Computational Intelligence and Neuroscience

Research Article

Recognition of Emotions Using Multichannel EEG Data and DBN-GC-Based Ensemble Deep Learning Framework

Pseudocodes for training the RBM with glia chain.

	Input: a training sample x₁, learning rate ε, glia effect vector g
	Output: weight matrix W, bias vector of the visible layer b, bias vector of the hidden layer c
	Start training
	Initialize v⁽⁰⁾: v⁽⁰⁾= x₁ (initialize state value of the visible layer v)
	for j = 1 : m (for all hidden units)
	(h_j⁽⁰⁾ = 1 \| v⁽⁰⁾) = σ ( + c_j)
	End for
	Extract a sample h⁽⁰⁾ according to h⁽⁰⁾∼ (h⁽⁰⁾ \| v⁽⁰⁾)
	for i = 1 : (for all visible units)
	( = 1 \| h⁽⁰⁾) = σ (Σ_jW_ijh_j⁽⁰⁾ + b_i)
	End for
	Extract a sample v⁽¹⁾ according to v⁽¹⁾∼ (v⁽¹⁾ \| h⁽⁰⁾)
	for j = 1 : m (for all hidden units, calculate the output value without glia effect)
	h_j^(1)∗ = Σ_iW_ij⁽¹⁾+ c_j
	End for
	update the glia effect vector g
	for j = 1 : m (for all hidden units, calculate the output value with glia effect)
	⁽¹⁾ = σ (^(1)∗ + )
	End for
	Update parameters
	W ⟵ W + ε ( (h⁽⁰⁾ = 1 \| v⁽⁰⁾)v^(0)T − p(h⁽¹⁾ = 1 \| v⁽¹⁾)v^(1)T)
	b ⟵ b + ε (v⁽⁰⁾ − v⁽¹⁾)
	c ⟵ c + ε ( (h⁽⁰⁾ = 1 \| v⁽⁰⁾) − (h⁽¹⁾ = 1 \| v⁽¹⁾))