Mocha
Training LeNet on MNIST
Preparing the Data
Defining the Network Architecture
Configuring the Backend and Building the Network
Configuring the Solver
Coffee Breaks for the Solver
Training
Remarks
Alex’s CIFAR-10 tutorial in Mocha
Caffe’s Tutorial and Code
Preparing the Data
Computation and Loss Layers
Constructing the Network
Configuring the Solver
Training
Pure Julia on CPU
CPU with Native Extension
CUDA with cuDNN
Image Classification with Pre-trained Model
Pre-training with Stacked De-noising Auto-encoders
(Stacked) Denoising Auto-encoders
Experiment Configuration
Pre-training
Fine Tuning
Comparison with Random Initialization
Networks
Overview
Network Architecture
Layer Implementation
Mocha Network Topology Tips
Shared Parameters
Shared Blobs
Layers
Overview
Data Layers
Computation Layers
Loss Layers
Statistics Layers
Utility Layers
Neurons (Activation Functions)
Initializers
Regularizers
Norm Constraints
Data Transformers
Solvers
General Solver Parameters
Learning Rate Policy
Momentum Policy
Solver Algorithms
Solver Coffee Breaks
Coffee Lounge
Built-in Coffee Breaks
Mocha Backends
Pure Julia CPU Backend
CPU Backend with Native Extension
OpenMP on Mac OS X
Native Extension on Windows
Compile Native Extension Manually
CUDA Backend
Recompiling Kernels
Tools
Importing Trained Model from Caffe
Overview
Exporting Caffe’s Snapshot to HDF5
Importing the HDF5 Snapshot to Mocha
Mocha’s HDF5 Snapshot Format
Exporting Caffe’s Mean File
Image Classifier
Blob
Constructors and Destructors
Accessing Properties of a Blob
Accessing Data of a Blob
Layer
Defining a Layer
Characterizing a Layer
Layer Computation API
Layer Parameters
Layer Activation Function
Mocha
Docs
»
Edit on GitHub
Index
A
|
B
|
C
|
D
|
E
|
F
|
G
|
H
|
I
|
K
|
L
|
M
|
N
|
O
|
P
|
R
|
S
|
T
|
V
|
W
|
X
A
AccuracyLayer (built-in class)
ArgmaxLayer (built-in class)
AsyncHDF5DataLayer (built-in class)
auto_scale (DropoutLayer attribute)
B
backward() (built-in function)
batch_size (HDF5DataLayer attribute)
(MemoryDataLayer attribute)
bias_cons (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
bias_init (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
bias_lr (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
bias_regu (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
bottoms (AccuracyLayer attribute)
(ArgmaxLayer attribute)
(ChannelPoolingLayer attribute)
(ConcatLayer attribute)
(ConvolutionLayer attribute)
(CropLayer attribute)
(DropoutLayer attribute)
(ElementWiseLayer attribute)
(HDF5OutputLayer attribute)
(IdentityLayer attribute)
(Index2OnehotLayer attribute)
(InnerProductLayer attribute)
(LRNLayer attribute)
(MemoryOutputLayer attribute)
(MultinomialLogisticLossLayer attribute)
(PoolingLayer attribute)
(PowerLayer attribute)
(ReshapeLayer attribute)
(SoftlabelSoftmaxLossLayer attribute)
(SoftmaxLayer attribute)
(SoftmaxLossLayer attribute)
(SplitLayer attribute)
(SquareLossLayer attribute)
(TiedInnerProductLayer attribute)
C
channel_dim (ChannelPoolingLayer attribute)
ChannelPoolingLayer (built-in class)
chunk_size (AsyncHDF5DataLayer attribute)
ConcatLayer (built-in class)
ConstantInitializer (built-in class)
ConvolutionLayer (built-in class)
copy!() (built-in function)
crop_size (CropLayer attribute)
CropLayer (built-in class)
D
data (MemoryDataLayer attribute)
datasets (HDF5OutputLayer attribute)
DataTransformers.Scale (built-in class)
DataTransformers.SubMean (built-in class)
destroy() (built-in function)
dim (AccuracyLayer attribute)
(ArgmaxLayer attribute)
(ConcatLayer attribute)
(Index2OnehotLayer attribute)
(MultinomialLogisticLossLayer attribute)
(SoftlabelSoftmaxLossLayer attribute)
(SoftmaxLayer attribute)
(SoftmaxLossLayer attribute)
DropoutLayer (built-in class)
dump_statistics() (built-in function)
E
ElementWiseLayer (built-in class)
eltype() (built-in function)
erase!() (built-in function)
every_n_iter (L2Cons attribute)
F
filename (HDF5OutputLayer attribute)
fill!() (built-in function)
filter_cons (ConvolutionLayer attribute)
filter_init (ConvolutionLayer attribute)
filter_lr (ConvolutionLayer attribute)
filter_regu (ConvolutionLayer attribute)
force_overwrite (HDF5OutputLayer attribute)
forward() (built-in function)
G
GaussianInitializer (built-in class)
get_fea_size() (built-in function)
get_height() (built-in function)
get_num() (built-in function)
get_width() (built-in function)
H
HDF5DataLayer (built-in class)
HDF5OutputLayer (built-in class)
I
IdentityLayer (built-in class)
Index2OnehotLayer (built-in class)
InnerProductLayer (built-in class)
K
kernel (ChannelPoolingLayer attribute)
(ConvolutionLayer attribute)
(LRNLayer attribute)
(PoolingLayer attribute)
L
L1Regu (built-in class)
L2Cons (built-in class)
L2Regu (built-in class)
length() (built-in function)
load_from (SolverParameters attribute)
lr_policy (SolverParameters attribute)
LRNLayer (built-in class)
LRPolicy.DecayOnValidation (built-in class)
LRPolicy.Exp (built-in class)
LRPolicy.Fixed (built-in class)
LRPolicy.Inv (built-in class)
LRPolicy.Staged (built-in class)
,
[1]
LRPolicy.Step (built-in class)
M
make_blob() (built-in function)
,
[1]
,
[2]
make_zero_blob() (built-in function)
max_iter (SolverParameters attribute)
mean (GaussianInitializer attribute)
mean_blob (DataTransformers.SubMean attribute)
mean_file (DataTransformers.SubMean attribute)
MemoryDataLayer (built-in class)
MemoryOutputLayer (built-in class)
mode (LRNLayer attribute)
mom_policy (SolverParameters attribute)
MomPolicy.Fixed (built-in class)
MomPolicy.Linear (built-in class)
MomPolicy.Step (built-in class)
MultinomialLogisticLossLayer (built-in class)
N
n_class (Index2OnehotLayer attribute)
n_filter (ConvolutionLayer attribute)
n_group (ConvolutionLayer attribute)
ndims() (built-in function)
Nesterov (built-in class)
neuron (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
Neurons.Identity (built-in class)
Neurons.ReLU (built-in class)
Neurons.Sigmoid (built-in class)
no_copy (SplitLayer attribute)
NoCons (built-in class)
NoRegu (built-in class)
normalize (MultinomialLogisticLossLayer attribute)
(SoftmaxLossLayer attribute)
NullInitializer (built-in class)
O
operation (ElementWiseLayer attribute)
output_dim (InnerProductLayer attribute)
P
pad (ChannelPoolingLayer attribute)
(ConvolutionLayer attribute)
(PoolingLayer attribute)
param_key (ConvolutionLayer attribute)
(InnerProductLayer attribute)
(TiedInnerProductLayer attribute)
pooling (ChannelPoolingLayer attribute)
(PoolingLayer attribute)
PoolingLayer (built-in class)
power (LRNLayer attribute)
(PowerLayer attribute)
PowerLayer (built-in class)
R
random_crop (CropLayer attribute)
random_mirror (CropLayer attribute)
RandomMaskLayer (built-in class)
ratio (DropoutLayer attribute)
regu_coef (SolverParameters attribute)
reset_statistics() (built-in function)
reshape_blob() (built-in function)
ReshapeLayer (built-in class)
S
scale (DataTransformers.Scale attribute)
(LRNLayer attribute)
(PowerLayer attribute)
setup_layer() (built-in function)
SGD (built-in class)
shape (ReshapeLayer attribute)
shift (LRNLayer attribute)
(PowerLayer attribute)
show_iter (TrainingSummary attribute)
show_lr (TrainingSummary attribute)
show_mom (TrainingSummary attribute)
show_obj_val (TrainingSummary attribute)
shuffle (HDF5DataLayer attribute)
shutdown() (built-in function)
size() (built-in function)
,
[1]
Snapshot (built-in class)
SoftlabelSoftmaxLossLayer (built-in class)
SoftmaxLayer (built-in class)
SoftmaxLossLayer (built-in class)
SolverParameters (built-in class)
source (HDF5DataLayer attribute)
SplitLayer (built-in class)
SquareLossLayer (built-in class)
std (GaussianInitializer attribute)
stride (ChannelPoolingLayer attribute)
(ConvolutionLayer attribute)
(PoolingLayer attribute)
T
threshold (L2Cons attribute)
tied_param_key (TiedInnerProductLayer attribute)
TiedInnerProductLayer (built-in class)
tops (ArgmaxLayer attribute)
(ChannelPoolingLayer attribute)
(ConcatLayer attribute)
(ConvolutionLayer attribute)
(CropLayer attribute)
(ElementWiseLayer attribute)
(HDF5DataLayer attribute)
(IdentityLayer attribute)
(Index2OnehotLayer attribute)
(InnerProductLayer attribute)
(LRNLayer attribute)
(MemoryDataLayer attribute)
(PoolingLayer attribute)
(PowerLayer attribute)
(ReshapeLayer attribute)
(SoftmaxLayer attribute)
(SplitLayer attribute)
(TiedInnerProductLayer attribute)
TrainingSummary (built-in class)
transformers (HDF5DataLayer attribute)
(MemoryDataLayer attribute)
V
ValidationPerformance (built-in class)
value (ConstantInitializer attribute)
W
weight (MultinomialLogisticLossLayer attribute)
(SoftlabelSoftmaxLossLayer attribute)
(SoftmaxLossLayer attribute)
(SquareLossLayer attribute)
weight_cons (InnerProductLayer attribute)
weight_init (InnerProductLayer attribute)
weight_lr (InnerProductLayer attribute)
weight_regu (InnerProductLayer attribute)
weights (MultinomialLogisticLossLayer attribute)
(SoftmaxLossLayer attribute)
X
XavierInitializer (built-in class)
Read the Docs
v: v0.0.8
Versions
latest
stable
v0.0.8
v0.0.7
v0.0.6
v0.0.5
v0.0.4
v0.0.3
v0.0.2
v0.0.1
Downloads
On Read the Docs
Project Home
Builds
Free document hosting provided by
Read the Docs
.