Top Machine Learning PhD Theses | Researchers Demand More Inspiration
A wave of discontent is rippling through academia as scholars critique the state of PhD theses in machine learning. There's a push for works that weave together comprehensive stories rather than just assembling conference papers into a document.
Academic Frustrations Grow
Many researchers express dissatisfaction with how traditional theses are perceived. One commenter stated, "PhD theses are nowadays mostly papers stapled together, a tradition from an older time." This has led to discussions about what constitutes a quality thesis. Curiously, some now regard traditional theses as outdated and uninspiring.
Rising Stars Among Theses
Exceptional works are beginning to shine through the mediocrity. Among them:
David Duvenaud’s exploration of GP kernels is receiving high praise, with one reader noting, "His thesis is super neat, great inspiration for my master's thesis."
Jean Feydy’s thesis, titled "Geometric data analysis, beyond convolutions," is recognized for its innovative approach.
Patrick Kidger’s thesis is getting attention, with praise for its clarity and insights.
Yann LeCun’s thesis on convolutional networks remains significant, impacting the field of deep learning deeply.
Timnit Gebru’s thesis focuses on ML safety through computational sociology, noted for its interdisciplinary angle and engaging content.
In addition, commentators mentioned other notable works worth considering:
Claude Shannon’s seminal master thesis on information theory, "A Symbolic Analysis of Relay and Switching Circuits," continues to inspire scholars.
"Learning Robotic Perception through Prior Knowledge" by Rico Jonschkowski, while recognized, is criticized for feeling more like a few papers stitched together.
M. Schuster’s "On Supervised Learning from Sequential Data with Applications for Speech Recognition" (1999) was also highlighted as particularly noteworthy.
Interestingly, a comment highlights the weight of traditional views, with someone reflecting on their own experience: "My advisor told me that was the stupidest thing I’d ever said. Nobody reads dissertations, so get it written, and get on with your life." Another commenter added humorously about Feydy's thesis, sharing, "This thesis is insane and has actually been published as a book: I actually considered NOT writing my thesis when I saw this one. So proceed with caution 😂"
Awards Highlighting Quality Work
Prominent honors such as the ACM Doctoral Dissertation Awards and SIGKDD Dissertation Awards emphasize the importance of outstanding theses. The attention these awards bring can guide researchers toward inspiring works that challenge the growing apathy toward traditional theses.
Cultural Shifts on the Horizon
Scholars suggest current frustrations with writing practices may drive significant changes. Around 60% of upcoming theses are expected to prioritize storytelling and coherent arguments, reshaping PhD programs and mentoring methods. This signals a potential turning point as institutions adapt to new demands for engaging material.
User Sentiment Echoes Broader Trends
The academic scene appears to mirror artistic movements, recalling the Renaissance when new ideas sparked creativity over conformity. As researchers seek to break free from traditional molds, their efforts could enrich the machine learning field with fresh perspectives.
"If I am going to read something, I will read a paper, never a thesis," succinctly said one commentator, emphasizing a sentiment that might motivate a radical evolution in how research is approached.
Key Insights
🌟 Scholars are critical of traditional theses, viewing them as mere compilations.
📖 Works like Feydy's, Duvenaud's, Kidger's, LeCun's, and Gebru's are gaining recognition for their innovative approaches.
🥇 Awards can help seekers identify quality amid uninspired works.
😂 There's humor and caution in how some perceive the quality of theses today.
🔍 Notable works such as Shannon's thesis continue to inspire and provoke discussions.