00:16:11 Anon. Capacitance: yes 00:17:14 Anon. VC Dimension: Object detection 00:17:29 Anon. Transformer: spatial invariance 00:19:10 Anon. Kalman Filter: distribution 00:20:22 Anon. ResNet50: shared parameters 00:20:24 Anon. pdb.set_trace(): It shared weights 00:22:59 Anon. pdb.set_trace(): Is this similar to the idea of YOLO object detection architecture 00:24:17 Anti (TA): YOLO is a CNN, if that is what you are asking 00:25:07 Anon. pdb.set_trace(): I am wondering about the architecture of the YOLO CNN 00:25:12 Anon. Transformer: Are all SSDs CNN based? 00:27:01 Mansi Anand (TA): YOLO CNNs are much more deeper networks. The idea behind is CNN itself. https://medium.com/@ODSC/overview-of-the-yolo-object-detection-algorithm-7b52a745d3e0#:~:text=YOLO%20is%20a%20clever%20convolutional,and%20probabilities%20for%20each%20region. 00:27:19 Anon. pdb.set_trace(): Thank you! 00:27:38 Mansi Anand (TA): @Sai what exactly are you referring to SSDs here? 00:27:58 Anon. Transformer: Single shot detector networks 00:28:04 Anti (TA): the vast majority of those used in practice are cnns 00:28:39 Anon. Transformer: Cool,thanks! 00:29:07 Mansi Anand (TA): Similar to YOLO, again it is CNN based deeper architecture. https://arxiv.org/pdf/1512.02325.pdf 00:32:35 Anon. RCNN: the subnets share the parameters 00:40:46 Anon. pdb.set_trace(): Thats with a stride of 1 right? 00:48:29 Anti (TA): I missed the specifics of that example; if we are moving one pixel at a time (regardless of image patch size), the stride is 1 00:50:26 Anon. PCA: Cable News Network hmmm 00:52:32 Anon. Matrix: the video of these experiments are on youtube, for anyone interested 00:54:45 Anon. Transformer: No 01:04:56 Mansi Anand (TA): 10 more secs 01:06:23 Anon. pdb.set_trace(): No 01:06:25 Anon. is_available(): no 01:06:55 Anon. is_available(): what about the Pokemon neuron? pretty sure that's a thing 01:07:10 Anon. SGD: I think it is the same funda 01:12:55 Anon. Sum-Product: distributed scanning 01:17:25 Anon. VC Dimension: Why all of the s cells in the first layer focus on the same region? 01:18:12 Anon. SVM: fires together => wires together 01:18:13 Anon. Dot Product: I don’t see a pattern of sharing parameters in this structure, am I understanding correctly? 01:19:48 Anon. SVM: it wouldn’t learn? 01:23:46 Mansi Anand (TA): 10 more secs 01:24:12 Anon. Validate: does this mean that each object has its own C-cell? 01:25:21 Anon. SVM: infer from context what it should be 01:33:37 Anon. VC Dimension: How does the model manage to learn anything without supervision? 01:34:29 Anon. grad_fn: Not a TA, but I was thinking about that too. I think these outputs could be inputs to another part of the brain. 01:35:52 Anon. grad_fn: That other part may be a decision layer. 01:36:18 Anon. SVM: So the more the eyes see something, the better it recognizes it? 01:36:21 Anon. Dot Product: Does Fukushima consider the process from visual neurons connecting to brain cells? We conveniently added a decision layer to make the model learnable, but in actual human brain it doesn’t work this way right? 01:38:27 Anon. Dot Product: Ok 01:38:33 Anon. VC Dimension: Thank you