Timezone: »

Shared Visual Representations in Human and Machine Intelligence
Arturo Deza · Joshua Peterson · Apurva Ratan Murty · Tom Griffiths

Fri Dec 13 08:00 AM -- 07:00 PM (PST) @ West 220 - 222
Event URL: http://www.svrhm2019.com/ »

The goal of the Shared Visual Representations in Human and Machine Intelligence (SVRHM) workshop is to disseminate relevant, parallel findings in the fields of computational neuroscience, psychology, and cognitive science that may inform modern machine learning methods.

In the past few years, machine learning methods—especially deep neural networks—have widely permeated the vision science, cognitive science, and neuroscience communities. As a result, scientific modeling in these fields has greatly benefited, producing a swath of potentially critical new insights into human learning and intelligence, which remains the gold standard for many tasks. However, the machine learning community has been largely unaware of these cross-disciplinary insights and analytical tools, which may help to solve many of the current problems that ML theorists and engineers face today (e.g., adversarial attacks, compression, continual learning, and unsupervised learning).

Thus we propose to invite leading cognitive scientists with strong computational backgrounds to disseminate their findings to the machine learning community with the hope of closing the loop by nourishing new ideas and creating cross-disciplinary collaborations.

See more information at the official conference website: https://www.svrhm2019.com/
Follow us on twitter for announcements: https://twitter.com/svrhm2019

Fri 8:50 a.m. - 9:00 a.m. [iCal]
Opening Remarks
Arturo Deza, Joshua Peterson, Apurva Ratan Murty, Tom Griffiths
Fri 9:00 a.m. - 9:25 a.m. [iCal]

Despite recent progress in artificial intelligence, humans and animals vastly surpass machine agents in their ability to quickly learn about their environment. While humans generalize to new concepts from small numbers of examples, state-of-the-art artificial neural networks still require huge amounts of supervision. We hypothesize that humans benefit from such data-efficiency because their internal representations support a much wider set tasks (such as planning and decision-making) which often require making predictions about future events. Using the curvature of natural videos as a measure of predictability, we find that human perceptual representations are indeed more predictable than their inputs, whereas current deep neural networks are not. Conversely, by optimizing neural networks for an information-theoretic measure of predictability, we arrive at artificial classifiers whose data-efficiency greatly surpasses that of purely supervised ones. Learning predictable representations may therefore enable artificial systems to perceive the world in a manner that is closer to biological ones.

Olivier Henaff
Fri 9:25 a.m. - 9:50 a.m. [iCal]

Despite the advances in modern deep learning approaches, we are still quite far from the generality, robustness and data efficiency of biological intelligence. In this talk I will suggest that this gap may be narrowed by re-focusing from implicit representation learning prevalent in end-to-end deep learning approaches to explicit unsupervised representation learning. In particular, I will discuss the value of disentangled visual representations acquired in an unsupervised manner loosely inspired by biological intelligence. In particular, this talk will connect disentangling with the ideas of symmetry transformations from physics to make a claim that disentangled representations reflect important world structure. I will then go over a few first demonstrations of how such representations can be useful in practice for continual learning, acquiring reinforcement learning (RL) policies that are more robust to transfer scenarios that standard RL approaches, and building abstract compositional visual concepts which make possible imagination of meaningful and diverse samples beyond the training data distribution.

Irina Higgins
Fri 9:50 a.m. - 10:10 a.m. [iCal]
Coffee Break (Break)
Fri 10:10 a.m. - 10:35 a.m. [iCal]

A "distribution mismatch" dataset for comparing representational similarity in ANNs and the brain

Wu Xiao
Fri 10:35 a.m. - 11:00 a.m. [iCal]
Feathers, wings and the future of computer vision research (Talk)
Bill Freeman
Fri 11:00 a.m. - 11:25 a.m. [iCal]
Taxonomic structure in learning from few positive examples (Talk)
Erin Grant
Fri 11:25 a.m. - 11:50 a.m. [iCal]

The classification performance of deep neural networks has begun to asymptote at near-perfect levels on natural image benchmarks. However, their ability to generalize outside the training set and their robustness to adversarial attacks have not. Humans, by contrast, exhibit robust and graceful generalization far outside their set of training samples. In this talk, I will discuss one strategy for translating these properties to machine-learning classifiers: training them to be uncertain in the same way as humans, rather than always right. When we integrate human uncertainty into training paradigms by using human guess distributions as labels, we find the generalize better and are more robust to adversarial attacks. Rather than expect all image datasets to come with such labels, we instead intend our CIFAR10H dataset to be used as a gold standard, with which algorithmic means of capturing the same information can be evaluated. To illustrate this, I present one automated method that does so—deep prototype models inspired by the cognitive science literature.

Ruairidh Battleday
Fri 11:50 a.m. - 12:15 p.m. [iCal]
Making the next generation of machine learning datasets: ObjectNet a new object recognition benchmark (Talk)
Andrei Barbu
Fri 12:15 p.m. - 12:40 p.m. [iCal]
The building blocks of vision (Talk)
Michael Tarr
Fri 2:00 p.m. - 3:00 p.m. [iCal]
Poster Session
Ethan Harris, Tom White, Oh Hyeon Choung, Takashi Shinozaki, Dipan Pal, Katherine Hermann, Judy Borowski, Camilo Fosco, Chaz Firestone, Vijay Veerabadran, Ben Lahner, Chaitanya Ryali, Fenil Doshi, Pulkit Singh, Sharon Zhou, Michel Besserve, Michael Chang, Anelise Newman, Mahesan Niranjan, Jonathon Hare, Daniela Mihai, Marios Savvides, Simon Kornblith, Christina M Funke, Aude Oliva, Virginia R de Sa, Dmitry Krotov, Colin Conwell, George Alvarez, Alex Kolchinski, Shengjia Zhao, Mitchell Gordon, Michael Bernstein, Stefano Ermon, Arash Mehrjou, Bernhard Schölkopf, JD Co-Reyes, Michael Janner, Jiajun Wu, Josh Tenenbaum, Sergey Levine, Yalda Mohsenzadeh, Zhenglong Zhou
Fri 3:00 p.m. - 3:30 p.m. [iCal]

"Cross-disciplinary research experiences and tips for Graduate School Admissions Panelists"

Panelists: Erin Grant (UC Berkeley) Nadine Chang (CMU) Ruairidh Battleday (Princeton) Sophia Sanborn (UC Berkeley) Nikhil Parthasarathy (NYU)

Erin Grant, Ruairidh Battleday, Sophia Sanborn, Nadine Chang, Nikhil Parthasarathy
Fri 3:30 p.m. - 3:55 p.m. [iCal]
Object representation in the human visual system (Talk)
Talia Konkle
Fri 3:55 p.m. - 4:20 p.m. [iCal]
Cognitive computational neuroscience of vision (Talk)
Nikolaus Kriegeskorte
Fri 4:20 p.m. - 4:45 p.m. [iCal]
Perturbation-based remodeling of visual neural network representations (Talk)
Matthias Bethge
Fri 4:45 p.m. - 5:10 p.m. [iCal]
Local gain control and perceptual invariances (Talk)
Eero Simoncelli
Fri 5:10 p.m. - 6:00 p.m. [iCal]

Panelists: Irina Higgins (DeepMind), Talia Konkle (Harvard), Nikolaus Kriegeskorte (Columbia), Matthias Bethge (Universität Tübingen)

Irina Higgins, Talia Konkle, Matthias Bethge, Nikolaus Kriegeskorte
Fri 6:00 p.m. - 6:10 p.m. [iCal]

Best Paper Award Prize (NVIDIA Titan RTX) and Best Poster Award Prize (Oculus Quest)

Arturo Deza, Joshua Peterson, Apurva Ratan Murty, Tom Griffiths
Fri 6:10 p.m. - 7:00 p.m. [iCal]

Sponsored by MIT Quest for Intelligence

Author Information

Arturo Deza (Harvard University)
Joshua Peterson (Princeton University)
Apurva Ratan Murty (Massachusetts Institute of Technology)
Tom Griffiths (Princeton University)

More from the Same Authors