Convergence to minima for the continuous version of Backtracking Gradient Descent
This addresses the issue of gradient descent getting stuck at saddle points, which is a common problem in non-convex optimization for machine learning practitioners, though it is incremental as it builds on existing backtracking methods.
The paper tackles the problem of ensuring convergence to minima rather than saddle points in gradient descent by proposing a continuous version of Backtracking Gradient Descent, proving that for almost all initial points, the sequence either diverges or converges to a critical point, avoiding saddle points with probability one.
The main result of this paper is: {\bf Theorem.} Let $f:\mathbb{R}^k\rightarrow \mathbb{R}$ be a $C^{1}$ function, so that $\nabla f$ is locally Lipschitz continuous. Assume moreover that $f$ is $C^2$ near its generalised saddle points. Fix real numbers $δ_0>0$ and $0<α<1$. Then there is a smooth function $h:\mathbb{R}^k\rightarrow (0,δ_0]$ so that the map $H:\mathbb{R}^k\rightarrow \mathbb{R}^k$ defined by $H(x)=x-h(x)\nabla f(x)$ has the following property: (i) For all $x\in \mathbb{R}^k$, we have $f(H(x)))-f(x)\leq -αh(x)||\nabla f(x)||^2$. (ii) For every $x_0\in \mathbb{R}^k$, the sequence $x_{n+1}=H(x_n)$ either satisfies $\lim_{n\rightarrow\infty}||x_{n+1}-x_n||=0$ or $ \lim_{n\rightarrow\infty}||x_n||=\infty$. Each cluster point of $\{x_n\}$ is a critical point of $f$. If moreover $f$ has at most countably many critical points, then $\{x_n\}$ either converges to a critical point of $f$ or $\lim_{n\rightarrow\infty}||x_n||=\infty$. (iii) There is a set $\mathcal{E}_1\subset \mathbb{R}^k$ of Lebesgue measure $0$ so that for all $x_0\in \mathbb{R}^k\backslash \mathcal{E}_1$, the sequence $x_{n+1}=H(x_n)$, {\bf if converges}, cannot converge to a {\bf generalised} saddle point. (iv) There is a set $\mathcal{E}_2\subset \mathbb{R}^k$ of Lebesgue measure $0$ so that for all $x_0\in \mathbb{R}^k\backslash \mathcal{E}_2$, any cluster point of the sequence $x_{n+1}=H(x_n)$ is not a saddle point, and more generally cannot be an isolated generalised saddle point. Some other results are proven.