class SHAInet::CNN

• SHAInet::CNN
• Reference
• Object

Defined in:

shainet/cnn/cnn.cr

Constant Summary

COST_FUNCTIONS = ["mse", "c_ent", "exp", "hel_d", "kld", "gkld", "ita_sai_d"]

Log = ::Log.for(self)

Constructors

• .new

Instance Method Summary

• #add_conv(filters_num : Int32, window_size : Int32, stride : Int32, padding : Int32, activation_function : ActivationFunction = SHAInet.none)
• #add_dropout(drop_percent : Int32 = 5)
• #add_fconnect(l_size : Int32, activation_function : ActivationFunction = SHAInet.none)
• #add_input(input_volume : Array(Int32))
• #add_maxpool(pool : Int32, stride : Int32)
• #add_relu(l_relu_slope : Float64 = 0.0)
• #add_softmax(range : Range(Int32, Int32) = (0..-1))
• #alpha : Float64

  Parameters for Adam

• #alpha=(alpha : Float64)

  Parameters for Adam

• #beta1 : Float64
• #beta2 : Float64
• #check_nan : Bool
• #delta_max : Float64

  Parameters for Rprop

• #delta_max=(delta_max : Float64)

  Parameters for Rprop

• #delta_min : Float64

  Parameters for Rprop

• #delta_min=(delta_min : Float64)

  Parameters for Rprop

• #epsilon : Float64
• #error_signal : Array(Float64)
• #etah_minus : Float64

  Parameters for Rprop

• #etah_minus=(etah_minus : Float64)

  Parameters for Rprop

• #etah_plus : Float64

  Parameters for Rprop

• #etah_plus=(etah_plus : Float64)

  Parameters for Rprop

• #evaluate(input_data : Array(Array(Array(GenNum))), expected_output : Array(GenNum), cost_function : CostFunction = SHAInet.quadratic_cost)
• #get_cost_proc(function_name : String) : CostFunction
• #inspect(what : String | Symbol)
• #layers : Array(CNNLayer | ConvLayer)

  General network parameters

• #learning_rate : Float64

  Parameters for SGD + Momentum

• #learning_rate=(learning_rate : Float64)

  Parameters for SGD + Momentum

• #log_summary(e)
• #mean_error : Float64
• #momentum : Float64

  Parameters for SGD + Momentum

• #momentum=(momentum : Float64)

  Parameters for SGD + Momentum

• #output : Array(Float64)
• #prev_mean_error : Float64
• #propagate_backwards
• #run(input_data : Array(Array(Array(GenNum))), stealth : Bool = true) : Array(Float64)
• #time_step : Int32
• #total_error : Float64
• #train(data : NamedTuple(input: Array(Array(Array(Float64))), output: Array(Float64)), training_type : Symbol | String, cost_function : Symbol | String | CostFunction = :mse, epochs : Int32 = 1, error_threshold : Float64 = 0.0, log_each : Int32 = 1000)

  Online train, updates weights/biases after each data point (stochastic gradient descent)

• #train_batch(data : Array(CNNPair), training_type : Symbol | String, cost_function : Symbol | String | CostFunction = :mse, epochs : Int32 = 1, error_threshold : Float64 = 0.0, log_each : Int32 = 1000, mini_batch_size : Int32 | Nil = nil)

  Batch train, updates weights/biases using a gradient sum from all data points in the batch (using gradient descent) ameba:disable Metrics/CyclomaticComplexity

• #update_output_gradients(cost_function_derivatives : Array(Float64))

  Update the output layer gradients manually

• #update_wb(training_type : Symbol | String, batch : Bool = false)

  Go over all layers and update the weights and biases, based on learning type chosen

• #verify_data(data : Array(CNNPair))

Constructor Detail

Instance Method Detail