Fig. 7: Using TERP to explain and check the reliability of Att-BLSTM model trained on AG’s news corpus to predict the news story titled “AI predicts protein structures”. | Nature Communications

Fig. 7: Using TERP to explain and check the reliability of Att-BLSTM model trained on AG’s news corpus to predict the news story titled “AI predicts protein structures”.

From: Thermodynamics-inspired explanations of artificial intelligence

Fig. 7

a\({{{{\mathcal{U}}}}}^{\, j}\) vs. j, b\({{{{\mathcal{S}}}}}^{\, j}\) vs. j, cθ j vs. j, d ζ j vs. j plots showing the optimal explanation appears at j = 2, due to the maximum drop in θ j as j is increased from 1 to 2.

Back to article page