00:14:26 Reshmi Ghosh (TA): Mute yourselves when you join! 00:18:53 Anon. Threshold: Best local linear approimation 00:19:13 Anon. Hodgkin-Huxley: gradient of function 00:19:13 Anon. Variance: The change in x with respect to the change in y 00:19:14 Anon. Sample: Rate of change of a function 00:19:16 Anon. Sigmoid: slope of a function 00:19:18 Anon. ICA: Rate of change of a variable with respect to another 00:19:28 Anon. YOLOv2: the change of a function dependent on change of inputs 00:19:36 Anon. DFS: a s.t. delta y = a * delta x 00:23:15 Anon. Variance: Alpha will be a vector 00:23:17 Anon. Algorithm: derivative would be a vector 00:23:47 Anon. Voltage-gate: gradient 00:23:47 Anon. Algorithm: derivative in all directions of the input 00:23:57 Anon. Variance: You need one output 00:24:03 Anon. Voltage-gate: column vector 00:24:05 Anon. Variance: So you need to do a dot product 00:24:20 Anon. LR: column vector 00:24:52 Anon. YOLOv3: is f a vector in this question? 00:25:01 Anon. YOLOv2: derivative is a vector with each component for each input. derivative is transpose of f(x) if vector 00:25:21 Jinhyung David Park (TA): @Nicky just a scalar 00:25:24 Anon. Supervised: Nx1 00:25:29 Anon. Markov Chain: n*1 00:25:51 Anon. YOLOv3: the derivative must be a row vector 00:26:11 Anon. Whyami: 1x N 00:26:14 Anon. Supervised: 1*N 00:26:16 Anon. LR: 1XN 00:29:07 Anon. Layer: alpha1*delx1 00:30:04 Anon. YOLOv3: just to be clear, if f is not a scalar and is instead a vector, then the shape of alpha would no longer be a 1xn correct? 00:31:07 Anon. YOLOv3: for example if f was 2x1 alpha would be 2xn 00:33:29 Anxiang Zhang (TA): 7 seconds 00:34:37 Anon. Autograd: I wasn't able to see the pole. Why was that ? 00:34:46 Reshmi Ghosh (TA): That is weird 00:35:20 Reshmi Ghosh (TA): We will share the questions later on, but if this problem persists let us know. 00:35:40 Anon. Autograd: Ok thank you, I will. 00:36:25 Anon. Algorithm: no inflection point 00:39:45 Anon. Weight: 0 00:39:47 Anon. Markov Chain: 0 00:40:41 Anon. Layer: constant value 00:40:42 Anon. Giant Squid Neuron: 0 00:44:14 Anon. Indifferentiable: tangent hyperplane to the function? 00:44:40 Anon. Giant Squid Neuron: Tangent vector 00:44:45 Anon. Autograd: The direction of fastest increase 00:44:47 Anon. Max Pool: direction increase faster 00:44:48 Anon. Markov Chain: the direction to which you should travel along that can increase the value of y most quickly 00:47:15 Anon. Layer: 0 00:47:15 Anon. Supervised: 0 00:47:15 Anon. Unsupervised: 0 00:47:16 Anon. Algorithm: 0 00:47:17 Anon. Voltage-gate: 0 00:47:17 Anon. Notagrad: 0 00:47:30 Anon. Residual: 90 00:47:30 Anon. Algorithm: 90 00:47:31 Anon. Loss Function: 90 00:47:32 Anon. Supervised: 90 00:47:33 Anon. Weight: 180 00:47:34 Anon. Unsupervised: 180 00:47:38 Anon. Giant Squid Neuron: 180 00:47:39 Anon. Loss Function: 180 00:47:41 Anon. Residual: 180 00:49:02 Anon. Giant Squid Neuron: Same direction 00:49:09 Anon. Weight: Same as the delta vector 00:53:56 Reshmi Ghosh (TA): Poll folks 00:54:11 Anon. Autograd: I can't see it please 00:54:22 Reshmi Ghosh (TA): Damn 00:54:27 Anxiang Zhang (TA): 5 seconds 00:54:35 Reshmi Ghosh (TA): I will get back to you after the lecture @Dereje 00:54:49 Anon. Sample: Try updating your zoom? 00:54:51 Anon. Transformer: since we’re waiting, I think the slides posted online are not the same as what are being shown 00:55:03 Reshmi Ghosh (TA): Yes they do get changed 00:55:26 Reshmi Ghosh (TA): We modify things until the last minute:) because who doesn’t? 00:55:35 Anon. Dropout: This is Last part of slides for Lec 3 00:56:08 Reshmi Ghosh (TA): BY THE WAY, When Bhiksha says I won’t explain it but will use it ~ potential quiz question. You should refer these slides back. Although no guarantee that it will appear in Quiz 3 (as I haven’t seen the questions myself ):P 00:57:57 Anon. Voltage-gate: - 00:57:58 Anon. Spiking NN: negative 00:57:58 Anon. Markov Chain: neg 00:57:58 Anon. Unsupervised: -ve 00:58:00 Anon. Weight: negative 00:58:01 Anon. Supervised: neg 00:58:10 Anon. Spiking NN: right 00:58:43 Anon. Weight: + 00:58:45 Anon. Supervised: + 00:58:55 Anon. Giant Squid Neuron: -< 00:58:58 Anon. Weight: left 00:58:59 Anon. Supervised: left 00:59:02 Anon. LTI: left 00:59:04 Anon. Leakage: left 01:00:29 Anon. Voltage-gate: xk - step*deltax 01:03:14 Anon. Dropout: large 01:03:23 Anon. Giant Squid Neuron: big 01:03:25 Anon. Markov Chain: big 01:03:41 Anon. LR: small 01:03:43 Anon. Leakage: mall 01:03:43 Anon. Voltage-gate: small 01:03:53 Anon. LR: overshhot 01:11:14 Anon. YOLOv2: step 01:38:21 Anxiang Zhang (TA): 5 seconds