We present theory of synaptic neural balance and we show experimentally that synaptic neural balance can improve deep learning speed, and accuracy, even in data-scarce environments. Given an additive cost function (regularizer) of the synaptic weights, a neuron is said to be in balance if the total cost of its incoming weights is equal to the total cost of its outgoing weights. For large classes of networks, activation functions, and regularizers, neurons can be balanced fully or partially using scaling operations that do not change their functionality. Furthermore, these balancing operations are associated with a strictly convex optimization problem with a single optimum and can be carried out in any order. In our simulations, we systematically observe that: (1) Fully balancing before training results in better performance as compared to several other training approaches; (2) Interleaving partial (layer-wise) balancing and stochastic gradient descent steps during training results in faster learning convergence and better overall accuracy (with L1 balancing converging faster than L2 balancing); and (3) When given limited training data, neural balanced models outperform plain or regularized models; and this is observed in both feedforward and recurrent networks. In short, the evidence supports that neural balancing operations could be added to the arsenal of methods used to regularize and train neural networks. Furthermore, balancing operations are entirely local and can be carried out asynchronously, making them plausible for biological or neuromorphic systems.

We develop a general theory of synaptic neural balance and how it can emerge or be enforced in neural networks . For a given additive cost function R (regularizer), a neuron is said to be in balance if the total cost of its input weights is equal to the total cost of its output weights. The basic example is provided by feedforward networks of ReLU units trained with L 2 regularizers, which exhibit balance after proper training. The theory explains this phenomenon and extends it in several directions. The first direction is the extension to bilinear and other activation functions . The second direction is the extension to more general regularizers, including all L p ( p > 0 ) regularizers. The third direction is the extension to non-layered architectures, recurrent architectures, convolutional architectures, as well as architectures with mixed activation functions and to different balancing algorithms. Gradient descent on the error function alone does not converge in general to a balanced state, where every neuron is in balance, even when starting from a balanced state. However, gradient descent on the regularized error function ought to converge to a balanced state, and thus network balance can be used to assess learning progress. The theory is based on two local neuronal operations: scaling which is commutative, and balancing which is not commutative. Finally, and most importantly, given any set of weights, when local balancing operations are applied to each neuron in a stochastic manner, global order always emerges through the convergence of the stochastic balancing algorithm to the same unique set of balanced weights. The reason for this convergence is the existence of an underlying strictly convex optimization problem where the relevant variables are constrained to a linear, only architecture-dependent, manifold. Simulations show that balancing neurons prior to learning, or during learning in alternation with gradient descent steps, can improve learning speed and performance thereby expanding the arsenal of available training tools. Scaling and balancing operations are entirely local and thus physically plausible in biological and neuromorphic neural networks .

In the Inertial Confinement Fusion (ICF) process, roughly a 2mm spherical shell made of high-density carbon is used as a target for laser beams, which compress and heat it to energy levels needed for high fusion yield in nuclear fusion. These shells are polished meticulously to meet the standards for a fusion shot. However, the polishing of these shells involves multiple stages, with each stage taking several hours. To make sure that the polishing process is advancing in the right direction, we are able to measure the shell surface roughness. This measurement, however, is very labor-intensive, time-consuming, and requires a human operator. To help improve the polishing process we have released the first dataset to the public that consists of raw vibration signals with the corresponding polishing surface roughness changes. We show that this dataset can be used with a variety of neural network based methods for prediction of the change of polishing surface roughness, hence eliminating the need for the time-consuming manual process. This is the first dataset of its kind to be released in public and its use will allow the operator to make any necessary changes to the ICF polishing process for optimal results. This dataset contains the raw vibration data of multiple polishing runs with their extracted statistical features and the corresponding surface roughness values. Additionally, to generalize the prediction models to different polishing conditions, we also apply domain adaptation techniques to improve prediction accuracy for conditions unseen by the trained model. The dataset is available in \url{https://junzeliu.github.io/Diamond-Polishing-Dataset/}.

Stochastic gradient Markov Chain Monte Carlo (SGMCMC) is a scalable algorithm for asymptotically exact Bayesian inference in parameter-rich models, such as Bayesian neural networks. However, since mixing can be slow in high dimensions, practitioners often resort to variational inference (VI). Unfortunately, VI makes strong assumptions on both the factorization and functional form of the posterior. To relax these assumptions, this work proposes a new non-parametric variational inference scheme that combines ideas from both SGMCMC and coordinate-ascent VI. The approach relies on a new Langevin-type algorithm that operates on a "self-averaged" posterior energy function, where parts of the latent variables are averaged over samples from earlier iterations of the Markov chain. This way, statistical dependencies between coordinates can be broken in a controlled way, allowing the chain to mix faster. This scheme can be further modified in a "dropout" manner, leading to even more scalability. We test our scheme for ResNet-20 on CIFAR-10, SVHN, and FMNIST. In all cases, we find improvements in convergence speed and/or final accuracy compared to SGMCMC and parametric VI.

This work addresses adversarial robustness in deep learning by considering deep networks with stochastic local winner-takes-all (LWTA) activations. This type of network units result in sparse representations from each model layer, as the units are organized in blocks where only one unit generates a non-zero output. The main operating principle of the introduced units lies on stochastic arguments, as the network performs posterior sampling over competing units to select the winner. We combine these LWTA arguments with tools from the field of Bayesian non-parametrics, specifically the stick-breaking construction of the Indian Buffet Process, to allow for inferring the sub-part of each layer that is essential for modeling the data at hand. Then, inference is performed by means of stochastic variational Bayes. We perform a thorough experimental evaluation of our model using benchmark datasets. As we show, our method achieves high robustness to adversarial perturbations, with state-of-the-art performance in powerful adversarial attack schemes.