Building Neural Network Models That Can Reason

Автор: Microsoft Research

Загружено: 2019-07-30

Просмотров: 29634

Описание:

Deep learning has had enormous success on perceptual tasks but still struggles in providing a model for inference. To address this gap, we have been developing networks that support memory, attention, composition, and reasoning. Our MACnet and NSM designs provide a strong prior for explicitly iterative reasoning, enabling them to learn explainable, structured reasoning, as well as achieve good generalization from a modest amount of data. The Neural State Machine (NSM) design also emphasizes the use of a more symbolic form of internal computation, represented as attention over symbols, which have distributed representations. Such designs impose structural priors on the operation of networks and encourage certain kinds of modularity and generalization. We demonstrate the models’ strength, robustness, and data efficiency on the CLEVR dataset for visual reasoning (Johnson et al. 2016), VQA-CP, which emphasizes disentanglement (Agrawal et al. 2018), and our own GQA (Hudson and Manning 2019). Joint work with Drew Hudson.

See more at https://www.microsoft.com/en-us/resea...

Building Neural Network Models That Can Reason

Доступные форматы для скачивания:

Скачать видео mp4

Информация по загрузке:

Скачать аудио mp3

Похожие видео

Efficient Robot Skill Learning: Grounded Simulation Learning and Imitation Learning from Observation

Efficient Robot Skill Learning: Grounded Simulation Learning and Imitation Learning from Observation

Quantum Computing for Computer Scientists

Quantum Computing for Computer Scientists

Но что такое нейронная сеть? | Глава 1. Глубокое обучение

Но что такое нейронная сеть? | Глава 1. Глубокое обучение

LLM и GPT - как работают большие языковые модели? Визуальное введение в трансформеры

LLM и GPT - как работают большие языковые модели? Визуальное введение в трансформеры

The Thousand Brains Theory

The Thousand Brains Theory

Как сжимаются изображения? [46 МБ ↘↘ 4,07 МБ] JPEG в деталях

Как сжимаются изображения? [46 МБ ↘↘ 4,07 МБ] JPEG в деталях

Language Understanding and LLMs with Christopher Manning - 686

Language Understanding and LLMs with Christopher Manning - 686

Microsoft Research Forum | Season 2, Episode 2

Microsoft Research Forum | Season 2, Episode 2

From Deep Learning of Disentangled Representations to Higher-level Cognition

From Deep Learning of Disentangled Representations to Higher-level Cognition

Dendrites: Why Biological Neurons Are Deep Neural Networks

Dendrites: Why Biological Neurons Are Deep Neural Networks

Краткое объяснение больших языковых моделей

Краткое объяснение больших языковых моделей

Machine Learning Zero to Hero (Google I/O'19)

Machine Learning Zero to Hero (Google I/O'19)

MSR Distinguished Lecture Series: First-person Perception and Interaction

MSR Distinguished Lecture Series: First-person Perception and Interaction

MIT AGI: Building machines that see, learn, and think like people (Josh Tenenbaum)

MIT AGI: Building machines that see, learn, and think like people (Josh Tenenbaum)

The Harsh Reality of Being an ML Researcher

The Harsh Reality of Being an ML Researcher

Christopher D Manning: A Neural Network Model That Can Reason (ICLR 2018 invited talk)

Christopher D Manning: A Neural Network Model That Can Reason (ICLR 2018 invited talk)

Преломление и «замедление» света | По мотивам лекции Ричарда Фейнмана

Преломление и «замедление» света | По мотивам лекции Ричарда Фейнмана

Как Сделать Настольный ЭЛЕКТРОЭРОЗИОННЫЙ Станок?

Как Сделать Настольный ЭЛЕКТРОЭРОЗИОННЫЙ Станок?

Примет ли Россия сделку или это все блеф?

Примет ли Россия сделку или это все блеф?

The Power of Self-Learning Systems

The Power of Self-Learning Systems