Generalizing in the Real World with Representation Learning

2025-05-06 0 0 8.19MB 152 页 10玖币

侵权投诉

POLYTECHNIQUE MONTRÉAL

aﬃliée à l’Université de Montréal

Generalizing in the Real World with Representation Learning

TEGAN MAHARAJ

Département de génie informatique et génie logiciel

Thèse présentée en vue de l’obtention du diplôme de Philosophiæ Doctor

Génie informatique

Mai 2022

©Tegan Maharaj, 2022.

arXiv:2210.09925v1 [cs.LG] 18 Oct 2022

POLYTECHNIQUE MONTRÉAL

aﬃliée à l’Université de Montréal

Cette thèse intitulée :

Generalizing in the Real World with Representation Learning

présentée par Tegan MAHARAJ

en vue de l’obtention du diplôme de Philosophiæ Doctor

a été dûment acceptée par le jury d’examen constitué de :

Michel DESMARAIS, président

Christopher PAL, membre et directeur de recherche

Aishwarya AGRAWAL, membre

Margaret MITCHELL, membre externe

iii

“We want AI agents that can discover like we can, not which contain what we have

discovered.”

– Rich Sutton, The Bitter Lesson (2019)

“There is something at work in my soul, which I do not understand.”

– Mary Shelley, Frankenstein, (1818)

“Curiously enough, the only thing that went through the mind of the bowl of petunias

as it fell was Oh no, not again. Many people have speculated that if we knew exactly

why the bowl of petunias had thought that we would know a lot more about the nature

of the Universe than we do now.”

– Douglas Adams, The Hitchhiker’s Guide to the Galaxy, (1979)

DEDICATION

This thesis is dedicated to all those once and future of Mila,

for being part of that emergent wonderland

imperfect utopia of creative and collaborative science.

Especially

to Chris,

for taking a chance on me,

and to David,

for believing in me.

ACKNOWLEDGEMENTS

I have the great good fortune to write the ﬁnal drafts of this thesis as an Assistant Professor

at the University of Toronto. A huge thank you to all my new colleagues for welcoming me to

the Faculty of Information, Schwartz Reisman Institute, and Vector - I’m excited and proud

to start the next chapter of my career in such an illustrious and interdisciplinary place.

It’s easy looking back from my dream job to say all the work in this thesis (and much more

done during my degree that didn’t ﬁt in here) was worth it. But perhaps more surprisingly,

it would also have been easy for me to do that at almost any point in the last 6 years. Things

were hard sometimes, but I learned so much, and met so many amazing people in this journey

- while I doubted myself at times, I don’t think I ever doubted that I was doing the best

thing I could be. It’s the ﬁrst time I’ve felt that way for years at a time, and it’s a pretty

great feeling. For this, I am grateful to more people than I can cover here, but I’m going to

give it the old college try.

I am ﬁrst and foremost grateful to my family for their invaluable love and support - including

all my extended family and family by choice. From a young age you’ve made me feel the

world is full of kind, interesting, intelligent, creative people who will be there for each other

and for me. I’m lucky to have you all. I’m especially grateful to my parents, Laurie and

Ken, for always taking the time to describe, explain, educate, and discuss; to Keir, Aaricia,

Amiani, Amy, Kathleen, Carlos, Chuck, Corinne, Jordan, Ben, and Lisa, for making me feel

at home wherever I am; and to David, Don, Val, Susan, Clint, and Gretchen for welcoming

me into your families.

I’m grateful to Yoshua Bengio, for setting up a lab truly founded on the best of science -

curiosity-driven, egalitarian, and humanitarian. I knew I wanted to apply to Mila (then Lisa)

after reading Representation Learning [Yoshua Bengio, Aaron Courville, Pascal Vincent], but

I might never have done so without the encouragement of several wonderful people - thank you

Guillaume Alain, Umut Şimşekli, Orhan Firat, Chris Maddison, and Jamie Kiros for taking

the time to talk with an interested Master’s student. Thank you especially to Jeﬀ Clune, for

making me believe it was possible, and for almost a decade of long-distance friendship. Even

after all that encouragement, I might have given up on the whole thing if not for Kyle Gill -

Thank you.

I’m so grateful to everyone at Mila, for helping me feel at home in the ﬁeld of AI and in the

wider world. This is a special place. Particular shoutouts to:

·Chris Beckham for breaking the Mila ice with me;

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

POLYTECHNIQUEMONTRÉALaliéeàl'UniversitédeMontréalGeneralizingintheRealWorldwithRepresentationLearningTEGANMAHARAJDépartementdegénieinformatiqueetgénielogicielThèseprésentéeenvuedel'obtentiondudiplômedePhilosophiæDoctorGénieinformatiqueMai2022©TeganMaharaj,2022.POLYTECHNIQUEMONTRÉALaliéeàl'Universi...

展开>> 收起<<

Generalizing in the Real World with Representation Learning.pdf

共152页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Generalizing in the Real World with Representation Learning

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: