Interesting theoretical associations have been established by recent papers\nbetween the fields of active learning and stochastic convex optimization due to\nthe common role of feedback in sequential querying mechanisms. In this paper,\nwe continue this thread in two parts by exploiting these relations for the\nfirst time to yield novel algorithms in both fields, further motivating the\nstudy of their intersection. First, inspired by a recent optimization algorithm\nthat was adaptive to unknown uniform convexity parameters, we present a new\nactive learning algorithm for one-dimensional thresholds that can yield minimax\nrates by adapting to unknown noise parameters. Next, we show that one can\nperform $d$-dimensional stochastic minimization of smooth uniformly convex\nfunctions when only granted oracle access to noisy gradient signs along any\ncoordinate instead of real-valued gradients, by using a simple randomized\ncoordinate descent procedure where each line search can be solved by\n$1$-dimensional active learning, provably achieving the same error convergence\nrate as having the entire real-valued gradient. Combining these two parts\nyields an algorithm that solves stochastic convex optimization of uniformly\nconvex and smooth functions using only noisy gradient signs by repeatedly\nperforming active learning, achieves optimal rates and is adaptive to all\nunknown convexity and smoothness parameters.\n