Exercise 15.14: Gaussian Distributions And Maximum Entropy chapter 15 Let Q=N(0,σ2) and P∈Qσ2 be arbitrary. Then 0≤KL(P‖Q)=−H(P)−∫plogqdν where the latter term equals (1)−∫plogqdν=12log(2πσ2)+12σ2VP(X). Hence H(P)≤12[1+log(2πσ2)]=H(Q) by VP(X)≤σ2. « Exercise 15.13: KL Divergence for Multivariate Gaussian Exercise 15.15: Sharper Bound for Variable Selection in Sparse PCA » Published on 29 January 2022. Please enable JavaScript to view the comments powered by Disqus.