In neural network, synaptic strength could be seen as probability to transmit impulse, and function could exist from transmission probability to synaptic strength. If the function satisfies constraint such as continuity and monotonicity, neural network would always go to one unique fixed point. An biological image classifier is proposed to utilize this fixed point.

Transfer learning has become a cornerstone of computer vision with the advent of ImageNet features, yet little work has been done to evaluate the performance of ImageNet architectures across different datasets. An implicit hypothesis in modern computer vision research is that models that perform better on ImageNet necessarily perform better on other vision tasks. However, this hypothesis has never been systematically tested. Here, we compare the performance of 13 classification models on 12 image classification tasks in three settings: as fixed feature extractors, fine-tuned, and trained from random initialization. We find that, when networks are used as fixed feature extractors, ImageNet accuracy is only weakly predictive of accuracy on other tasks ($r^2=0.24$). In this setting, ResNets consistently outperform networks that achieve higher accuracy on ImageNet. When networks are fine-tuned, we observe a substantially stronger correlation ($r^2 = 0.86$). We achieve state-of-the-art performance on eight image classification tasks simply by fine-tuning state-of-the-art ImageNet architectures, outperforming previous results based on specialized methods for transfer learning. Finally, we observe that, on three small fine-grained image classification datasets, networks trained from random initialization perform similarly to ImageNet-pretrained networks. Together, our results show that ImageNet architectures generalize well across datasets, with small improvements in ImageNet accuracy producing improvements across other tasks, but ImageNet features are less general than previously suggested.

GANS are powerful generative models that are able to model the manifold of natural images. We leverage this property to perform manifold regularization by approximating the Laplacian norm using a Monte Carlo approximation that is easily computed with the GAN. When incorporated into the feature-matching GAN of Improved GAN, we achieve state-of-the-art results for GAN-based semi-supervised learning on the CIFAR-10 dataset, with a method that is significantly easier to implement than competing methods.

Spectral clustering is a celebrated algorithm that partitions objects based on pairwise similarity information. While this approach has been successfully applied to a variety of domains, it comes with limitations. The reason is that there are many other applications in which only \emph{multi}-way similarity measures are available. This motivates us to explore the multi-way measurement setting. In this work, we develop two algorithms intended for such setting: Hypergraph Spectral Clustering (HSC) and Hypergraph Spectral Clustering with Local Refinement (HSCLR). Our main contribution lies in performance analysis of the poly-time algorithms under a random hypergraph model, which we name the weighted stochastic block model, in which objects and multi-way measures are modeled as nodes and weights of hyperedges, respectively. Denoting by $n$ the number of nodes, our analysis reveals the following: (1) HSC outputs a partition which is better than a random guess if the sum of edge weights (to be explained later) is $\Omega(n)$; (2) HSC outputs a partition which coincides with the hidden partition except for a vanishing fraction of nodes if the sum of edge weights is $\omega(n)$; and (3) HSCLR exactly recovers the hidden partition if the sum of edge weights is on the order of $n \log n$. Our results improve upon the state of the arts recently established under the model and they firstly settle the order-wise optimal results for the binary edge weight case. Moreover, we show that our results lead to efficient sketching algorithms for subspace clustering, a computer vision application. Lastly, we show that HSCLR achieves the information-theoretic limits for a special yet practically relevant model, thereby showing no computational barrier for the case.

We present a novel approach for parallel computation in the context of machine learning that we call "Tell Me Something New" (TMSN). This approach involves a set of independent workers that use broadcast to update each other when they observe "something new". TMSN does not require synchronization or a head node and is highly resilient against failing machines or laggards. We demonstrate the utility of TMSN by applying it to learning boosted trees. We show that our implementation is 10 times faster than XGBoost and LightGBM on the splice-site prediction problem.

Markov chain algorithms are ubiquitous in machine learning and statistics and many other disciplines. In this work we present a novel estimator applicable to several classes of Markov chains, dubbed Markov chain importance sampling (MCIS). For a broad class of Metropolis-Hastings algorithms, MCIS efficiently makes use of rejected proposals. For discretized Langevin diffusions, it provides a novel way of correcting the discretization error. Our estimator satisfies a central limit theorem and improves on error per CPU cycle, often to a large extent. As a by-product it enables estimating the normalizing constant, an important quantity in Bayesian machine learning and statistics.

We study the classic $k$-means/median clustering, which are fundamental problems in unsupervised learning, in the setting where data are partitioned across multiple sites, and where we are allowed to discard a small portion of the data by labeling them as outliers. We propose a simple approach based on constructing small summary for the original dataset. The proposed method is time and communication efficient, has good approximation guarantees, and can identify the global outliers effectively. To the best of our knowledge, this is the first practical algorithm with theoretical guarantees for distributed clustering with outliers. Our experiments on both real and synthetic data have demonstrated the clear superiority of our algorithm against all the baseline algorithms in almost all metrics.

Most classifiers operate by selecting the maximum of an estimate of the conditional distribution $p(y|x)$ where $x$ stands for the features of the instance to be classified and $y$ denotes its label. This often results in a hubristic bias: overconfidence in the assignment of a definite label. Usually, the observations are concentrated on a small volume but the classifier provides definite predictions for the entire space. We propose constructing conformal prediction sets [vovk2005algorithmic] which contain a set of labels rather than a single label. These conformal prediction sets contain the true label with probability $1-\alpha$. Our construction is based on $p(x|y)$ rather than $p(y|x)$ which results in a classifier that is very cautious: it outputs the null set - meaning `I don't know' --- when the object does not resemble the training examples. An important property of our approach is that classes can be added or removed without having to retrain the classifier. We demonstrate the performance on the ImageNet ILSVRC dataset using high dimensional features obtained from state of the art convolutional neural networks.

In this paper, we propose a new adaptive stochastic gradient Langevin dynamics (ASGLD) algorithmic framework and its two specialized versions, namely adaptive stochastic gradient (ASG) and adaptive gradient Langevin dynamics(AGLD), for non-convex optimization problems. All proposed algorithms can escape from saddle points with at most $O(\log d)$ iterations, which is nearly dimension-free. Further, we show that ASGLD and ASG converge to a local minimum with at most $O(\log d/\epsilon^4)$ iterations. Also, ASGLD with full gradients or ASGLD with a slowly linearly increasing batch size converge to a local minimum with iterations bounded by $O(\log d/\epsilon^2)$, which outperforms existing first-order methods.

Particle filters sequentially approximate posterior distributions by sampling representative points and updating them independently. The idea is applied in various domains, e.g. reasoning with uncertainty in robotics. A remaining challenge is constructing probabilistic models of the system, which can be especially hard for complex sensors, e.g. a camera. We introduce the Particle Filter Networks (PF-nets) that encode both a learned probabilistic system model and the particle filter algorithm in a single neural network architecture. The unified representation allows learning models end-to-end, circumventing the difficulties of conventional model-based methods. We applied PF-nets to a challenging visual localization task that requires matching visual features from camera images with the geometry encoded in a 2-D floor map. In preliminary experiments end-to-end PF-nets consistently outperformed alternative learning architectures, as well as conventional model-based methods.