Our lower bound analysis shows that the sample complexities ofBSGD cannot be improved for general convexobjectives and nonconvexobjectivesexcept for smooth nonconvexobjectiveswith Lipschitz continuous gradient estimator.

Conditional stochastic optimization covers a variety of applications ranging from invariant learning and causal inference to meta-learning. However, constructing unbiased gradient estimators for such problems is challenging due to the composition structure. As an alternative, we propose a biased stochastic gradient descent (BSGD) algorithm and study the bias-variance tradeoff under different structural assumptions. We establish the sample complexities of BSGD for strongly convex, convex, and weakly convex objectives under smooth and non-smooth conditions. Our lower bound analysis shows that the sample complexities of BSGD cannot be improved for general convex objectives and nonconvex objectives except for smooth nonconvex objectives with Lipschitz continuous gradient estimator. For this special setting, we propose an accelerated algorithm called biased SpiderBoost (BSpiderBoost) that matches the lower bound complexity. We further conduct numerical experiments on invariant logistic regression and model-agnostic meta-learning to illustrate the performance of BSGD and BSpiderBoost.

Our lower bound analysis shows that the sample complexities of BSGD cannot be improved for general convex objectives and nonconvex objectives except for smooth nonconvex objectives with Lipschitz continuous gradient estimator.

But this view of differentiable learning ignores two things.

But this view of differentiable learning ignores two things.

Strengths: To the best of my knowledge, the BSGD algorithm is the first stochastic-gradient based algorithm that directly solves CSO problem itself. The two most relevant work that focus on CSO are [12] and [24]; [12] solves a saddle-point problem reformulation of CSO, while [24] resorts to providing sample complexities for SAA approach to solve general CSO problem. With respect to the SAA approach presented in [24], BSGD method improves in sample complexities (they remove the dependence on d) when F is general convex, matching the lower bounds they provide. Although BSGD is not optimal when F is strongly convex and smooth, it matches the complexities of SAA approach[24]. They also argue about the settings in which BSGD may not be optimal, providing a transparent evaluation of their algorithm.

We study the power of learning via mini-batch stochastic gradient descent (SGD) on the population loss, and batch Gradient Descent (GD) on the empirical loss, of a differentiable model or neural network, and ask what learning problems can be learnt using these paradigms. We show that SGD and GD can always simulate learning with statistical queries (SQ), but their ability to go beyond that depends on the precision $\rho$ of the gradient calculations relative to the minibatch size $b$ (for SGD) and sample size $m$ (for GD). With fine enough precision relative to minibatch size, namely when $b \rho$ is small enough, SGD can go beyond SQ learning and simulate any sample-based learning algorithm and thus its learning power is equivalent to that of PAC learning; this extends prior work that achieved this result for $b=1$. Similarly, with fine enough precision relative to the sample size $m$, GD can also simulate any sample-based learning algorithm based on $m$ samples. In particular, with polynomially many bits of precision (i.e. when $\rho$ is exponentially small), SGD and GD can both simulate PAC learning regardless of the mini-batch size. On the other hand, when $b \rho^2$ is large enough, the power of SGD is equivalent to that of SQ learning.

Finding methods for making generalizable predictions is a fundamental problem of machine learning. By looking into similarities between the prediction problem for unknown data and the lossless compression we have found an approach that gives a solution. In this paper we propose a compression principle that states that an optimal predictive model is the one that minimizes a total compressed message length of all data and model definition while guarantees decodability. Following the compression principle we use Bayesian approach to build probabilistic models of data and network definitions. A method to approximate Bayesian integrals using a sequence of variational approximations is implemented as an optimizer for hyper-parameters: Bayesian Stochastic Gradient Descent (BSGD). Training with BSGD is completely defined by setting only three parameters: number of epochs, the size of the dataset and the size of the minibatch, which define a learning rate and a number of iterations. We show that dropout can be used for a continuous dimensionality reduction that allows to find optimal network dimensions as required by the compression principle.