Posts in category: Google AI

Posted by Amir Yazdanbakhsh, Research Scientist, and Vijay Janapa Reddi, Visiting Researcher, Google Research

Computer Architecture research has a long history of developing simulators and tools to evaluate and shape the design of computer systems. For example, the SimpleScalar simulator was introduced in the late 1990s and allowed researchers to explore various microarchitectural ideas. Computer architecture simulators and tools, such as gem5, DRAMSys, and many more have played a significant role in advancing computer architecture research. Since then, these shared resources and infrastructure have benefited industry and academia and have enabled researchers to systematically build on each other’s work, leading to significant advances in the field.

Nonetheless, computer architecture research is evolving, with industry and academia turning towards machine learning (ML) optimization to meet stringent domain-specific requirements, such as ML for computer architecture, ML for TinyML acceleration, DNN accelerator datapath, memory controllers, power consumption, security, and privacy. Although prior work has demonstrated the benefits of ML in design optimization, the lack of strong, reproducible baselines hinders fair and objective comparison across different methods and poses several challenges to their deployment. To ensure steady progress, it is imperative to understand and tackle these challenges collectively.

To alleviate these challenges, in “ArchGym: An Open-Source Gymnasium for Machine Learning Assisted Architecture Design”, accepted at ISCA 2023, we introduced ArchGym, which includes a variety of computer architecture simulators and ML algorithms. Enabled by ArchGym, our results indicate that with a sufficiently large number of samples, any of a diverse collection of ML algorithms are capable of finding the optimal set of architecture design parameters for each target problem; no one solution is necessarily better than another. These results further indicate that selecting the optimal hyperparameters for a given ML algorithm is essential for finding the optimal architecture design, but choosing them is non-trivial. We release the code and dataset across multiple computer architecture simulations and ML algorithms.

Challenges in ML-assisted architecture research

ML-assisted architecture research poses several challenges, including:

1. For a specific ML-assisted computer architecture problem (e.g., finding an optimal solution for a DRAM controller) there is no systematic way to identify optimal ML algorithms or hyperparameters (e.g., learning rate, warm-up steps, etc.). There is a wider range of ML and heuristic methods, from random walk to reinforcement learning (RL), that can be employed for design space exploration (DSE). While these methods have shown noticeable performance improvement over their choice of baselines, it is not evident whether the improvements are because of the choice of optimization algorithms or hyperparameters.

   Thus, to ensure reproducibility and facilitate widespread adoption of ML-aided architecture DSE, it is necessary to outline a systematic benchmarking methodology.

2. While computer architecture simulators have been the backbone of architectural innovations, there is an emerging need to address the trade-offs between accuracy, speed, and cost in architecture exploration. The accuracy and speed of performance estimation widely varies from one simulator to another, depending on the underlying modeling details (e.g., cycle–accurate vs. ML–based proxy models). While analytical or ML-based proxy models are nimble by virtue of discarding low-level details, they generally suffer from high prediction error. Also, due to commercial licensing, there can be strict limits on the number of runs collected from a simulator. Overall, these constraints exhibit distinct performance vs. sample efficiency trade-offs, affecting the choice of optimization algorithm for architecture exploration.

   It is challenging to delineate how to systematically compare the effectiveness of various ML algorithms under these constraints.

3. Finally, the landscape of ML algorithms is rapidly evolving and some ML algorithms need data to be useful. Additionally, rendering the outcome of DSE into meaningful artifacts such as datasets is critical for drawing insights about the design space.

   In this rapidly evolving ecosystem, it is consequential to ensure how to amortize the overhead of search algorithms for architecture exploration. It is not apparent, nor systematically studied how to leverage exploration data while being agnostic to the underlying search algorithm.

ArchGym design

ArchGym addresses these challenges by providing a unified framework for evaluating different ML-based search algorithms fairly. It comprises two main components: 1) the ArchGym environment and 2) the ArchGym agent. The environment is an encapsulation of the architecture cost model — which includes latency, throughput, area, energy, etc., to determine the computational cost of running the workload, given a set of architectural parameters — paired with the target workload(s). The ArchGym agent is an encapsulation of the ML algorithm used for the search and consists of hyperparameters and a guiding policy. The hyperparameters are intrinsic to the algorithm for which the model is to be optimized and can significantly influence performance. The policy, on the other hand, determines how the agent selects a parameter iteratively to optimize the target objective.

Notably, ArchGym also includes a standardized interface that connects these two components, while also saving the exploration data as the ArchGym Dataset. At its core, the interface entails three main signals: hardware state, hardware parameters, and metrics. These signals are the bare minimum to establish a meaningful communication channel between the environment and the agent. Using these signals, the agent observes the state of the hardware and suggests a set of hardware parameters to iteratively optimize a (user-defined) reward. The reward is a function of hardware performance metrics, such as performance, energy consumption, etc. 

ArchGym comprises two main components: the ArchGym environment and the ArchGym agent. The ArchGym environment encapsulates the cost model and the agent is an abstraction of a policy and hyperparameters. With a standardized interface that connects these two components, ArchGym provides a unified framework for evaluating different ML-based search algorithms fairly while also saving the exploration data as the ArchGym Dataset.

ML algorithms could be equally favorable to meet user-defined target specifications

Using ArchGym, we empirically demonstrate that across different optimization objectives and DSE problems, at least one set of hyperparameters exists that results in the same hardware performance as other ML algorithms. A poorly selected (random selection) hyperparameter for the ML algorithm or its baseline can lead to a misleading conclusion that a particular family of ML algorithms is better than another. We show that with sufficient hyperparameter tuning, different search algorithms, even random walk (RW), are able to identify the best possible normalized reward. However, note that finding the right set of hyperparameters may require exhaustive search or even luck to make it competitive.

With a sufficient number of samples, there exists at least one set of hyperparameters that results in the same performance across a range of search algorithms. Here the dashed line represents the maximum normalized reward. Cloud-1, cloud-2, stream, and random indicate four different memory traces for DRAMSys (DRAM subsystem design space exploration framework).

Dataset construction and high-fidelity proxy model training

Creating a unified interface using ArchGym also enables the creation of datasets that can be used to design better data-driven ML-based proxy architecture cost models to improve the speed of architecture simulation. To evaluate the benefits of datasets in building an ML model to approximate architecture cost, we leverage ArchGym’s ability to log the data from each run from DRAMSys to create four dataset variants, each with a different number of data points. For each variant, we create two categories: (a) Diverse Dataset (DD), which represents the data collected from different agents (ACO, GA, RW, and BO), and (b) ACO only, which shows the data collected exclusively from the ACO agent, both of which are released along with ArchGym. We train a proxy model on each dataset using random forest regression with the objective to predict the latency of designs for a DRAM simulator. Our results show that:

1. As we increase the dataset size, the average normalized root mean squared error (RMSE) slightly decreases.
2. However, as we introduce diversity in the dataset (e.g., collecting data from different agents), we observe 9× to 42× lower RMSE across different dataset sizes.

Diverse dataset collection across different agents using ArchGym interface.

The impact of a diverse dataset and dataset size on the normalized RMSE.

The need for a community-driven ecosystem for ML-assisted architecture research

While, ArchGym is an initial effort towards creating an open-source ecosystem that (1) connects a broad range of search algorithms to computer architecture simulators in an unified and easy-to-extend manner, (2) facilitates research in ML-assisted computer architecture, and (3) forms the scaffold to develop reproducible baselines, there are a lot of open challenges that need community-wide support. Below we outline some of the open challenges in ML-assisted architecture design. Addressing these challenges requires a well coordinated effort and a community driven ecosystem.

Key challenges in ML-assisted architecture design.

We call this ecosystem Architecture 2.0. We outline the key challenges and a vision for building an inclusive ecosystem of interdisciplinary researchers to tackle the long-standing open problems in applying ML for computer architecture research. If you are interested in helping shape this ecosystem, please fill out the interest survey.

Conclusion

ArchGym is an open source gymnasium for ML architecture DSE and enables an standardized interface that can be readily extended to suit different use cases. Additionally, ArchGym enables fair and reproducible comparison between different ML algorithms and helps to establish stronger baselines for computer architecture research problems.

We invite the computer architecture community as well as the ML community to actively participate in the development of ArchGym. We believe that the creation of a gymnasium-type environment for computer architecture research would be a significant step forward in the field and provide a platform for researchers to use ML to accelerate research and lead to new and innovative designs.

Acknowledgements

This blogpost is based on joint work with several co-authors at Google and Harvard University. We would like to acknowledge and highlight Srivatsan Krishnan (Harvard) who contributed several ideas to this project in collaboration with Shvetank Prakash (Harvard), Jason Jabbour (Harvard), Ikechukwu Uchendu (Harvard), Susobhan Ghosh (Harvard), Behzad Boroujerdian (Harvard), Daniel Richins (Harvard), Devashree Tripathy (Harvard), and Thierry Thambe (Harvard).  In addition, we would also like to thank James Laudon, Douglas Eck, Cliff Young, and Aleksandra Faust for their support, feedback, and motivation for this work. We would also like to thank John Guilyard for the animated figure used in this post. Amir Yazdanbakhsh is now a Research Scientist at Google DeepMind and Vijay Janapa Reddi is an Associate Professor at Harvard.

Read More

Posted by Malaya Jules, Program Manager, Google

This week, the 61st annual meeting of the Association for Computational Linguistics (ACL), a premier conference covering a broad spectrum of research areas that are concerned with computational approaches to natural language, is taking place online.

As a leader in natural language processing and understanding, and a Diamond Level sponsor of ACL 2023, Google will showcase the latest research in the field with over 50 publications, and active involvement in a variety of workshops and tutorials.

If you’re registered for ACL 2023, we hope that you’ll visit the Google booth to learn more about the projects at Google that go into solving interesting problems for billions of people. You can also learn more about Google’s participation below (Google affiliations in bold).

Board and Organizing Committee

Spotlight papers

Papers

Industry Track papers

ACL Findings papers

Tutorials

Workshops

* Work done while at Google

Read More

Posted by Sanjay Subramanian, PhD student, UC Berkeley, and Arsha Nagrani, Research Scientist, Google Research, Perception Team

Visual question answering (VQA) is a machine learning task that requires a model to answer a question about an image or a set of images. Conventional VQA approaches need a large amount of labeled training data consisting of thousands of human-annotated question-answer pairs associated with images. In recent years, advances in large-scale pre-training have led to the development of VQA methods that perform well with fewer than fifty training examples (few-shot) and without any human-annotated VQA training data (zero-shot). However, there is still a significant performance gap between these methods and state-of-the-art fully supervised VQA methods, such as MaMMUT and VinVL. In particular, few-shot methods struggle with spatial reasoning, counting, and multi-hop reasoning. Furthermore, few-shot methods have generally been limited to answering questions about single images.

To improve accuracy on VQA examples that involve complex reasoning, in “Modular Visual Question Answering via Code Generation,” to appear at ACL 2023, we introduce CodeVQA, a framework that answers visual questions using program synthesis. Specifically, when given a question about an image or set of images, CodeVQA generates a Python program (code) with simple visual functions that allow it to process images, and executes this program to determine the answer. We demonstrate that in the few-shot setting, CodeVQA outperforms prior work by roughly 3% on the COVR dataset and 2% on the GQA dataset.

CodeVQA

The CodeVQA approach uses a code-writing large language model (LLM), such as PALM, to generate Python programs (code). We guide the LLM to correctly use visual functions by crafting a prompt consisting of a description of these functions and fewer than fifteen “in-context” examples of visual questions paired with the associated Python code for them. To select these examples, we compute embeddings for the input question and of all of the questions for which we have annotated programs (a randomly chosen set of fifty). Then, we select questions that have the highest similarity to the input and use them as in-context examples. Given the prompt and question that we want to answer, the LLM generates a Python program representing that question.

We instantiate the CodeVQA framework using three visual functions: (1) query, (2) get_pos, and (3) find_matching_image.

• Query, which answers a question about a single image, is implemented using the few-shot Plug-and-Play VQA (PnP-VQA) method. PnP-VQA generates captions using BLIP — an image-captioning transformer pre-trained on millions of image-caption pairs — and feeds these into a LLM that outputs the answers to the question.
• Get_pos, which is an object localizer that takes a description of an object as input and returns its position in the image, is implemented using GradCAM. Specifically, the description and the image are passed through the BLIP joint text-image encoder, which predicts an image-text matching score. GradCAM takes the gradient of this score with respect to the image features to find the region most relevant to the text.
• Find_matching_image, which is used in multi-image questions to find the image that best matches a given input phrase, is implemented by using BLIP text and image encoders to compute a text embedding for the phrase and an image embedding for each image. Then the dot products of the text embedding with each image embedding represent the relevance of each image to the phrase, and we pick the image that maximizes this relevance.

The three functions can be implemented using models that require very little annotation (e.g., text and image-text pairs collected from the web and a small number of VQA examples). Furthermore, the CodeVQA framework can be easily generalized beyond these functions to others that a user might implement (e.g., object detection, image segmentation, or knowledge base retrieval).

Illustration of the CodeVQA method. First, a large language model generates a Python program (code), which invokes visual functions that represent the question. In this example, a simple VQA method (query) is used to answer one part of the question, and an object localizer (get_pos) is used to find the positions of the objects mentioned. Then the program produces an answer to the original question by combining the outputs of these functions.

Results

The CodeVQA framework correctly generates and executes Python programs not only for single-image questions, but also for multi-image questions. For example, if given two images, each showing two pandas, a question one might ask is, “Is it true that there are four pandas?” In this case, the LLM converts the counting question about the pair of images into a program in which an object count is obtained for each image (using the query function). Then the counts for both images are added to compute a total count, which is then compared to the number in the original question to yield a yes or no answer.

We evaluate CodeVQA on three visual reasoning datasets: GQA (single-image), COVR (multi-image), and NLVR2 (multi-image). For GQA, we provide 12 in-context examples to each method, and for COVR and NLVR2, we provide six in-context examples to each method. The table below shows that CodeVQA improves consistently over the baseline few-shot VQA method on all three datasets.

We find that in GQA, CodeVQA’s accuracy is roughly 30% higher than the baseline on spatial reasoning questions, 4% higher on “and” questions, and 3% higher on “or” questions. The third category includes multi-hop questions such as “Are there salt shakers or skateboards in the picture?”, for which the generated program is shown below.

img = open_image("Image13.jpg")
salt_shakers_exist = query(img, "Are there any salt shakers?")
skateboards_exist = query(img, "Are there any skateboards?")
if salt_shakers_exist == "yes" or skateboards_exist == "yes":
    answer = "yes"
else:
    answer = "no"

In COVR, we find that CodeVQA’s gain over the baseline is higher when the number of input images is larger, as shown in the table below. This trend indicates that breaking the problem down into single-image questions is beneficial.

Conclusion

We present CodeVQA, a framework for few-shot visual question answering that relies on code generation to perform multi-step visual reasoning. Exciting directions for future work include expanding the set of modules used and creating a similar framework for visual tasks beyond VQA. We note that care should be taken when considering whether to deploy a system such as CodeVQA, since vision-language models like the ones used in our visual functions have been shown to exhibit social biases. At the same time, compared to monolithic models, CodeVQA offers additional interpretability (through the Python program) and controllability (by modifying the prompts or visual functions), which are useful in production systems.

Acknowledgements

This research was a collaboration between UC Berkeley’s Artificial Intelligence Research lab (BAIR) and Google Research, and was conducted by Sanjay Subramanian, Medhini Narasimhan, Kushal Khangaonkar, Kevin Yang, Arsha Nagrani, Cordelia Schmid, Andy Zeng, Trevor Darrell, and Dan Klein.

Read More

Posted by Kuniaki Saito, Student Researcher, Google Research, Cloud AI Team, and Kihyuk Sohn, Research Scientist, Google Research

Image retrieval plays a crucial role in search engines. Typically, their users rely on either image or text as a query to retrieve a desired target image. However, text-based retrieval has its limitations, as describing the target image accurately using words can be challenging. For instance, when searching for a fashion item, users may want an item whose specific attribute, e.g., the color of a logo or the logo itself, is different from what they find in a website. Yet searching for the item in an existing search engine is not trivial since precisely describing the fashion item by text can be challenging. To address this fact, composed image retrieval (CIR) retrieves images based on a query that combines both an image and a text sample that provides instructions on how to modify the image to fit the intended retrieval target. Thus, CIR allows precise retrieval of the target image by combining image and text.

However, CIR methods require large amounts of labeled data, i.e., triplets of a 1) query image, 2) description, and 3) target image. Collecting such labeled data is costly, and models trained on this data are often tailored to a specific use case, limiting their ability to generalize to different datasets.

To address these challenges, in “Pic2Word: Mapping Pictures to Words for Zero-shot Composed Image Retrieval”, we propose a task called zero-shot CIR (ZS-CIR). In ZS-CIR, we aim to build a single CIR model that performs a variety of CIR tasks, such as object composition, attribute editing, or domain conversion, without requiring labeled triplet data. Instead, we propose to train a retrieval model using large-scale image-caption pairs and unlabeled images, which are considerably easier to collect than supervised CIR datasets at scale. To encourage reproducibility and further advance this space, we also release the code.

Description of existing composed image retrieval model.

We train a composed image retrieval model using image-caption data only. Our model retrieves images aligned with the composition of the query image and text.

Method overview

We propose to leverage the language capabilities of the language encoder in the contrastive language-image pre-trained model (CLIP), which excels at generating semantically meaningful language embeddings for a wide range of textual concepts and attributes. To that end, we use a lightweight mapping sub-module in CLIP that is designed to map an input picture (e.g., a photo of a cat) from the image embedding space to a word token (e.g., “cat”) in the textual input space. The whole network is optimized with the vision-language contrastive loss to again ensure the visual and text embedding spaces are as close as possible given a pair of an image and its textual description. Then, the query image can be treated as if it is a word. This enables the flexible and seamless composition of query image features and text descriptions by the language encoder. We call our method Pic2Word and provide an overview of its training process in the figure below. We want the mapped token s to represent the input image in the form of word token. Then, we train the mapping network to reconstruct the image embedding in the language embedding, p. Specifically, we optimize the contrastive loss proposed in CLIP computed between the visual embedding v and the textual embedding p.

Training of the mapping network (fM) using unlabeled images only. We optimize only the mapping network with a frozen visual and text encoder.

Given the trained mapping network, we can regard an image as a word token and pair it with the text description to flexibly compose the joint image-text query as shown in the figure below.

With the trained mapping network, we regard the image as a word token and pair it with the text description to flexibly compose the joint image-text query.

Evaluation

We conduct a variety of experiments to evaluate Pic2Word’s performance on a variety of CIR tasks.

Domain conversion

We first evaluate the capability of compositionality of the proposed method on domain conversion — given an image and the desired new image domain (e.g., sculpture, origami, cartoon, toy), the output of the system should be an image with the same content but in the new desired image domain or style. As illustrated below, we evaluate the ability to compose the category information and domain description given as an image and text, respectively. We evaluate the conversion from real images to four domains using ImageNet and ImageNet-R.

To compare with approaches that do not require supervised training data, we pick three approaches: (i) image only performs retrieval only with visual embedding, (ii) text only employs only text embedding, and (iii) image + text averages the visual and text embedding to compose the query. The comparison with (iii) shows the importance of composing image and text using a language encoder. We also compare with Combiner, which trains the CIR model on Fashion-IQ or CIRR.

We aim to convert the domain of the input query image into the one described with text, e.g., origami.

As shown in figure below, our proposed approach outperforms baselines by a large margin.

Results (recall@10, i.e., the percentage of relevant instances in the first 10 images retrieved.) on composed image retrieval for domain conversion.

Fashion attribute composition

Next, we evaluate the composition of fashion attributes, such as the color of cloth, logo, and length of sleeve, using the Fashion-IQ dataset. The figure below illustrates the desired output given the query.

Overview of CIR for fashion attributes.

In the figure below, we present a comparison with baselines, including supervised baselines that utilized triplets for training the CIR model: (i) CB uses the same architecture as our approach, (ii) CIRPLANT, ALTEMIS, MAAF use a smaller backbone, such as ResNet50. Comparison to these approaches will give us the understanding on how well our zero-shot approach performs on this task.

Although CB outperforms our approach, our method performs better than supervised baselines with smaller backbones. This result suggests that by utilizing a robust CLIP model, we can train a highly effective CIR model without requiring annotated triplets.

Results (recall@10, i.e., the percentage of relevant instances in the first 10 images retrieved.) on composed image retrieval for Fashion-IQ dataset (higher is better). Light blue bars train the model using triplets. Note that our approach performs on par with these supervised baselines with shallow (smaller) backbones.

Qualitative results

We show several examples in the figure below. Compared to a baseline method that does not require supervised training data (text + image feature averaging), our approach does a better job of correctly retrieving the target image.

Qualitative results on diverse query images and text description.

Conclusion and future work

In this article, we introduce Pic2Word, a method for mapping pictures to words for ZS-CIR. We propose to convert the image into a word token to achieve a CIR model using only an image-caption dataset. Through a variety of experiments, we verify the effectiveness of the trained model on diverse CIR tasks, indicating that training on an image-caption dataset can build a powerful CIR model. One potential future research direction is utilizing caption data to train the mapping network, although we use only image data in the present work.

Acknowledgements

This research was conducted by Kuniaki Saito, Kihyuk Sohn, Xiang Zhang, Chun-Liang Li, Chen-Yu Lee, Kate Saenko, and Tomas Pfister. Also thanks to Zizhao Zhang and Sergey Ioffe for their valuable feedback.

Read More

AI is the most profound technology that humanity is working on today. It’s a critical part of solving big societal challenges and helping to make our everyday lives bett…Read More

28 institutions from around the world join forces to showcase Leonardo da Vinci’s unparalleled legacy, blending art, science, and AI innovationRead More

Posted by Fabian Pedregosa and Eleni Triantafillou, Research Scientists, Google

Deep learning has recently driven tremendous progress in a wide array of applications, ranging from realistic image generation and impressive retrieval systems to language models that can hold human-like conversations. While this progress is very exciting, the widespread use of deep neural network models requires caution: as guided by Google’s AI Principles, we seek to develop AI technologies responsibly by understanding and mitigating potential risks, such as the propagation and amplification of unfair biases and protecting user privacy.

Fully erasing the influence of the data requested to be deleted is challenging since, aside from simply deleting it from databases where it’s stored, it also requires erasing the influence of that data on other artifacts such as trained machine learning models. Moreover, recent research [1, 2] has shown that in some cases it may be possible to infer with high accuracy whether an example was used to train a machine learning model using membership inference attacks (MIAs). This can raise privacy concerns, as it implies that even if an individual’s data is deleted from a database, it may still be possible to infer whether that individual’s data was used to train a model.

Given the above, machine unlearning is an emergent subfield of machine learning that aims to remove the influence of a specific subset of training examples — the “forget set” — from a trained model. Furthermore, an ideal unlearning algorithm would remove the influence of certain examples while maintaining other beneficial properties, such as the accuracy on the rest of the train set and generalization to held-out examples. A straightforward way to produce this unlearned model is to retrain the model on an adjusted training set that excludes the samples from the forget set. However, this is not always a viable option, as retraining deep models can be computationally expensive. An ideal unlearning algorithm would instead use the already-trained model as a starting point and efficiently make adjustments to remove the influence of the requested data.

Today we’re thrilled to announce that we’ve teamed up with a broad group of academic and industrial researchers to organize the first Machine Unlearning Challenge. The competition considers a realistic scenario in which after training, a certain subset of the training images must be forgotten to protect the privacy or rights of the individuals concerned. The competition will be hosted on Kaggle, and submissions will be automatically scored in terms of both forgetting quality and model utility. We hope that this competition will help advance the state of the art in machine unlearning and encourage the development of efficient, effective and ethical unlearning algorithms.

Machine unlearning applications

Machine unlearning has applications beyond protecting user privacy. For instance, one can use unlearning to erase inaccurate or outdated information from trained models (e.g., due to errors in labeling or changes in the environment) or remove harmful, manipulated, or outlier data.

The field of machine unlearning is related to other areas of machine learning such as differential privacy, life-long learning, and fairness. Differential privacy aims to guarantee that no particular training example has too large an influence on the trained model; a stronger goal compared to that of unlearning, which only requires erasing the influence of the designated forget set. Life-long learning research aims to design models that can learn continuously while maintaining previously-acquired skills. As work on unlearning progresses, it may also open additional ways to boost fairness in models, by correcting unfair biases or disparate treatment of members belonging to different groups (e.g., demographics, age groups, etc.).

Anatomy of unlearning. An unlearning algorithm takes as input a pre-trained model and one or more samples from the train set to unlearn (the “forget set”). From the model, forget set, and retain set, the unlearning algorithm produces an updated model. An ideal unlearning algorithm produces a model that is indistinguishable from the model trained without the forget set.

Challenges of machine unlearning

The problem of unlearning is complex and multifaceted as it involves several conflicting objectives: forgetting the requested data, maintaining the model’s utility (e.g., accuracy on retained and held-out data), and efficiency. Because of this, existing unlearning algorithms make different trade-offs. For example, full retraining achieves successful forgetting without damaging model utility, but with poor efficiency, while adding noise to the weights achieves forgetting at the expense of utility.

Furthermore, the evaluation of forgetting algorithms in the literature has so far been highly inconsistent. While some works report the classification accuracy on the samples to unlearn, others report distance to the fully retrained model, and yet others use the error rate of membership inference attacks as a metric for forgetting quality [4, 5, 6].

We believe that the inconsistency of evaluation metrics and the lack of a standardized protocol is a serious impediment to progress in the field — we are unable to make direct comparisons between different unlearning methods in the literature. This leaves us with a myopic view of the relative merits and drawbacks of different approaches, as well as open challenges and opportunities for developing improved algorithms. To address the issue of inconsistent evaluation and to advance the state of the art in the field of machine unlearning, we’ve teamed up with a broad group of academic and industrial researchers to organize the first unlearning challenge.

Announcing the first Machine Unlearning Challenge

We are pleased to announce the first Machine Unlearning Challenge, which will be held as part of the NeurIPS 2023 Competition Track. The goal of the competition is twofold. First, by unifying and standardizing the evaluation metrics for unlearning, we hope to identify the strengths and weaknesses of different algorithms through apples-to-apples comparisons. Second, by opening this competition to everyone, we hope to foster novel solutions and shed light on open challenges and opportunities.

The competition will be hosted on Kaggle and run between mid-July 2023 and mid-September 2023. As part of the competition, today we’re announcing the availability of the starting kit. This starting kit provides a foundation for participants to build and test their unlearning models on a toy dataset.

The competition considers a realistic scenario in which an age predictor has been trained on face images, and, after training, a certain subset of the training images must be forgotten to protect the privacy or rights of the individuals concerned. For this, we will make available as part of the starting kit a dataset of synthetic faces (samples shown below) and we’ll also use several real-face datasets for evaluation of submissions. The participants are asked to submit code that takes as input the trained predictor, the forget and retain sets, and outputs the weights of a predictor that has unlearned the designated forget set. We will evaluate submissions based on both the strength of the forgetting algorithm and model utility. We will also enforce a hard cut-off that rejects unlearning algorithms that run slower than a fraction of the time it takes to retrain. A valuable outcome of this competition will be to characterize the trade-offs of different unlearning algorithms.

Excerpt images from the Face Synthetics dataset together with age annotations. The competition considers the scenario in which an age predictor has been trained on face images like the above, and, after training, a certain subset of the training images must be forgotten.

For evaluating forgetting, we will use tools inspired by MIAs, such as LiRA. MIAs were first developed in the privacy and security literature and their goal is to infer which examples were part of the training set. Intuitively, if unlearning is successful, the unlearned model contains no traces of the forgotten examples, causing MIAs to fail: the attacker would be unable to infer that the forget set was, in fact, part of the original training set. In addition, we will also use statistical tests to quantify how different the distribution of unlearned models (produced by a particular submitted unlearning algorithm) is compared to the distribution of models retrained from scratch. For an ideal unlearning algorithm, these two will be indistinguishable.

Conclusion

Machine unlearning is a powerful tool that has the potential to address several open problems in machine learning. As research in this area continues, we hope to see new methods that are more efficient, effective, and responsible. We are thrilled to have the opportunity via this competition to spark interest in this field, and we are looking forward to sharing our insights and findings with the community.

Acknowledgements

The authors of this post are now part of Google DeepMind. We are writing this blog post on behalf of the organization team of the Unlearning Competition: Eleni Triantafillou*, Fabian Pedregosa* (*equal contribution), Meghdad Kurmanji, Kairan Zhao, Gintare Karolina Dziugaite, Peter Triantafillou, Ioannis Mitliagkas, Vincent Dumoulin, Lisheng Sun Hosoya, Peter Kairouz, Julio C. S. Jacques Junior, Jun Wan, Sergio Escalera and Isabelle Guyon.

Read More

Posted by Yang Zhao and Tingbo Hou, Software Engineers, Core ML

In recent years, diffusion models have shown great success in text-to-image generation, achieving high image quality, improved inference performance, and expanding our creative inspiration. Nevertheless, it is still challenging to efficiently control the generation, especially with conditions that are difficult to describe with text.

Today, we announce MediaPipe diffusion plugins, which enable controllable text-to-image generation to be run on-device. Expanding upon our prior work on GPU inference for on-device large generative models, we introduce new low-cost solutions for controllable text-to-image generation that can be plugged into existing diffusion models and their Low-Rank Adaptation (LoRA) variants.

Text-to-image generation with control plugins running on-device.

Background

With diffusion models, image generation is modeled as an iterative denoising process. Starting from a noise image, at each step, the diffusion model gradually denoises the image to reveal an image of the target concept. Research shows that leveraging language understanding via text prompts can greatly improve image generation. For text-to-image generation, the text embedding is connected to the model via cross-attention layers. Yet, some information is difficult to describe by text prompts, e.g., the position and pose of an object. To address this problem, researchers add additional models into the diffusion to inject control information from a condition image.

Common approaches for controlled text-to-image generation include Plug-and-Play, ControlNet, and T2I Adapter. Plug-and-Play applies a widely used denoising diffusion implicit model (DDIM) inversion approach that reverses the generation process starting from an input image to derive an initial noise input, and then employs a copy of the diffusion model (860M parameters for Stable Diffusion 1.5) to encode the condition from an input image. Plug-and-Play extracts spatial features with self-attention from the copied diffusion, and injects them into the text-to-image diffusion. ControlNet creates a trainable copy of the encoder of a diffusion model, which connects via a convolution layer with zero-initialized parameters to encode conditioning information that is conveyed to the decoder layers. However, as a result, the size is large, half that of the diffusion model (430M parameters for Stable Diffusion 1.5). T2I Adapter is a smaller network (77M parameters) and achieves similar effects in controllable generation. T2I Adapter only takes the condition image as input, and its output is shared across all diffusion iterations. Yet, the adapter model is not designed for portable devices.

The MediaPipe diffusion plugins

To make conditioned generation efficient, customizable, and scalable, we design the MediaPipe diffusion plugin as a separate network that is:

• Plugable: It can be easily connected to a pre-trained base model.
• Trained from scratch: It does not use pre-trained weights from the base model.
• Portable: It runs outside the base model on mobile devices, with negligible cost compared to the base model inference.

The MediaPipe diffusion plugin is a portable on-device model for text-to-image generation. It extracts multiscale features from a conditioning image, which are added to the encoder of a diffusion model at corresponding levels. When connecting to a text-to-image diffusion model, the plugin model can provide an extra conditioning signal to the image generation. We design the plugin network to be a lightweight model with only 6M parameters. It uses depth-wise convolutions and inverted bottlenecks from MobileNetv2 for fast inference on mobile devices.

Overview of the MediaPipe diffusion model plugin. The plugin is a separate network, whose output can be plugged into a pre-trained text-to-image generation model. Features extracted by the plugin are applied to the associated downsampling layer of the diffusion model (blue).

Unlike ControlNet, we inject the same control features in all diffusion iterations. That is, we only run the plugin once for one image generation, which saves computation. We illustrate some intermediate results of a diffusion process below. The control is effective at every diffusion step and enables controlled generation even at early steps. More iterations improve the alignment of the image with the text prompt and generate more detail.

Illustration of the generation process using the MediaPipe diffusion plugin.

Examples

In this work, we developed plugins for a diffusion-based text-to-image generation model with MediaPipe Face Landmark, MediaPipe Holistic Landmark, depth maps, and Canny edge. For each task, we select about 100K images from a web-scale image-text dataset, and compute control signals using corresponding MediaPipe solutions. We use refined captions from PaLI for training the plugins.

Face Landmark

The MediaPipe Face Landmarker task computes 478 landmarks (with attention) of a human face. We use the drawing utils in MediaPipe to render a face, including face contour, mouth, eyes, eyebrows, and irises, with different colors. The following table shows randomly generated samples by conditioning on face mesh and prompts. As a comparison, both ControlNet and Plugin can control text-to-image generation with given conditions.

Face-landmark plugin for text-to-image generation, compared with ControlNet.

Holistic Landmark

MediaPipe Holistic Landmarker task includes landmarks of body pose, hands, and face mesh. Below, we generate various stylized images by conditioning on the holistic features.

Holistic-landmark plugin for text-to-image generation.

Depth

Depth-plugin for text-to-image generation.

Canny Edge

Canny-edge plugin for text-to-image generation.

Evaluation

We conduct a quantitative study of the face landmark plugin to demonstrate the model’s performance. The evaluation dataset contains 5K human images. We compare the generation quality as measured by the widely used metrics, Fréchet Inception Distance (FID) and CLIP scores. The base model is a pre-trained text-to-image diffusion model. We use Stable Diffusion v1.5 here.

As shown in the following table, both ControlNet and the MediaPipe diffusion plugin produce much better sample quality than the base model, in terms of FID and CLIP scores. Unlike ControlNet, which needs to run at every diffusion step, the MediaPipe plugin only runs once for each image generated. We measured the performance of the three models on a server machine (with Nvidia V100 GPU) and a mobile phone (Galaxy S23). On the server, we run all three models with 50 diffusion steps, and on mobile, we run 20 diffusion steps using the MediaPipe image generation app. Compared with ControlNet, the MediaPipe plugin shows a clear advantage in inference efficiency while preserving the sample quality.

We test the performance of the plugin on a wide range of mobile devices from mid-tier to high-end. We list the results on some representative devices in the following table, covering both Android and iOS.

Conclusion

In this work, we present MediaPipe, a portable plugin for conditioned text-to-image generation. It injects features extracted from a condition image to a diffusion model, and consequently controls the image generation. Portable plugins can be connected to pre-trained diffusion models running on servers or devices. By running text-to-image generation and plugins fully on-device, we enable more flexible applications of generative AI.

Acknowledgments

We’d like to thank all team members who contributed to this work: Raman Sarokin and Juhyun Lee for the GPU inference solution; Khanh LeViet, Chuo-Ling Chang, Andrei Kulik, and Matthias Grundmann for leadership. Special thanks to Jiuqiang Tang, Joe Zou and Lu wang, who made this technology and all the demos running on-device.

Read More

Posted by Kuniaki Saito, Student Researcher, Cloud AI Team, and Kihyuk Sohn, Research Scientist, Perception Team

Pre-training visual language (VL) models on web-scale image-caption datasets has recently emerged as a powerful alternative to traditional pre-training on image classification data. Image-caption datasets are considered to be more “open-domain” because they contain broader scene types and vocabulary words, which result in models with strong performance in few- and zero-shot recognition tasks. However, images with fine-grained class descriptions can be rare, and the class distribution can be imbalanced since image-caption datasets do not go through manual curation. By contrast, large-scale classification datasets, such as ImageNet, are often curated and can thus provide fine-grained categories with a balanced label distribution. While it may sound promising, directly combining caption and classification datasets for pre-training is often unsuccessful as it can result in biased representations that do not generalize well to various downstream tasks.

In “Prefix Conditioning Unifies Language and Label Supervision”, presented at CVPR 2023, we demonstrate a pre-training strategy that uses both classification and caption datasets to provide complementary benefits. First, we show that naïvely unifying the datasets results in sub-optimal performance on downstream zero-shot recognition tasks as the model is affected by dataset bias: the coverage of image domains and vocabulary words is different in each dataset. We address this problem during training through prefix conditioning, a novel simple and effective method that uses prefix tokens to disentangle dataset biases from visual concepts. This approach allows the language encoder to learn from both datasets while also tailoring feature extraction to each dataset. Prefix conditioning is a generic method that can be easily integrated into existing VL pre-training objectives, such as Contrastive Language-Image Pre-training (CLIP) or Unified Contrastive Learning (UniCL).

High-level idea

We note that classification datasets tend to be biased in at least two ways: (1) the images mostly contain single objects from restricted domains, and (2) the vocabulary is limited and lacks the linguistic flexibility required for zero-shot learning. For example, the class embedding of “a photo of a dog” optimized for ImageNet usually results in a photo of one dog in the center of the image pulled from the ImageNet dataset, which does not generalize well to other datasets containing images of multiple dogs in different spatial locations or a dog with other subjects.

By contrast, caption datasets contain a wider variety of scene types and vocabularies. As shown below, if a model simply learns from two datasets, the language embedding can entangle the bias from the image classification and caption dataset, which can decrease the generalization in zero-shot classification. If we can disentangle the bias from two datasets, we can use language embeddings that are tailored for the caption dataset to improve generalization.

Top: Language embedding entangling the bias from image classification and caption dataset. Bottom: Language embeddings disentangles the bias from two datasets.

Prefix conditioning

Prefix conditioning is partially inspired by prompt tuning, which prepends learnable tokens to the input token sequences to instruct a pre-trained model backbone to learn task-specific knowledge that can be used to solve downstream tasks. The prefix conditioning approach differs from prompt tuning in two ways: (1) it is designed to unify image-caption and classification datasets by disentangling the dataset bias, and (2) it is applied to VL pre-training while the standard prompt tuning is used to fine-tune models. Prefix conditioning is an explicit way to specifically steer the behavior of model backbones based on the type of datasets provided by users. This is especially helpful in production when the number of different types of datasets is known ahead of time.

During training, prefix conditioning learns a text token (prefix token) for each dataset type, which absorbs the bias of the dataset and allows the remaining text tokens to focus on learning visual concepts. Specifically, it prepends prefix tokens for each dataset type to the input tokens that inform the language and visual encoder of the input data type (e.g., classification vs. caption). Prefix tokens are trained to learn the dataset-type-specific bias, which enables us to disentangle that bias in language representations and utilize the embedding learned on the image-caption dataset during test time, even without an input caption.

We utilize prefix conditioning for CLIP using a language and visual encoder. During test time, we employ the prefix used for the image-caption dataset since the dataset is supposed to cover broader scene types and vocabulary words, leading to better performance in zero-shot recognition.

Illustration of the Prefix Conditioning.

Experimental results

We apply prefix conditioning to two types of contrastive loss, CLIP and UniCL, and evaluate their performance on zero-shot recognition tasks compared to models trained with ImageNet21K (IN21K) and Conceptual 12M (CC12M). CLIP and UniCL models trained with two datasets using prefix conditioning show large improvements in zero-shot classification accuracy.

Zero-shot classification accuracy of models trained with only IN21K or CC12M compared to CLIP and UniCL models trained with both two datasets using prefix conditioning (“Ours”).

Study on test-time prefix

The table below describes the performance change by the prefix used during test time. We demonstrate that by using the same prefix used for the classification dataset (“Prompt”), the performance on the classification dataset (IN-1K) improves. When using the same prefix used for the image-caption dataset (“Caption”), the performance on other datasets (Zero-shot AVG) improves. This analysis illustrates that if the prefix is tailored for the image-caption dataset, it achieves better generalization of scene types and vocabulary words.

Analysis of the prefix used for test-time.

Study on robustness to image distribution shift

We study the shift in image distribution using ImageNet variants. We see that the “Caption” prefix performs better than “Prompt” in ImageNet-R (IN-R) and ImageNet-Sketch (IN-S), but underperforms in ImageNet-V2 (IN-V2). This indicates that the “Caption” prefix achieves generalization on domains far from the classification dataset. Therefore, the optimal prefix probably differs by how far the test domain is from the classification dataset.

Analysis on the robustness to image-level distribution shift. IN: ImageNet, IN-V2: ImageNet-V2, IN-R: Art, Cartoon style ImageNet, IN-S: ImageNet Sketch.

Conclusion and future work

We introduce prefix conditioning, a technique for unifying image caption and classification datasets for better zero-shot classification. We show that this approach leads to better zero-shot classification accuracy and that the prefix can control the bias in the language embedding. One limitation is that the prefix learned on the caption dataset is not necessarily optimal for the zero-shot classification. Identifying the optimal prefix for each test dataset is an interesting direction for future work.

Acknowledgements

This research was conducted by Kuniaki Saito, Kihyuk Sohn, Xiang Zhang, Chun-Liang Li, Chen-Yu Lee, Kate Saenko, and Tomas Pfister. Thanks to Zizhao Zhang and Sergey Ioffe for their valuable feedback.

Read More