# Neural Net

---

`NeuralNet` in SINGA represents an instance of user's neural net model. As the
neural net typically consists of a set of layers, `NeuralNet` comprises
a set of unidirectionally connected [Layer](layer.html)s.
This page describes how to convert an user's neural net into
the configuration of `NeuralNet`.

<img src="../_static/images/model-category.png" align="center" width="200px"/>
<span><strong>Figure 1 - Categorization of popular deep learning models.</strong></span>

## Net structure configuration

Users configure the `NeuralNet` by listing all layers of the neural net and
specifying each layer's source layer names. Popular deep learning models can be
categorized as Figure 1. The subsequent sections give details for each
category.

### Feed-forward models

<div align = "left">
<img src="../_static/images/mlp-net.png" align="center" width="200px"/>
<span><strong>Figure 2 - Net structure of a MLP model.</strong></span>
</div>

Feed-forward models, e.g., CNN and MLP, can easily get configured as their layer
connections are undirected without circles. The
configuration for the MLP model shown in Figure 1 is as follows,

    net {
      layer {
        name : 'data"
        type : kData
      }
      layer {
        name : 'image"
        type : kImage
        srclayer: 'data'
      }
      layer {
        name : 'label"
        type : kLabel
        srclayer: 'data'
      }
      layer {
        name : 'hidden"
        type : kHidden
        srclayer: 'image'
      }
      layer {
        name : 'softmax"
        type : kSoftmaxLoss
        srclayer: 'hidden'
        srclayer: 'label'
      }
    }

### Energy models

<img src="../_static/images/rbm-rnn.png" align="center" width="500px"/>
<span><strong>Figure 3 - Convert connections in RBM and RNN.</strong></span>


For energy models including RBM, DBM,
etc., their connections are undirected (i.e., Category B). To represent these models using
`NeuralNet`, users can simply replace each connection with two directed
connections, as shown in Figure 3a. In other words, for each pair of connected layers, their source
layer field should include each other's name.
The full [RBM example](rbm.html) has
detailed neural net configuration for a RBM model, which looks like

    net {
      layer {
        name : "vis"
        type : kVisLayer
        param {
          name : "w1"
        }
        srclayer: "hid"
      }
      layer {
        name : "hid"
        type : kHidLayer
        param {
          name : "w2"
          share_from: "w1"
        }
        srclayer: "vis"
      }
    }

### RNN models

For recurrent neural networks (RNN), users can remove the recurrent connections
by unrolling the recurrent layer.  For example, in Figure 3b, the original
layer is unrolled into a new layer with 4 internal layers. In this way, the
model is like a normal feed-forward model, thus can be configured similarly.
The [RNN example](rnn.html) has a full neural net
configuration for a RNN model.


## Configuration for multiple nets

Typically, a training job includes three neural nets for
training, validation and test phase respectively. The three neural nets share most
layers except the data layer, loss layer or output layer, etc..  To avoid
redundant configurations for the shared layers, users can uses the `exclude`
filed to filter a layer in the neural net, e.g., the following layer will be
filtered when creating the testing `NeuralNet`.


    layer {
      ...
      exclude : kTest # filter this layer for creating test net
    }


## Neural net partitioning

A neural net can be partitioned in different ways to distribute the training
over multiple workers.

### Batch and feature dimension

<img src="../_static/images/partition_fc.png" align="center" width="400px"/>
<span><strong>Figure 4 - Partitioning of a fully connected layer.</strong></span>


Every layer's feature blob is considered a matrix whose rows are feature
vectors. Thus, one layer can be split on two dimensions. Partitioning on
dimension 0 (also called batch dimension) slices the feature matrix by rows.
For instance, if the mini-batch size is 256 and the layer is partitioned into 2
sub-layers, each sub-layer would have 128 feature vectors in its feature blob.
Partitioning on this dimension has no effect on the parameters, as every
[Param](param.html) object is replicated in the sub-layers. Partitioning on dimension
1 (also called feature dimension) slices the feature matrix by columns. For
example, suppose the original feature vector has 50 units, after partitioning
into 2 sub-layers, each sub-layer would have 25 units. This partitioning may
result in [Param](param.html) object being split, as shown in
Figure 4. Both the bias vector and weight matrix are
partitioned into two sub-layers.


### Partitioning configuration

There are 4 partitioning schemes, whose configurations are give below,

  1. Partitioning each singe layer into sub-layers on batch dimension (see
  below). It is enabled by configuring the partition dimension of the layer to
  0, e.g.,

          # with other fields omitted
          layer {
            partition_dim: 0
          }

  2. Partitioning each singe layer into sub-layers on feature dimension (see
  below).  It is enabled by configuring the partition dimension of the layer to
  1, e.g.,

          # with other fields omitted
          layer {
            partition_dim: 1
          }

  3. Partitioning all layers into different subsets. It is enabled by
  configuring the location ID of a layer, e.g.,

          # with other fields omitted
          layer {
            location: 1
          }
          layer {
            location: 0
          }


  4. Hybrid partitioning of strategy 1, 2 and 3. The hybrid partitioning is
  useful for large models. An example application is to implement the
  [idea proposed by Alex](http://arxiv.org/abs/1404.5997).
  Hybrid partitioning is configured like,

          # with other fields omitted
          layer {
            location: 1
          }
          layer {
            location: 0
          }
          layer {
            partition_dim: 0
            location: 0
          }
          layer {
            partition_dim: 1
            location: 0
          }

Currently SINGA supports strategy-2 well. Other partitioning strategies are
are under test and will be released in later version.

## Parameter sharing

Parameters can be shared in two cases,

  * sharing parameters among layers via user configuration. For example, the
  visible layer and hidden layer of a RBM shares the weight matrix, which is configured through
  the `share_from` field as shown in the above RBM configuration. The
  configurations must be the same (except name) for shared parameters.

  * due to neural net partitioning, some `Param` objects are replicated into
  different workers, e.g., partitioning one layer on batch dimension. These
  workers share parameter values. SINGA controls this kind of parameter
  sharing automatically, users do not need to do any configuration.

  * the `NeuralNet` for training and testing (and validation) share most layers
  , thus share `Param` values.

If the shared `Param` instances resident in the same process (may in different
threads), they use the same chunk of memory space for their values. But they
would have different memory spaces for their gradients. In fact, their
gradients will be averaged by the stub or server.

## Advanced user guide

### Creation

    static NeuralNet* NeuralNet::Create(const NetProto& np, Phase phase, int num);

The above function creates a `NeuralNet` for a given phase, and returns a
pointer to the `NeuralNet` instance. The phase is in {kTrain,
kValidation, kTest}. `num` is used for net partitioning which indicates the
number of partitions.  Typically, a training job includes three neural nets for
training, validation and test phase respectively. The three neural nets share most
layers except the data layer, loss layer or output layer, etc.. The `Create`
function takes in the full net configuration including layers for training,
validation and test.  It removes layers for phases other than the specified
phase based on the `exclude` field in
[layer configuration](layer.html):

    layer {
      ...
      exclude : kTest # filter this layer for creating test net
    }

The filtered net configuration is passed to the constructor of `NeuralNet`:

    NeuralNet::NeuralNet(NetProto netproto, int npartitions);

The constructor creates a graph representing the net structure firstly in

    Graph* NeuralNet::CreateGraph(const NetProto& netproto, int npartitions);

Next, it creates a layer for each node and connects layers if their nodes are
connected.

    void NeuralNet::CreateNetFromGraph(Graph* graph, int npartitions);

Since the `NeuralNet` instance may be shared among multiple workers, the
`Create` function returns a pointer to the `NeuralNet` instance .

### Parameter sharing

 `Param` sharing
is enabled by first sharing the Param configuration (in `NeuralNet::Create`)
to create two similar (e.g., the same shape) Param objects, and then calling
(in `NeuralNet::CreateNetFromGraph`),

    void Param::ShareFrom(const Param& from);

It is also possible to share `Param`s of two nets, e.g., sharing parameters of
the training net and the test net,

    void NeuralNet:ShareParamsFrom(NeuralNet* other);

It will call `Param::ShareFrom` for each Param object.

### Access functions
`NeuralNet` provides a couple of access function to get the layers and params
of the net:

    const std::vector<Layer*>& layers() const;
    const std::vector<Param*>& params() const ;
    Layer* name2layer(string name) const;
    Param* paramid2param(int id) const;


### Partitioning


#### Implementation

SINGA partitions the neural net in `CreateGraph` function, which creates one
node for each (partitioned) layer. For example, if one layer's partition
dimension is 0 or 1, then it creates `npartition` nodes for it; if the
partition dimension is -1, a single node is created, i.e., no partitioning.
Each node is assigned a partition (or location) ID. If the original layer is
configured with a location ID, then the ID is assigned to each newly created node.
These nodes are connected according to the connections of the original layers.
Some connection layers will be added automatically.
For instance, if two connected sub-layers are located at two
different workers, then a pair of bridge layers is inserted to transfer the
feature (and gradient) blob between them. When two layers are partitioned on
different dimensions, a concatenation layer which concatenates feature rows (or
columns) and a slice layer which slices feature rows (or columns) would be
inserted. These connection layers help making the network communication and
synchronization transparent to the users.

#### Dispatching partitions to workers

Each (partitioned) layer is assigned a location ID, based on which it is dispatched to one
worker. Particularly, the pointer to the `NeuralNet` instance is passed
to every worker within the same group, but each worker only computes over the
layers that have the same partition (or location) ID as the worker's ID.  When
every worker computes the gradients of the entire model parameters
(strategy-2), we refer to this process as data parallelism.  When different
workers compute the gradients of different parameters (strategy-3 or
strategy-1), we call this process model parallelism.  The hybrid partitioning
leads to hybrid parallelism where some workers compute the gradients of the
same subset of model parameters while other workers compute on different model
parameters.  For example, to implement the hybrid parallelism in for the
[DCNN model](http://arxiv.org/abs/1404.5997), we set `partition_dim = 0` for
lower layers and `partition_dim = 1` for higher layers.