Beyond imitation: Zero-shot task transfer on robots by learning concepts as cognitive programs

Lázaro-Gredilla, Miguel; Lin, Dianhuan; Guntupalli, J. Swaroop; George, Dileep

doi:10.1126/scirobotics.aav3150

Cited by 53 publications

(50 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To avoid this effect, further research on learning the concepts themselves would be desirable. Some attempts have been initially applied in visual concepts, which mention the possibility of applying them to different applications, such as textual or auditory scenarios [83].…”

Section: Applications In Technological Financial and Other Successfmentioning

confidence: 99%

Deep Learning and Big Data in Healthcare: A Double Review for Critical Beginners

et al. 2019

View full text Add to dashboard Cite

In the last few years, there has been a growing expectation created about the analysis of large amounts of data often available in organizations, which has been both scrutinized by the academic world and successfully exploited by industry. Nowadays, two of the most common terms heard in scientific circles are Big Data and Deep Learning. In this double review, we aim to shed some light on the current state of these different, yet somehow related branches of Data Science, in order to understand the current state and future evolution within the healthcare area. We start by giving a simple description of the technical elements of Big Data technologies, as well as an overview of the elements of Deep Learning techniques, according to their usual description in scientific literature. Then, we pay attention to the application fields that can be said to have delivered relevant real-world success stories, with emphasis on examples from large technology companies and financial institutions, among others. The academic effort that has been put into bringing these technologies to the healthcare sector are then summarized and analyzed from a twofold view as follows: first, the landscape of application examples is globally scrutinized according to the varying nature of medical data, including the data forms in electronic health recordings, medical time signals, and medical images; second, a specific application field is given special attention, in particular the electrocardiographic signal analysis, where a number of works have been published in the last two years. A set of toy application examples are provided with the publicly-available MIMIC dataset, aiming to help the beginners start with some principled, basic, and structured material and available code. Critical discussion is provided for current and forthcoming challenges on the use of both sets of techniques in our future healthcare.

show abstract

Section: Applications In Technological Financial and Other Successfmentioning

confidence: 99%

Deep Learning and Big Data in Healthcare: A Double Review for Critical Beginners

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Grid cell outputs provide a periodic tiling of uniform space, which is advantageous for learning and navigating maps when other sensory cues are absent. Similarly encoding snapshots from a graphical model for vision [37] as the input to this sequencer might enable the learning of visuo-spatial concepts and visual routines [38], and model the bi-directional influence hip-275 pocampus has on the visual cortex. We believe these ideas are promising paths for future exploration.…”

mentioning

confidence: 99%

Learning cognitive maps as structured graphs for vicarious evaluation

Rikhye¹,

Gothoskar²,

Guntupalli³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Cognitive maps enable us to learn the layout of environments, encode and retrieve episodic memories, and navigate vicariously for mental evaluation of options. A unifying model of cognitive maps will need to 5 explain how the maps can be learned scalably with sensory observations that are non-unique over multiple spatial locations (aliased), retrieved efficiently in the face of uncertainty, and form the fabric of efficient hierarchical planning. We propose learning higher-order graphs -structured in a specific way that allows efficient learning, hierarchy formation, and inference -as the general principle that connects these different desiderata. We show that these graphs can be learned efficiently from experienced sequences using a 10 cloned Hidden Markov Model (CHMM), and uncertainty-aware planning can be achieved using messagepassing inference. Using diverse experimental settings, we show that CHMMs can be used to explain the emergence of context-specific representations, formation of transferable structural knowledge, transitive inference, shortcut finding in novel spaces, remapping of place cells, and hierarchical planning. Structured higher-order graph learning and probabilistic inference might provide a simple unifying framework for un-15 derstanding hippocampal function, and a pathway for relational abstractions in artificial intelligence.properties of place cells and grid cells [8]. Yet another recent model casts spatial and non-spatial problems as a connected graph with neural responses as efficient representations of this graph [9]. Unfortunately, 30 both these models fail to explain several experimental observations such as the discovery of place cells that encode routes [10,11], remapping in place cells [12], a recent discovery of place cells that do not encode goal value [13], and flexible planning after learning the environment.Here, we propose that learning higher-order graphs of sequential events might be an underlying principle of cognitive maps, and propose a specific representational structure that aids in learning, memory integration, 35 retrieval of episodes, and navigation. In particular, we demonstrate that this representational structure can be represented as a probabilistic sequence model -the cloned Hidden Markov Model (CHMM). We show that sequence learning in CHMMs can explain a variety of cognitive maps phenomena such as discovering spatial maps from random walks under aliased and disjoint sensory experiences, transferable structural knowledge, finding shortcuts, and hierarchical planning and physiological findings such as remapping of place cells, 40 and route-specific encoding. Notably, all these properties emerge from a simple model that is easy to train, scale, and perform inference on. Cloned Hidden Markov Model as a model of cognitive mapsCHMMs are based on Dynamic Markov Coding (DMC) [14], an idea for representing higher-order sequences by splitting, or cloning, observed states. For example, a first order Markov chain representing the 45 sequences A-C-E and B-C-D will also assi...

show abstract

“…Problems in perception still need dynamic inference, which means that the reasoning components will need to go all the way down to sensory regions, so that perception and cognition can work together. In our opinion, hybrid models are more likely to be a combination of graphical models, graph-structured neural networks, causal inference, and probabilistic programs (Lázaro-Gredilla et al, 2019). Neural networks will help to accelerate inference and learning in many parts of these hybrid models.…”

Section: Do Hybrid Models Imply Neural Network For Perception and Symentioning

confidence: 99%

From CAPTCHA to Commonsense: How Brain Can Teach Us About Artificial Intelligence

George¹,

Lázaro-Gredilla²,

Guntupalli³

2020

Front. Comput. Neurosci.

Self Cite

View full text Add to dashboard Cite

Despite the recent progress in AI powered by deep learning in solving narrow tasks, we are not close to human intelligence in its flexibility, versatility, and efficiency. Efficient learning and effective generalization come from inductive biases, and building Artificial General Intelligence (AGI) is an exercise in finding the right set of inductive biases that make fast learning possible while being general enough to be widely applicable in tasks that humans excel at. To make progress in AGI, we argue that we can look at the human brain for such inductive biases and principles of generalization. To that effect, we propose a strategy to gain insights from the brain by simultaneously looking at the world it acts upon and the computational framework to support efficient learning and generalization. We present a neuroscience-inspired generative model of vision as a case study for such approach and discuss some open problems about the path to AGI.

show abstract

Beyond imitation: Zero-shot task transfer on robots by learning concepts as cognitive programs

Cited by 53 publications

References 57 publications

Deep Learning and Big Data in Healthcare: A Double Review for Critical Beginners

Deep Learning and Big Data in Healthcare: A Double Review for Critical Beginners

Learning cognitive maps as structured graphs for vicarious evaluation

From CAPTCHA to Commonsense: How Brain Can Teach Us About Artificial Intelligence

Contact Info

Product

Resources

About