fMRI Scanning at MIT: For Kids MITCBMM 2022-12-06 | What to expect when your child is participating in an fMRI study at MIT
Mapping, analyzing, and emulating brain computations MITCBMM 2022-12-02 | Ed Boyden - MIT BCS, MIT MAS, MIT BE, MIT Quest
Welcome (afternoon) MITCBMM 2022-12-02 | Dan Huttenlocher, Dean of MIT Schwarzman School of Computing
Insights into AI algorithms drawn from hippocampal function MITCBMM 2022-12-02 | Matt Wilson - MIT BCS, MIT CBMMIla Fiete - MIT BCS, MIT Quest, MIT CBMM
Embodied Intelligence Mission MITCBMM 2022-12-02 | Leslie Kaelbling - MIT EECS, MIT CSAIL, MIT Quest, MIT CBMMNancy Kanwisher - MIT BCS, MIT Quest, MIT CBMMMehrdad Jazayeri - MIT BCS, MIT Quest
Quest | CBMM history and future MITCBMM 2022-12-02 | Tomaso Poggio - MIT BCS, MIT CSAIL, MIT Quest, Co-Director of the Center for Brains, Minds, and Machines
Tutorial on Statistical Inference On Representational Geometries MITCBMM 2022-12-02 | Heiko Schütt, NYURepresentational similarity analysis (RSA) is a popular method for comparing representations when a mapping between them is not available. One important comparison RSA is used for is between neuronal measurements and models of brain computation like deep neural networks. RSA is a two step process, first a matrix of pairwise dissimilarities between conditions is computed. This matrix is then a summary of the representational geometry, which can be compared directly between different representations as it has the same dimensions. In the first half of this tutorial, I will go through some recent advancements for RSA that improve the reliability and statistical accuracy of RSA substantially: First, I will explain the reasoning for cross-validated distance measures for computing the dissimilarity matrix and for whitened similarity measures to compare them to each other. Then, I will explain why simultaneous generalization to new subjects and new stimuli is hard and a solution based on bootstrapping. And finally, I will explain necessary cross-validation based extensions for flexible models. In the second half of this tutorial, I will give a guide how to run these analyses using our new rsatoolbox in python by going through demo notebooks that illustrate the functionality. Relevant papers:Schütt et al., 2021: Statistical inference on representational geometriesWalther et al., 2016: Reliability of dissimilarity measures for multi-voxel pattern analysisDiedrichsen et al., 2021: Comparing representational geometries using whitened unbiased-distance-matrix similarity GitHub repository: github.com/rsagroup/rsatoolboxGitHub demo repository: github.com/rsagroup/rsatoolbox/tree/main/demos
Developing Intelligence Mission MITCBMM 2022-12-02 | Joshua Tenenbaum - MIT BCS, MIT CSAIL, MIT Quest, MIT CBMMRebecca Saxe - MIT BCS, MIT Quest, MIT CBMMLaura Schulz - MIT BCS, MIT Quest, MIT CBMM
Introduction and Welcome MITCBMM 2022-12-02 | Jim DiCarlo, MIT BCS, MIT Quest, MIT CBMMNergis Mavalvala, Dean of MIT School of Science
Language Mission MITCBMM 2022-12-02 | Ev Fedorenko - MIT BCS, MIT QuestJacob Andreas - MIT CSAIL, MIT Quest
Building and evaluating multi-system functional brain models MITCBMM 2022-12-02 | Robert Guangyu Yang - MIT BCS, MIT EECS, MIT Quest, MIT CBMM
What would it mean to understand intelligence? MITCBMM 2022-12-02 | James DiCarlo, Director of the MIT Quest for Intelligence , Co-Director of the Center for Brains, Minds, and Machines
What to expect when participating in an fMRI study at MIT MITCBMM 2022-11-28 | For adult studies using visual stimuli only (ie no earphone use).
Cell Type-Specific Transcriptomics MITCBMM 2022-11-22 | Sebastian Pineda, MITTutorial on transcriptomic assays - TRAP and snRNA-seq sequencing with Sebastian PinedaHigh-throughput sequencing assays have become ubiquitous and indispensable tools in molecular neurobiology. They provide a means to investigate gene expression, dissect gene interactions and pathways, and examine how these change in model systems and disease. Recent developments in transcriptomics have yielded methods that allow us to survey gene expression with cell-type specificity at varying degrees of resolution. In this tutorial, I will give a brief overview of two common cell-type specific transcriptomic assays: TRAP (translating ribosome affinity purification) and snRNA-seq (single-nucleus RNA) sequencing. I will introduce the types of data that these assays generate and discuss the advantages and use cases of each. The bulk of the tutorial will focus on a typical data preprocessing and analysis workflow covering the topics dimensionality reduction, count normalization, clustering, and cell type annotation, as well as key downstream analysis concepts such as differential gene expression and pathway enrichment. Lastly, I will showcase recently developed tools and software packages that facilitate more sophisticated methods such as batch correction, data imputation, and network construction to simplify and enhance large-scale transcriptomics data analysis.Sebastian's website: https://heiman.mit.edu/team/pineda-sebastian/
What to expect when participating in an fMRI study at MIT MITCBMM 2022-11-18 | For adult studies using audio stimuli (ie earphone use).
Reintegrating AI: Skills, Symbols, and the Sensorimotor Dilemma MITCBMM 2022-10-25 | George Konidaris, Brown UniversityAbstract: AI is, at once, an immensely successful field---generating remarkable ongoing innovation that powers whole industries---and a complete failure. Despite more than 50 years of study, the field has never settled on a widely accepted, or even well-formulated, definition of its primary scientific goal: designing a general intelligence. Instead it consists of siloed subfields studying isolated aspects of intelligence, each of which is important but none of which can reasonably claim to address the problem as a whole. But intelligence is not a collection of loosely related capabilities; AI is not about learning or planning, reasoning or vision, grasping or language---it is about all of these capabilities, and how they work together to generate complex behavior. We cannot hope to make progress towards answering the overarching scientific question without a sincere and sustained effort to reintegrate the field.My talk will describe the current working hypothesis of the Brown Integrative, General-Purpose AI (bigAI) group, which takes the form of a decision-theoretic model that could plausibly generate the full range of intelligent behavior. Our approach is explicitly structuralist: we aim to understand how to structure intelligent agent by reintegrating, rather than discarding, existing subfields into a intellectually coherent single model. The model follows from the claim that general intelligence can only coherently be ascribed to a robot, not a computer, and that the resulting interaction with the world can be well-modeled as a decision process. Such a robot faces a sensorimotor dilemma: it must necessarily operate in a very rich sensorimotor space---one sufficient to support all the tasks it must solve, but that is therefore vastly overpowered for any single one. A core (but heretofore largely neglected) requirement for general intelligence is therefore the ability to autonomously formulate streamlined, task-specific representations, of the kind that single-task agents are typically assumed to be given. Our model also cleanly incorporates existing techniques developed in robotics, viewing them as innate knowledge about the structure of the world and the robot, and modeling them as the first few layers of a hierarchy of decision processes. Finally, our model suggests that language should ground to decision process formalisms, rather than abstract knowledge bases, text, or video, because they are the model that best characterizes the principal task facing both humans and robots.Speaker Bio: George Konidaris is an Associate Professor of Computer Science and director of the Intelligent Robot Lab at Brown, which forms part of bigAI (Brown Integrative, General AI). He is also the Chief Roboticist of Realtime Robotics, a startup based on his research on robot motion planning. Konidaris focuses on understanding how to design agents that learn abstraction hierarchies that enable fast, goal-oriented planning. He develops and applies techniques from machine learning, reinforcement learning, optimal control and planning to construct well-grounded hierarchies that result in fast planning for common cases, and are robust to uncertainty at every level of control.
Introduction to transformer architecture and discussion MITCBMM 2022-10-17 | Phillip Isola and Brian Cheung, MITThis is meant to be an informal discussion in which Phillip and Brian will give an overview of transformer networks and then we will open the floor for questions and discussion. It is likely that we will have another meeting at a later time discussing what transformers may contribute to neuroscience.
How brain computations can inspire new paths in AI - Part 2 MITCBMM 2022-09-19 | Gabriel Kreiman; Harvard University, Children's Hospital Boston
How brain computations can inspire new paths in AI - Part 1 MITCBMM 2022-09-19 | Gabriel Kreiman; Harvard University, Children's Hospital Boston
The quest to create engineer-quality models of the mechanisms of human visual object recognition PT2 MITCBMM 2022-09-15 | James DiCarlo, MIT
The quest to create engineer-quality models of the mechanisms of human visual object recognition PT1 MITCBMM 2022-09-12 | James DiCarlo, MIT
Clustering and generalization of abstract structures in reinforcement learning MITCBMM 2022-09-07 | Michael J. Frank, Brown UniversityPart of the 2022 Brains, Minds and Machines Summer Course
Towards complex language in partially observed environments MITCBMM 2022-09-01 | Stefanie Tellex, Brown UniversityPart of the Brains, Minds and Machines Summer Course 2022
A Conversation with Josh Tenenbaum MITCBMM 2022-09-01 | Josh Tenenbaum, MITPart of the Brains, Minds and Machines Summer Course 2022Links to videos to watch as prerequisites for this discussion:- youtube.com/watch?v=Qb1Mthx_DM4- youtube.com/watch?v=q9XR9Wbl7iI- https://cbmm.mit.edu/video/computational-models-cognition-reverse-engineering-common-sense-human-mind-and-brain-part-1- https://cbmm.mit.edu/video/computational-models-cognition-reverse-engineering-common-sense-human-mind-and-brain-part-2
Faces MITCBMM 2022-08-16 | Winrich Freiwald, The Rockefeller UniversityPart of the Brains, Minds and Machines Summer Course 2022
Error-driven Input Modulation: Solving the Credit Assignment Problem without a Backward Pass MITCBMM 2022-07-22 | CBMM researchers Giorgia Dellaferrera and Gabriel Kreiman discuss their latest paper, published in Proceedings of Machine Learning Research 2022 (https://proceedings.mlr.press/v162/dellaferrera22a.html), on neural networks using backpropogation and how they adapted them to be more biologically plausible.
Face neurons encode nonsemantic features MITCBMM 2022-06-23 | CBMM researchers Will Xiao and Gabriel Kreiman discuss their latest publication in PNAS and how their results suggest that so-called "face neurons" are better described as tuned to visual features rather than semantic categories.
Neurons detect cognitive boundaries to structure episodic memories in humans MITCBMM 2022-05-31 | CBMM researchers Dr. Jie Zheng and Prof. Gabriel Kreiman discuss their paper entitled "Neurons detect cognitive boundaries to structure episodic memories in humans" recently published in Nature Neuroscience on episodic memories where they show that neurons responded to abstract cognitive boundaries between different episodes.
GLMsingle: a toolbox for improving single-trial fMRI response estimates MITCBMM 2022-05-02 | Jacob Prince, MITAdvances in modern artificial intelligence have inspired a paradigm shift in human neuroscience, yielding large-scale functional magnetic resonance imaging (fMRI) datasets that provide high-resolution brain responses to tens of thousands of naturalistic visual stimuli. Because such experiments necessarily involve brief stimulus durations and few repetitions of each stimulus, achieving sufficient signal-to-noise ratio can be a major challenge. This tutorial will introduce GLMsingle, a scalable, user-friendly toolbox available in MATLAB and Python that enables accurate estimation of single-trial fMRI responses (glmsingle.org). Requiring only fMRI time-series data and a design matrix as inputs, GLMsingle integrates three techniques for improving the accuracy of trial-wise general linear model (GLM) beta estimates. First, for each voxel, a custom hemodynamic response function (HRF) is identified from a library of candidate functions. Second, cross-validation is used to derive a set of noise regressors from voxels unrelated to the experimental paradigm. Third, to improve the stability of beta estimates for closely spaced trials, betas are regularized on a voxel-wise basis using ridge regression. Validation analyses show that GLMsingle substantially improves the reliability of beta estimates across visually-responsive cortex, and that these improvements translate into tangible benefits for higher-level analyses relevant to systems and cognitive neuroscience. This tutorial will provide a practical overview of GLMsingle, aimed at making the toolbox accessible to a wide range of fMRI users. We will review key concepts in GLM modeling of fMRI data, discuss how the different components of GLMsingle seek to optimize different steps in the signal estimation procedure, and walk users through a Google Colab demo illustrating the ease of applying GLMsingle to example fMRI data.Slides: osf.io/7q5y3GitHub repo: github.com/cvnlab/GLMsinglePreprint: biorxiv.org/content/10.1101/2022.01.31.478431v1Example scripts: github.com/cvnlab/GLMsingle/tree/main/examples
ThreeDWorld (TDW) Tutorial MITCBMM 2022-04-13 | Jeremy Schwartz and Seth Alter, MITIn this tutorial, Jeremy Schwartz will walk us through the features and capabilities of ThreeDWorld (threedworld.org), a high-fidelity, multi-modal platform for interactive physical simulation. Next, Seth Alter will conduct a tutorial lab session. The repository is available here - github.com/alters-mit/tdw_bcs_demo/tree/main - (please note that it is not needed to run the code in advance — only requirement is to have Python 3.6 or higher installed).
Using AI to accelerate scientific discovery MITCBMM 2022-04-11 | Dr. Demis Hassabis, Founder and CEO of DeepMindAbstract: The past decade has seen incredible advances in the field of Artificial Intelligence (AI). DeepMind has been in the vanguard of many of these big breakthroughs, pioneering the development of self-learning systems like AlphaGo, the first program to beat the world champion at the complex game of Go. Games have proven to be a great training ground for developing and testing AI algorithms, but the aim at DeepMind has always been to build general learning systems ultimately capable of solving important problems in the real world. I believe we are on the cusp of an exciting new era in science with AI poised to be a powerful tool for accelerating scientific discovery itself. We recently demonstrated this potential with our AlphaFold system, a solution to the 50-year grand challenge of protein structure prediction, culminating in the release of the most accurate and complete picture of the human proteome.Speaker Biography: Demis Hassabis is the Founder and CEO of DeepMind, the world’s leading AI research company that aims to solve intelligence to advance science and benefit humanity.Founded in London in 2010, DeepMind has achieved breakthrough results in many challenging AI domains from Atari games to StarCraft II, and has published over 1000 research papers - including more than two dozen in Nature and Science. In 2016, DeepMind developed AlphaGo, the first program to beat a world champion at the complex game of Go. In 2020, its AlphaFold program was heralded as a solution to the 50-year grand challenge of protein structure prediction and in 2021, DeepMind launched the AlphaFold Protein Structure Database, which offers the most complete and accurate picture of the human proteome to date. A chess prodigy, Demis reached master standard aged 13, and went on to program the multi-million selling simulation game Theme Park aged 17. After graduating from Cambridge University in computer science, he founded pioneering videogames company Elixir Studios, and completed a PhD in cognitive neuroscience at University College London. Science listed his neuroscience research on imagination as one of 2007’s top ten breakthroughs, and in 2021, AlphaFold2 was selected as the Breakthrough of the Year.He is a Fellow of the Royal Society and the Royal Academy of Engineering. In 2017 he featured in the Time 100 list of most influential people, and in 2018 he was awarded a CBE for services to science and technology.
How fly neurons compute the direction of visual motion MITCBMM 2022-03-22 | Alexander Borst, Max-Planck-Institute of Neurobiology, Martinsried, GermanyAbstract: Detecting the direction of image motion is important for visual navigation, predator avoidance and prey capture, and thus essential for the survival of all animals that have eyes. However, the direction of motion is not explicitly represented at the level of the photoreceptors: it rather needs to be computed by subsequent neural circuits, involving a comparison of the signals from neighboring photoreceptors over time. The exact nature of this process represents a classic example of neural computation and has been a longstanding question in the field. Only recently, much progress has been made in the fruit fly Drosophila by genetically targeting individual neuron types to block, activate or record from them. Our results obtained this way demonstrate that the local direction of motion is computed in two parallel ON and OFF pathways. Within each pathway, a retinotopic array of four direction-selective T4 (ON) and T5 (OFF) cells represents the four Cartesian components of local motion vectors (leftward, rightward, upward, downward). Since none of the presynaptic neurons is directionally selective, direction selectivity first emerges within T4 and T5 cells. Our present research focuses on the cellular and biophysical mechanisms by which the direction of image motion is computed in these neurons.
Continuous-time deconvolutional regression: A method for studying continuous dynamics in naturali... MITCBMM 2022-02-28 | [full title] Continuous-time deconvolutional regression: A method for studying continuous dynamics in naturalistic dataCory Shain, MITAbstract: Naturalistic experiments are of growing interest to neuroscientists and cognitive scientists. Naturalistic data can be hard to analyze because critical events can occur at irregular intervals, and measured responses to those events can overlap and interact in complex ways. For example, words come quickly enough during naturalistic reading and listening that the brain responses to words likely overlap in time, and inherent variability in word durations can make these responses difficult to identify from data. In this tutorial, I will present continuous-time deconvolutional regression (CDR), a new approach to analyzing naturalistic time series. CDR uses machine learning to estimate impulse response functions from data, but, unlike established methods like finite impulse response modeling, these functions are defined in continuous time. CDR can therefore directly estimate event-related responses in a range of naturalistic experiment types, including fMRI, EEG/MEG, and behavioral measures. The tutorial will demonstrate how to define, fit, and evaluate CDR models, how to test hypotheses in the CDR framework, how to visualize patterns with CDR, and how CDR can be used to relax a range of assumptions about time series data. These steps can be run from the command line using an open-source Python library, with no need for users to write any code.The sample data+models for the CDR tutorial available here: dropbox.com/sh/yz4l5745nxz8h7w/AACWgDwfKjhkXlK0L5eM-d25a?dl=0. The files data.zip and models.zip need to be downloaded and extracted.CDR can be installed with either Anaconda or python+pip. To install with pip, run: pip install github.com/coryshain/cdr/archive/refs/tags/v0.5.3.tar.gzConda installation instructions are available in the readme at github.com/coryshain/cdr.
When and How CNNs Generalize to Out-of-Distribution Category-Viewpoint Combinations MITCBMM 2022-02-22 | CBMM researchers and authors, Spandan Madan, Xavier Boix, Helen Ho, and Tomotake Sasaki discuss their latest research that was published in Nature Machine Intelligence.Paper and more information can be found here - https://cbmm.mit.edu/publications/when-and-how-convolutional-neural-networks-generalize-out-distribution-category
Deep neural network models of sound localization reveal how perception is adapted to real-world ... MITCBMM 2022-01-27 | [full title] Deep neural network models of sound localization reveal how perception is adapted to real-world environments [publication release video]CBMM researchers and authors, MIT graduate student Andrew Francl and MIT Prof. Josh McDermott, discuss their latest research as published in Nature Human Behavior.
Early concepts of intimacy: Young humans use saliva sharing to infer close relationships MITCBMM 2022-01-20 | CBMM researchers and authors, MIT/Harvard postdoc Ashley Thomas and MIT Prof. Rebecca Saxe discuss their newly released paper published in Science.
CBMM Seasons Greetings 2021 MITCBMM 2022-01-11 | Season's Greetings and Best Wishes for the New Year from your friends at the Center for Brains, Minds, and Machines!
DNN models reveal interplay of peripheral coding and stimulus statistics in pitch perception MITCBMM 2021-12-15 | Perception is thought to be shaped by the environments for which organisms are optimized. These influences are difficult to test in biological organisms but may be revealed by machine perceptual systems optimized under different conditions. CBMM researchers, MIT graduate students Mark Saddler and Ray Gonzalez, and Prof. Josh McDermott, MIT, investigated environmental and physiological influences on pitch perception, whose properties are commonly linked to peripheral neural coding limits. Hear from Mark and Josh about how they were able to accomplish this.For the paper and more - https://cbmm.mit.edu/publications/deep-neural-network-models-reveal-interplay-peripheral-coding-and-stimulus-statistics
Building a state space for song learning MITCBMM 2021-12-08 | Michale Fee, MITAbstract: Songbird vocalizations are produced by a sparse sequence of spike bursts in a motor circuit that controls the vocal output on a fast (10ms) timescale. This sparse sequence is also transmitted to song learning circuits, presumably to control the temporal specificity of vocal learning, a process thought to proceed by mechanisms similar to reinforcement learning (RL). Electrophysiological recordings in young birds have revealed that such sequences do not exist at the earliest stages of learning, and emerge only gradually during song acquisition. How does this sparse temporal basis, or state space, emerge during development? Songbirds learn their vocalizations by imitating the song of an adult bird, suggesting that the auditory memory of the tutor song may play a role in setting up sequences in the motor system, creating a state space custom built for a given tutor song. I will describe a model for how temporal sequences to support RL of this complex behavioral pattern may be constructed in the brain, and will propose a hypothesis for how the auditory system could shape these sequences to align with a memory of the tutor song, thus facilitating song evaluation.
Selective responses to faces, scenes, and bodies in the ventral visual pathway of infants MITCBMM 2021-11-15 | MIT graduate student Heather Kosakowski and MIT Prof. Rebecca Saxe discuss their latest paper in Current Biology where they show that 2- to 9-month-old human infants have face-, scene- and body- selective responses in FFA, PPA, and EBA, respectively, constraining theories of cortical development.
Boundary conditions for language in biological and artificial neural systems MITCBMM 2021-11-12 | Andrea E. Martin, Lise Meitner Group Leader, Max-Planck Institute for PsycholinguisticsAbstract: Human language is a fundamental biological signal with computational properties that are markedly different than in other perception-action systems: hierarchical relationships between units (e.g., phonemes, morphemes, words, phrases), and the unbounded ability to combine smaller units into larger ones. These and other formal properties have long made language difficult to account for from a biological systems perspective, and within models of cognition. I focus on this foundational puzzle – essentially “what does a system need to represent information that is both algebraic and statistical?” - and discuss the computational requirements, including the role of neural oscillations across time, for what I believe is necessary for a system to represent and process language. I build on examples from cognitive neuroimaging data and computational simulations, and outline a developing theory that integrates basic insights from linguistics and psycholinguistics with the currency of neural computation, in turn demarcating the boundary conditions for artificial systems making contact with human language.Research website: www.andreaemartin.com.
Wobbling, drooping, bouncing - Visual perception of materials and their properties MITCBMM 2021-11-09 | Dr. Vivian Paulun, Kanwisher and Tenenbaum Labs, MITAbstract: Visual inference of material properties like mass, compliance, elasticity or fragility is crucial to predicting and interacting with our environment. Yet, it is unclear how the brain achieves this remarkable ability. How materials move, flow, fold or deform, depends not only on their internal properties but also on many external factors. For example, the observable behavior of an elastic bouncing object depends on its elasticity but also on its initial position and velocity. Estimating elasticity requires disentangling these different contributions to the observed motion. Predicting the future path of the object requires a forward simulation given the estimated latent parameters. I will present a set of experiments which we investigated how accurately human observers estimate the elasticity of bouncing objects or predict their future path. Furthermore, I will discuss the nature of the visual information observers use as well as the limitations of their internal model.
Neurons that structure memories of ordered experience in human MITCBMM 2021-10-27 | Gabariel Kreiman (on behalf of Jie Zheng), Children's Hospital Boston, Harvard Medical SchoolAbstract: The process of constructing temporal associations among related events is essential to episodic memory. However, what neural mechanism helps accomplish this function remains unclear. To address this question, we recorded single unit activity in humans while subjects performed a temporal order memory task. During encoding, subjects watched a series of clips (i.e., each clip consisted of 4 events) and were later instructed to retrieve the ordinal information of event sequences. We found that hippocampal neurons in humans could index specific orders of events with increased neuronal firings (i.e., rate order cells) or clustered spike timing relative to theta phases (i.e., phase order cells), which are transferrable across different encoding experiences (e.g., different clips). Rate order cells also increased their firing rates when subjects correctly retrieved the temporal information of their preferred ordered events. Phase order cells demonstrated stronger phase precessions at event transitions during encoding for clips whose ordinal information was subsequently correct retrieved. These results not only highlight the critical role of the hippocampus in structuring memories of continuous event sequences but also suggest a potential neural code representing temporal associations among events.
Visual Search Asymmetry: Deep Nets and Humans Share Similar Inherent Biases MITCBMM 2021-10-27 | Mengmi Zhang, Children's Hospital Boston/Harvard Medical SchoolAbstract: Visual search is a ubiquitous and often challenging daily task, exemplified by looking for the car keys at home or a friend in a crowd. An intriguing property of some classical search tasks is an asymmetry such that finding a target A among distractors B can be easier than finding B among A. To elucidate the mechanisms responsible for asymmetry in visual search, we propose a computational model that takes a target and a search image as inputs and produces a sequence of eye movements until the target is found. The model integrates eccentricity-dependent visual recognition with target-dependent top-down cues. We compared the model against human behavior in six paradigmatic search tasks that show asymmetry in humans. Without prior exposure to the stimuli or task-specific training, the model provides a plausible mechanism for search asymmetry. We hypothesized that the polarity of search asymmetry arises from experience with the natural environment. We tested this hypothesis by training the model on an augmented version of ImageNet where the biases of natural images were either removed or reversed. The polarity of search asymmetry disappeared or was altered depending on the training protocol. This study highlights how classical perceptual properties can emerge in neural network models, without the need for task-specific training, but rather as a consequence of the statistical properties of the developmental diet fed to the model. Our work will be presented in the upcoming Neurips conference, 2021.
The neural architecture of language: Integrative modeling converges on predictive processing MITCBMM 2021-10-25 | Graduate student Martin Schimpf (MIT/CBMM) and Prof. Evelina Fedorenko (MIT) discuss their latest publication in PNAS on predictive computational modeling and neural processing comparisons.
