Abstract--Robots in many real-world settings have access to force/torque sensors in their gripper and tactile sensing is often necessary in tasks that involve contact-rich motion. In this work, we leverage surprise from mismatches in touch feedback to guide exploration in hard sparse-reward reinforcement learning tasks. Our approach, Touch-based Curiosity (ToC), learns what visible objects interactions are supposed to "feel" like. We encourage exploration by rewarding interactions where the expectation and the experience don't match. In our proposed method, an initial task-independent exploration phase is followed by an on-task learning phase, in which the original interactions are relabeled with on-task rewards. We test our approach on a range of touchintensive robot arm tasks (e.g. In the former, the environment is often fully observable, and the reward is dense and well-defined. In the Recent works in RL have focused on curiosity-driven latter, a large amount of work is required to design useful exploration through prediction-based surprise [6, 45, 48]. While it may be possible to hand-craft dense formulation, a forward dynamics models predicts the future, and reward signals for many real-world tasks, we believe that it's if its prediction is incorrect when compared to the real future, a worthwhile endeavor to investigate learning methods that do the agent is surprised and is thus rewarded.

Modelers use automatic differentiation of computation graphs to implement complex Deep Learning models without defining gradient computations. However, modelers often use sampling methods to estimate intractable expectations such as in Reinforcement Learning and Variational Inference. Current methods for estimating gradients through these sampling steps are limited: They are either only applicable to continuous random variables and differentiable functions, or can only use simple but high variance score-function estimators. To overcome these limitations, we introduce Storchastic, a new framework for automatic differentiation of stochastic computation graphs. Storchastic allows the modeler to choose from a wide variety of gradient estimation methods at each sampling step, to optimally reduce the variance of the gradient estimates. Furthermore, Storchastic is provably unbiased for estimation of any-order gradients, and generalizes variance reduction techniques to higher-order gradient estimates. Finally, we implement Storchastic as a PyTorch library.

Poetry generation has long been a challenge for artificial intelligence. In the scope of Japanese poetry generation, many researchers have paid attention to Haiku generation, but few have focused on Waka generation. To further explore the creative potential of natural language generation systems in Japanese poetry creation, we propose a novel Waka generation model, WakaVT, which automatically produces Waka poems given user-specified keywords. Firstly, an additive mask-based approach is presented to satisfy the form constraint. Secondly, the structures of Transformer and variational autoencoder are integrated to enhance the quality of generated content. Specifically, to obtain novelty and diversity, WakaVT employs a sequence of latent variables, which effectively captures word-level variability in Waka data. To improve linguistic quality in terms of fluency, coherence, and meaningfulness, we further propose the fused multilevel self-attention mechanism, which properly models the hierarchical linguistic structure of Waka. To the best of our knowledge, we are the first to investigate Waka generation with models based on Transformer and/or variational autoencoder. Both objective and subjective evaluation results demonstrate that our model outperforms baselines significantly.

The Neural Tangent Kernel (NTK) characterizes the behavior of infinitely wide neural nets trained under least squares loss by gradient descent (Jacot et al., 2018). However, despite its importance, the super-quadratic runtime of kernel methods limits the use of NTK in large-scale learning tasks. To accelerate kernel machines with NTK, we propose a near input sparsity time algorithm that maps the input data to a randomized low-dimensional feature space so that the inner product of the transformed data approximates their NTK evaluation. Furthermore, we propose a feature map for approximating the convolutional counterpart of the NTK (Arora et al., 2019), which can transform any image using a runtime that is only linear in the number of pixels. We show that in standard large-scale regression and classification tasks a linear regressor trained on our features outperforms trained NNs and Nystrom method with NTK kernels.

Recently, bound propagation based certified adversarial defense have been proposed for training neural networks with certifiable robustness guarantees. Despite state-of-the-art (SOTA) methods including interval bound propagation (IBP) and CROWN-IBP have per-batch training complexity similar to standard neural network training, to reach SOTA performance they usually need a long warmup schedule with hundreds or thousands epochs and are thus still quite costly for training. In this paper, we discover that the weight initialization adopted by prior works, such as Xavier or orthogonal initialization, which was originally designed for standard network training, results in very loose certified bounds at initialization thus a longer warmup schedule must be used. We also find that IBP based training leads to a significant imbalance in ReLU activation states, which can hamper model performance. Based on our findings, we derive a new IBP initialization as well as principled regularizers during the warmup stage to stabilize certified bounds during initialization and warmup stage, which can significantly reduce the warmup schedule and improve the balance of ReLU activation states. Additionally, we find that batch normalization (BN) is a crucial architectural element to build best-performing networks for certified training, because it helps stabilize bound variance and balance ReLU activation states. With our proposed initialization, regularizers and architectural changes combined, we are able to obtain 65.03% verified error on CIFAR-10 ($\epsilon=\frac{8}{255}$) and 82.13% verified error on TinyImageNet ($\epsilon=\frac{1}{255}$) using very short training schedules (160 and 80 total epochs, respectively), outperforming literature SOTA trained with a few hundreds or thousands epochs.

Alphas are stock prediction models capturing trading signals in a stock market. A set of effective alphas can generate weakly correlated high returns to diversify the risk. Existing alphas can be categorized into two classes: Formulaic alphas are simple algebraic expressions of scalar features, and thus can generalize well and be mined into a weakly correlated set. Machine learning alphas are data-driven models over vector and matrix features. They are more predictive than formulaic alphas, but are too complex to mine into a weakly correlated set. In this paper, we introduce a new class of alphas to model scalar, vector, and matrix features which possess the strengths of these two existing classes. The new alphas predict returns with high accuracy and can be mined into a weakly correlated set. In addition, we propose a novel alpha mining framework based on AutoML, called AlphaEvolve, to generate the new alphas. To this end, we first propose operators for generating the new alphas and selectively injecting relational domain knowledge to model the relations between stocks. We then accelerate the alpha mining by proposing a pruning technique for redundant alphas. Experiments show that AlphaEvolve can evolve initial alphas into the new alphas with high returns and weak correlations.

Open intent classification is a challenging task in dialogue systems. On the one hand, it should ensure the quality of known intent identification. On the other hand, it needs to detect the open (unknown) intent without prior knowledge. Current models are limited in finding the appropriate decision boundary to balance the performances of both known intents and the open intent. In this paper, we propose a post-processing method to learn the adaptive decision boundary (ADB) for open intent classification. We first utilize the labeled known intent samples to pre-train the model. Then, we automatically learn the adaptive spherical decision boundary for each known class with the aid of well-trained features. Specifically, we propose a new loss function to balance both the empirical risk and the open space risk. Our method does not need open intent samples and is free from modifying the model architecture. Moreover, our approach is surprisingly insensitive with less labeled data and fewer known intents. Extensive experiments on three benchmark datasets show that our method yields significant improvements compared with the state-of-the-art methods. The codes are released at https://github.com/thuiar/Adaptive-Decision-Boundary.

Introduced in May of 2020, we've been waiting on edge for the release of GPT-3 for mass use. In the nearly ten months since its initial release, we are starting to see more startups with access to this groundbreaking technology. We've covered natural language processing through our BERT series, so it's all-the-more fitting that we tap into GPT-3. Who knows, maybe this will be part of our AI-writing series. But for now, here's the low-down on GPT-3 and when we expect it'll be ready for mass use.

With April 1 almost here, a computer researcher wanted to see if an AI could generate some knee-slapping April Fools' Day pranks. Social distancing is still in effect so Janelle Shane tasked the popular language modeling software GPT-3 to come up with gags people could play on themselves. One version of the neural net suggested sawing your mattress in half or standing in line for a movie for 30 minutes and then just walking away. Another suggested '[creating] a secret language that only you and your cat can understand.' 'Take a bunch of jellybeans,' it offered as an alternative, 'put them in a Ziplock bag and tie the bag around your neck.'

Finally, there is a programming environment called nbdev, which allows users to create complete Python packages. The Mid level APIs are a key differentiator for fast.ai