AdaGrad Gradient Accumulation
3 explanations from your library.
10:30Prof. Prabir Kumar Biswas, IIT Kharagpur
From AdaGrad to RMSProp
“sum of the squares of the past gradients where this past gradient starts from time t equal to 0, ok. So, you are basically the operation that was done in Adagrad algorithm is r t”
Closest moment in this session: it teaches the idea without listing it as a key concept.
21:34IIT Madras (NPTEL)
Gradient Descent Variants and Momentum
“which will and then the Nesterov Accelerated Gradient, AdaGrad, AdaDelta, also RMSProp very similar initially, okay.”
Closest moment in this session: it teaches the idea without listing it as a key concept.
30:09IIT Madras (NPTEL)
Learning Rate Decay
“Epoch is when you have gone through your entire data set, epoch is when you have gone through your entire data set once, that is one epoch, right. Let us say you are using stochastic gradient descent or mini batch gradient descent, you have to run through your entire data set and that would be considered as one epoch.”
Closest moment in this session: it teaches the idea without listing it as a key concept.
Cloudinary under the hood2 Cloudinary URLs make this page. No render servers: each one is generated on request and cached.
Compare reel
3 explanations of this concept, from different sessions, spliced into one video with fl_splice. Each clip is labelled with its speaker.
…/video/upload/so_29,eo_44,w_1280,h_720,c_fill/l_video:pravaha:ff2c7367-5718-4b86-9d17-04328cf12206,fl_spl…/so_92.5,eo_108.1,w_1280,h_720,c_fill/fl_layer_apply/fl_layer_apply,g_north_west,x_40,y_40,so_30.6,eo_48.5/f_auto:video,q_auto/pravaha/1ba747ae-bdb0-43b5-bd11-8a0eda32a02c.mp4
- so_29
- starts at 29 s · and 5 more like it
- eo_44
- ends at 44 s · and 5 more like it
- w_1280
- 1280 px wide · and 2 more like it
- h_720
- 720 px tall · and 2 more like it
- c_fill
- crops to fill the frame exactly · and 2 more like it
- l_video:pravaha:ff2c7367-57…
- another video (pravaha/ff2c7367-5718-4b86-9d17-04328cf12206) as a layer
- fl_splice
- joins the next clip onto the end of this one · and 1 more like it
- fl_layer_apply
- places the layer defined just before · and 4 more like it
- l_video:pravaha:2b17046e-fb…
- another video (pravaha/2b17046e-fb2d-4a2c-ad2c-e84a76f8c28a) as a layer
- l_text:arial_34_bold:1%20%C…
- text layer “1 · Prof. Prabir Kumar Biswas” · and 2 more like it
- co_white
- text colour white · and 2 more like it
- b_rgb:0f766ecc
- background #0f766ecc · and 2 more like it
- g_north_west
- anchored to the top-left corner · and 2 more like it
- x_40
- 40 px from the side · and 2 more like it
- y_40
- 40 px from the edge · and 2 more like it
- f_auto:video
- best format for each device (e.g. AV1, WebM, MP4, WebP)
- q_auto
- AI-chosen quality: smallest file that still looks right
Explanation thumbnail
The frame at that moment, cropped around the speaker by AI.
…/video/upload/so_30.5,c_fill,ar_16:9,w_640,g_auto/f_auto,q_auto/pravaha/1ba747ae-bdb0-43b5-bd11-8a0eda32a02c.jpg
- so_30.5
- starts at 30.5 s
- c_fill
- crops to fill the frame exactly
- ar_16:9
- aspect ratio 16 : 9
- w_640
- 640 px wide
- g_auto
- AI picks the focus (the speaker or slide), not a blind centre crop
- f_auto
- best format for each device (e.g. AV1, WebM, MP4, WebP)
- q_auto
- AI-chosen quality: smallest file that still looks right
Want a guided order, basics first? Turn AdaGrad Gradient Accumulation into a short course →