Jasmine Collins

Aaron Gokaslan, A. Feder Cooper, Jasmine Collins et al.

We assemble a dataset of Creative-Commons-licensed (CC) images, which we use to train a set of open diffusion models that are qualitatively competitive with Stable Diffusion 2 (SD2). This task presents two challenges: (1) high-resolution CC images lack the captions necessary to train text-to-image generative models; (2) CC images are relatively scarce. In turn, to address these challenges, we use an intuitive transfer learning technique to produce a set of high-quality synthetic captions paired with curated CC images. We then develop a data- and compute-efficient training recipe that requires as little as 3% of the LAION-2B data needed to train existing SD2 models, but obtains comparable quality. These results indicate that we have a sufficient number of CC images (~70 million) for training high-quality models. Our training recipe also implements a variety of optimizations that achieve ~3X training speed-ups, enabling rapid model iteration. We leverage this recipe to train several high-quality text-to-image models, which we dub the CommonCanvas family. Our largest model achieves comparable performance to SD2 on a human evaluation, despite being trained on our CC dataset that is significantly smaller than LAION and using synthetic captions for training. We release our models, data, and code at https://github.com/mosaicml/diffusion/blob/main/assets/common-canvas.md

Rishab Parthasarathy, Jasmine Collins, Cory Stephenson

Automated evaluation of generative text-to-image models remains a challenging problem. Recent works have proposed using multimodal LLMs to judge the quality of images, but these works offer little insight into how multimodal LLMs make use of concepts relevant to humans, such as image style or composition, to generate their overall assessment. In this work, we study what attributes of an image--specifically aesthetics, lack of artifacts, anatomical accuracy, compositional correctness, object adherence, and style--are important for both LLMs and humans to make judgments on image quality. We first curate a dataset of human preferences using synthetically generated image pairs. We use inter-task correlation between each pair of image quality attributes to understand which attributes are related in making human judgments. Repeating the same analysis with LLMs, we find that the relationships between image quality attributes are much weaker. Finally, we study individual image quality attributes by generating synthetic datasets with a high degree of control for each axis. Humans are able to easily judge the quality of an image with respect to all of the specific image quality attributes (e.g. high vs. low aesthetic image), however we find that some attributes, such as anatomical accuracy, are much more difficult for multimodal LLMs to learn to judge. Taken together, these findings reveal interesting differences between how humans and multimodal LLMs perceive images.

Eliza Kosoy, Jasmine Collins, David M. Chan et al.

Research in developmental psychology consistently shows that children explore the world thoroughly and efficiently and that this exploration allows them to learn. In turn, this early learning supports more robust generalization and intelligent behavior later in life. While much work has gone into developing methods for exploration in machine learning, artificial agents have not yet reached the high standard set by their human counterparts. In this work we propose using DeepMind Lab (Beattie et al., 2016) as a platform to directly compare child and agent behaviors and to develop new exploration techniques. We outline two ongoing experiments to demonstrate the effectiveness of a direct comparison, and outline a number of open research questions that we believe can be tested using this methodology.

Jasmine Collins, Johannes Balle, Jonathon Shlens

A significant advance in accelerating neural network training has been the development of normalization methods, permitting the training of deep models both faster and with better accuracy. These advances come with practical challenges: for instance, batch normalization ties the prediction of individual examples with other examples within a batch, resulting in a network that is heavily dependent on batch size. Layer normalization and group normalization are data-dependent and thus must be continually used, even at test-time. To address the issues that arise from using explicit normalization techniques, we propose to replace existing normalization methods with a simple, secondary objective loss that we term a standardization loss. This formulation is flexible and robust across different batch sizes and surprisingly, this secondary objective accelerates learning on the primary training objective. Because it is a training loss, it is simply removed at test-time, and no further effort is needed to maintain normalized activations. We find that a standardization loss accelerates training on both small- and large-scale image classification experiments, works with a variety of architectures, and is largely robust to training across different batch sizes.

Jasmine Collins, Jascha Sohl-Dickstein, David Sussillo

Two potential bottlenecks on the expressiveness of recurrent neural networks (RNNs) are their ability to store information about the task in their parameters, and to store information about the input history in their units. We show experimentally that all common RNN architectures achieve nearly the same per-task and per-unit capacity bounds with careful training, for a variety of tasks and stacking depths. They can store an amount of task information which is linear in the number of parameters, and is approximately 5 bits per parameter. They can additionally store approximately one real number from their input history per hidden unit. We further find that for several tasks it is the per-task parameter capacity bound that determines performance. These results suggest that many previous results comparing RNN architectures are driven primarily by differences in training effectiveness, rather than differences in capacity. Supporting this observation, we compare training difficulty for several architectures, and show that vanilla RNNs are far more difficult to train, yet have slightly higher capacity. Finally, we propose two novel RNN architectures, one of which is easier to train than the LSTM or GRU for deeply stacked architectures.

Akosua Busia, Jasmine Collins, Navdeep Jaitly

Recently developed deep learning techniques have significantly improved the accuracy of various speech and image recognition systems. In this paper we adapt some of these techniques for protein secondary structure prediction. We first train a series of deep neural networks to predict eight-class secondary structure labels given a protein's amino acid sequence information and find that using recent methods for regularization, such as dropout and weight-norm constraining, leads to measurable gains in accuracy. We then adapt recent convolutional neural network architectures--Inception, ReSNet, and DenseNet with Batch Normalization--to the problem of protein structure prediction. These convolutional architectures make heavy use of multi-scale filter layers that simultaneously compute features on several scales, and use residual connections to prevent underfitting. Using a carefully modified version of these architectures, we achieve state-of-the-art performance of 70.0% per amino acid accuracy on the public CB513 benchmark dataset. Finally, we explore additions from sequence-to-sequence learning, altering the model to make its predictions conditioned on both the protein's amino acid sequence and its past secondary structure labels. We introduce a new method of ensembling such a conditional model with our convolutional model, an approach which reaches 70.6% Q8 accuracy on CB513. We argue that these results can be further refined for larger boosts in prediction accuracy through more sophisticated attempts to control overfitting of conditional models. We aim to release the code for these experiments as part of the TensorFlow repository.