|
ad0e0ea5a7
|
explicitly broadcast sigmoid
|
2017-08-19 22:04:47 +01:00 |
|
Mike J Innes
|
4a9dc40e7c
|
simplify organisation
|
2017-08-19 20:52:29 +01:00 |
|
Mike J Innes
|
318e503d9b
|
remove rnns
|
2017-08-19 20:52:17 +01:00 |
|
Mike J Innes
|
b18d8cd08f
|
rename chain.jl
|
2017-08-19 20:26:07 +01:00 |
|
Mike J Innes
|
0222103c7f
|
fix these later
|
2017-08-19 20:05:22 +01:00 |
|
Mike J Innes
|
bd6bffde48
|
silo the compiler
|
2017-08-19 20:04:21 +01:00 |
|
Mike J Innes
|
f8482ff80c
|
remove back!, update!
|
2017-08-18 10:18:45 +01:00 |
|
Mike J Innes
|
21089fea9c
|
remove some explicit back!
|
2017-08-18 01:19:10 +01:00 |
|
Mike J Innes
|
052cc52ada
|
remove inference, for now
|
2017-08-18 01:19:10 +01:00 |
|
Mike J Innes
|
8869f1f953
|
these are now useless
|
2017-08-18 01:05:46 +01:00 |
|
ylxdzsw
|
c9ae219613
|
simplify back! of Chain
|
2017-06-14 21:58:43 +08:00 |
|
ylxdzsw
|
cca21a617c
|
training julia models
|
2017-06-12 19:39:34 +08:00 |
|
Mike J Innes
|
099b9fbeb7
|
collect exports
|
2017-06-05 16:37:06 +01:00 |
|
Mike J Innes
|
837173d65b
|
model abstract is pretty useless
|
2017-06-05 16:08:23 +01:00 |
|
Mike J Innes
|
4685d2e672
|
strip down non-obvious exports
|
2017-06-05 15:47:26 +01:00 |
|
Mike J Innes
|
d1f370a2f1
|
move stateful
|
2017-06-02 16:02:47 +01:00 |
|
Mike J Innes
|
2b85c76785
|
preserve original param states
|
2017-06-02 15:46:24 +01:00 |
|
Mike J Innes
|
07c523a13d
|
fix flatten/softmax batching behaviour
|
2017-06-01 19:28:02 +01:00 |
|
Elliot Saba
|
dfa4dcccb5
|
Fix activation functions
|
2017-05-25 18:12:58 -07:00 |
|
Mike J Innes
|
e04dcbd460
|
reorganise recurrent stuff
|
2017-05-04 10:45:44 +01:00 |
|
Mike J Innes
|
36f4103d7d
|
cost is really a layer
|
2017-05-01 12:30:28 +01:00 |
|
Mike J Innes
|
14afe54143
|
fixes for recurrent networks
|
2017-04-19 17:17:37 +01:00 |
|
Mike J Innes
|
90edefe072
|
use broadcasting plus
|
2017-03-21 01:18:00 +00:00 |
|
Mike J Innes
|
eadf96605c
|
fix recurrent layer batching
|
2017-03-20 23:11:08 +00:00 |
|
Mike J Innes
|
6756ce7528
|
relu inferrence
|
2017-03-20 19:57:11 +00:00 |
|
Mike J Innes
|
d73e962da9
|
replace old shape inference system
|
2017-03-17 16:34:54 +00:00 |
|
Mike J Innes
|
2adc3cd18e
|
new struct syntax
|
2017-03-14 17:56:03 +00:00 |
|
Mike J Innes
|
6751657dfa
|
typealias / abstract deps
|
2017-03-14 16:51:31 +00:00 |
|
Mike J Innes
|
c4d815b5fc
|
move todo
|
2017-03-09 00:12:49 +00:00 |
|
Mike J Innes
|
6a3bed1e61
|
this is pretty useless
|
2017-03-08 15:38:55 +00:00 |
|
Mike J Innes
|
06f2ee2284
|
iterate over chain
|
2017-03-07 14:37:37 +00:00 |
|
Mike J Innes
|
2a57150bce
|
AvgPool shim
|
2017-03-06 17:21:35 +00:00 |
|
Mike J Innes
|
4d4979b401
|
better alternative to basemodel
|
2017-02-28 16:42:48 +00:00 |
|
Mike J Innes
|
5f1f2ebaa2
|
model storage notes
|
2017-02-28 16:41:33 +00:00 |
|
Mike J Innes
|
3fdffea37d
|
fix
|
2017-02-20 21:50:01 +00:00 |
|
Mike J Innes
|
d7fe525f4d
|
fix the build
|
2017-01-16 01:21:45 +01:00 |
|
Mike J Innes
|
c82716b535
|
move activations
|
2016-12-15 23:24:10 +00:00 |
|
Mike J Innes
|
6114b70f76
|
use regular +
|
2016-12-15 22:57:36 +00:00 |
|
Mike J Innes
|
62fd13bded
|
consistently use delta for gradients
|
2016-12-15 21:37:39 +00:00 |
|
Mike J Innes
|
4b64bf11a5
|
fix lstm
|
2016-12-15 20:53:08 +00:00 |
|
Mike J Innes
|
f31b539566
|
make these fit with julia semantics
|
2016-11-15 16:40:17 +00:00 |
|
Mike J Innes
|
9062792495
|
rename affine file
|
2016-11-15 00:09:53 +00:00 |
|
Mike J Innes
|
bdd05157e2
|
dense -> affine
|
2016-11-14 22:16:00 +00:00 |
|
Mike J Innes
|
e4a6ca5f9e
|
remove custom show
|
2016-11-13 15:35:20 +00:00 |
|
Mike J Innes
|
ad6e6b4116
|
update recurrent usage
|
2016-11-08 18:08:13 +00:00 |
|
Mike J Innes
|
d7d95feab8
|
actually get GRU working
|
2016-11-02 00:36:13 +00:00 |
|
Mike J Innes
|
85415d4244
|
throw GRU together
|
2016-11-01 14:42:41 +00:00 |
|
Mike J Innes
|
7cd94b4a5d
|
well, that was easy 😎
|
2016-10-31 11:01:19 +00:00 |
|
Mike J Innes
|
508364407e
|
simplify recurrent layer
|
2016-10-30 16:07:18 +00:00 |
|
Mike J Innes
|
81d9743836
|
export recurrent
|
2016-10-30 11:41:52 +00:00 |
|
Mike J Innes
|
89c4a6df31
|
this is no longer test code
|
2016-10-29 00:13:32 +01:00 |
|
Mike J Innes
|
eb78f67a93
|
refactor input model
|
2016-10-25 23:10:35 +01:00 |
|
Mike J Innes
|
b115d8ce3f
|
model -> net
|
2016-10-12 16:28:16 +01:00 |
|
Mike J Innes
|
a56af5d16e
|
reshape layer
|
2016-10-10 23:48:16 +01:00 |
|
Mike J Innes
|
438dc9d40a
|
fix conv2d shape inference
|
2016-10-10 23:20:40 +01:00 |
|
Mike J Innes
|
bf04b70ad1
|
Float32 by default
|
2016-10-04 22:36:56 +01:00 |
|
Mike J Innes
|
cc1ca4c3c2
|
Conv2D tweaks
|
2016-10-04 22:23:26 +01:00 |
|
Mike J Innes
|
2609d47ce9
|
work more nicely with TF batching
|
2016-10-04 21:10:50 +01:00 |
|
Mike J Innes
|
8961b4c10f
|
basic convnet example working
|
2016-09-06 18:11:15 +01:00 |
|
Mike J Innes
|
6503496c39
|
improve printing
|
2016-09-06 18:11:14 +01:00 |
|
Mike J Innes
|
d58fefb972
|
tweak note
|
2016-09-06 18:11:14 +01:00 |
|
Mike J Innes
|
19b5e8bd21
|
loop lifting
|
2016-09-06 18:11:14 +01:00 |
|
Mike J Innes
|
c92cff5dce
|
a bunch of stuff
|
2016-09-06 18:11:05 +01:00 |
|
Mike J Innes
|
afac5d8bfe
|
better default init
|
2016-09-06 18:10:21 +01:00 |
|
Mike J Innes
|
fd67383494
|
don't print reams of data
|
2016-09-06 18:10:21 +01:00 |
|
Mike J Innes
|
2635283bf1
|
small reorg
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
b8565a4cc3
|
update api
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
6808a92793
|
anonymous models
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
545d4480ed
|
tweaks
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
20b13b9a41
|
fix
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
c4b96dbc98
|
add gradients in place
|
2016-09-06 18:10:20 +01:00 |
|
Mike J Innes
|
cf2b168a55
|
support for inner layers
|
2016-09-06 18:10:19 +01:00 |
|
Mike J Innes
|
8e92403436
|
new params approach
|
2016-09-06 18:10:19 +01:00 |
|
Mike J Innes
|
cab43611e3
|
basic MXNet output
|
2016-09-06 18:10:19 +01:00 |
|
Mike J Innes
|
5995597ebc
|
dense layer
|
2016-09-06 18:10:19 +01:00 |
|
Mike J Innes
|
9986a1c163
|
remove old torch-esque code
|
2016-09-06 18:10:18 +01:00 |
|
Mike Innes
|
e5856d8b27
|
init
|
2016-09-06 18:10:18 +01:00 |
|