We consider active learning of deep neural networks. Most active learning works in this context have focused on studying effective querying mechanisms and assumed that an appropriate network architecture is a priori known for the problem at hand. We challenge this assumption and propose a novel active strategy whereby the learning algorithm searches for effective architectures on the fly, while actively learning. We apply our strategy using three known querying techniques (softmax response, MC-dropout, and coresets) and show that the proposed approach overwhelmingly outperforms active learning using fixed architectures.

For reliable transmission across a noisy communication channel, classical results from information theory show that it is asymptotically optimal to separate out the source and channel coding processes. However, this decomposition can fall short in the finite bit-length regime, as it requires non-trivial tuning of hand-crafted codes and assumes infinite computational power for decoding. In this work, we propose Neural Error Correcting and Source Trimming (\modelname) codes to jointly learn the encoding and decoding processes in an end-to-end fashion. By adding noise into the latent codes to simulate the channel during training, we learn to both compress and error-correct given a fixed bit-length and computational budget. We obtain codes that are not only competitive against several capacity-approaching channel codes, but also learn useful robust representations of the data for downstream tasks such as classification. Finally, we learn an extremely fast neural decoder, yielding almost an order of magnitude in speedup compared to standard decoding methods based on iterative belief propagation.

In recent years, three-dimensional convolutional neural network (3D CNN) is intensively applied in video analysis and receives good performance. However, 3D CNN leads to massive computation and storage consumption, which hinders its deployment on mobile and embedded devices. In this paper, we propose a three-dimensional regularization-based pruning method to assign different regularization parameters to different weight groups based on their importance to the network. Experiments show that the proposed method outperforms other popular methods in this area.

We introduce a theory-driven mechanism for learning a neural network model that performs generative topology design in one shot given a problem setting, circumventing the conventional iterative procedure that computational design tasks usually entail. The proposed mechanism can lead to machines that quickly response to new design requirements based on its knowledge accumulated through past experiences of design generation. Achieving such a mechanism through supervised learning would require an impractically large amount of problem-solution pairs for training, due to the known limitation of deep neural networks in knowledge generalization. To this end, we introduce an interaction between a student (the neural network) and a teacher (the optimality conditions underlying topology optimization): The student learns from existing data and is tested on unseen problems. Deviation of the student's solutions from the optimality conditions is quantified, and used to choose new data points for the student to learn from. We show through a compliance minimization problem that the proposed learning mechanism is significantly more data efficient than using a static dataset under the same computational budget.

In this paper, we propose a new method called ProfWeight for transferring information from a pre-trained deep neural network that has a high test accuracy to a simpler interpretable model or a very shallow network of low complexity and a priori low test accuracy. We are motivated by applications in interpretability and model deployment in severely memory constrained environments (like sensors). Our method uses linear probes to generate confidence scores through flattened intermediate representations. Our transfer method involves a theoretically justified weighting of samples during the training of the simple model using confidence scores of these intermediate layers. The value of our method is first demonstrated on CIFAR-10, where our weighting method significantly improves (3-4%) networks with only a fraction of the number of Resnet blocks of a complex Resnet model. We further demonstrate operationally significant results on a real manufacturing problem, where we dramatically increase the test accuracy of a CART model (the domain standard) by roughly 13%.

Person re-identification (ReID) is aimed at identifying the same person across videos captured from different cameras. In the view that networks extracting global features using ordinary network architectures are difficult to extract local features due to their weak attention mechanisms, researchers have proposed a lot of elaborately designed ReID networks, while greatly improving the accuracy, the model size and the feature extraction latency are also soaring. We argue that a relatively compact ordinary network extracting globally pooled features has the capability to extract discriminative local features and can achieve state-of-the-art precision if only the model's parameters are properly learnt. In order to reduce the difficulty in learning hard identity labels, we propose a novel knowledge distillation method: Factorized Distillation, which factorizes both feature maps and retrieval features of holistic ReID network to mimic representations of multiple partial ReID models, thus transferring the knowledge from partial ReID models to the holistic network. Experiments show that the performance of model trained with the proposed method can outperform state-of-the-art with relatively few network parameters.

Evolutionary deep intelligence has recently shown great promise for producing small, powerful deep neural network models via the organic synthesis of increasingly efficient architectures over successive generations. Existing evolutionary synthesis processes, however, have allowed the mating of parent networks independent of architectural alignment, resulting in a mismatch of network structures. We present a preliminary study into the effects of architectural alignment during evolutionary synthesis using a gene tagging system. Surprisingly, the network architectures synthesized using the gene tagging approach resulted in slower decreases in performance accuracy and storage size; however, the resultant networks were comparable in size and performance accuracy to the non-gene tagging networks. Furthermore, we speculate that there is a noticeable decrease in network variability for networks synthesized with gene tagging, indicating that enforcing a like-with-like mating policy potentially restricts the exploration of the search space of possible network architectures.

Behavioral skills or policies for autonomous agents are conventionally learned from reward functions, via reinforcement learning, or from demonstrations, via imitation learning. However, both modes of task specification have their disadvantages: reward functions require manual engineering, while demonstrations require a human expert to be able to actually perform the task in order to generate the demonstration. Instruction following from natural language instructions provides an appealing alternative: in the same way that we can specify goals to other humans simply by speaking or writing, we would like to be able to specify tasks for our machines. However, a single instruction may be insufficient to fully communicate our intent or, even if it is, may be insufficient for an autonomous agent to actually understand how to perform the desired task. In this work, we propose an interactive formulation of the task specification problem, where iterative language corrections are provided to an autonomous agent, guiding it in acquiring the desired skill. Our proposed language-guided policy learning algorithm can integrate an instruction and a sequence of corrections to acquire new skills very quickly. In our experiments, we show that this method can enable a policy to follow instructions and corrections for simulated navigation and manipulation tasks, substantially outperforming direct, non-interactive instruction following.

One obstacle to applying reinforcement learning algorithms to real-world problems is the lack of suitable reward functions. Designing such reward functions is difficult in part because the user only has an implicit understanding of the task objective. This gives rise to the agent alignment problem: how do we create agents that behave in accordance with the user's intentions? We outline a high-level research direction to solve the agent alignment problem centered around reward modeling: learning a reward function from interaction with the user and optimizing the learned reward function with reinforcement learning. We discuss the key challenges we expect to face when scaling reward modeling to complex and general domains, concrete approaches to mitigate these challenges, and ways to establish trust in the resulting agents.

Yes, they do. This work investigates a perspective for deep learning: whether different normalization layers in a ConvNet require different normalizers. This is the first step towards understanding this phenomenon. We allow each convolutional layer to be stacked before a switchable normalization (SN) that learns to choose a normalizer from a pool of normalization methods. Through systematic experiments in ImageNet, COCO, Cityscapes, and ADE20K, we answer three questions: (a) Is it useful to allow each normalization layer to select its own normalizer? (b) What impacts the choices of normalizers? (c) Do different tasks and datasets prefer different normalizers? Our results suggest that (1) using distinct normalizers improves both learning and generalization of a ConvNet; (2) the choices of normalizers are more related to depth and batch size, but less relevant to parameter initialization, learning rate decay, and solver; (3) different tasks and datasets have different behaviors when learning to select normalizers.