Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
York University associate professor Kosta Derpanis shared an early version of lecture slides that use cartoon animations, including a robot on a pixelated globe, to walk through policy gradients and reinforcement learning concepts such as cumulative rewards, trajectory gradients, and Monte Carlo approximations.
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
Derpanis notes the slides are in development because computer vision and robotics research are converging, though no full course rollout date or additional materials have been announced yet.
The 87-second preview highlights log-derivative tricks and reward objectives through simple character-driven scenes rather than static equations.
Users praise the professor sharing early Policy Gradients lecture slides for being insightful and courageous, with comments calling the material amazing and appreciating the honesty in the creation process.
#KostasThoughts: I’m a HUGE fan of using visual analogies as section title screens. They serve as visual bookmarks, helping the audience navigate the material while reinforcing a key idea and adding a touch of humour 😉
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
#KostasThoughts: One of my goals this fall is to get back into my home studio, start recording lectures, and share them publicly. At this point, creating lecture slides and giving in-person lectures come fairly easily. Recording and editing lectures, however, still feels daunting 😬
@jbhuang0604 has been a major inspiration. I really enjoy his online lecture series. Top-quality content 💪
Check out his YouTube channel!
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
#KostasThoughts
My lecture motto: Your eyeballs are mine.
One idea at a time. The focus of the slide should always match the focus of the speech.
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
#KostasThoughts: Someone recently asked me about automating slide creation. For me, creating slides is part of the learning process. Breaking a topic into its building blocks and putting it back together helps me understand it.
I hope my students learn from the lecture, but I learn a great deal through the creation (and presenting) process as well.
That said, I have started using automated tools for ideation, evaluating my slide compositions, and generating images and videos.
TLDR: If I don’t understand it, I won’t present it.
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
Jia-Bin's YouTube Channel: https://www.youtube.com/c/JiaBinHuang
#KostasThoughts: One of my goals this fall is to get back into my home studio, start recording lectures, and share them publicly. At this point, creating lecture slides and giving in-person lectures come fairly easily. Recording and editing lectures, however, still feels daunting 😬
@jbhuang0604 has been a major inspiration. I really enjoy his online lecture series. Top-quality content 💪
Check out his YouTube channel!
@CSProfKGD This looks amazing!!
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉
@CSProfKGD It still feels daunting for me. Putting yourself out there takes courage.
Looking forward to your lectures! We need more content creators!
#KostasThoughts: One of my goals this fall is to get back into my home studio, start recording lectures, and share them publicly. At this point, creating lecture slides and giving in-person lectures come fairly easily. Recording and editing lectures, however, still feels daunting 😬
@jbhuang0604 has been a major inspiration. I really enjoy his online lecture series. Top-quality content 💪
Check out his YouTube channel!
Someone recently asked me about automating slide creation. For me, creating slides is part of the learning process. Breaking a topic into its building blocks and putting it back together helps me understand it.
I hope my students learn from the lecture, but I learn a great deal through the creation (and presenting) process as well.
That said, I have started using automated tools for ideation, evaluating my slide compositions, and generating images and videos.
TLDR: If I don’t understand it, I won’t present it.
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉

@CSProfKGD This is the opposite of brainrot. Love it.

@CSProfKGD "...creating slides is part of the learning process." -- so true!💯
York University associate professor Kosta Derpanis shared an early version of lecture slides that use cartoon animations, including a robot on a pixelated globe, to walk through policy gradients and reinforcement learning concepts such as cumulative rewards, trajectory gradients, and Monte Carlo approximations.
Back to the lecture prep grind. VERY early version of my lecture on Policy Gradients.
With #computervision and #robotics research appearing to converge, at least based on #CVPR2026 and my European tour conversations, it feels like good timing 😉