00:29:28 Anon. Scalar: yep 00:29:30 Anon. PaddedSequence: yes 00:29:31 Anon. Convolution: yes 00:29:33 Anon. Hyperplane: yes 00:30:03 Anon. Matrix: yes 00:32:49 Anon. Loss Function: I don't see the button to "raise my hand" 00:33:01 Anon. print(‘Hello world!’): Click on participants 00:33:03 Anon. Fourier Transform: In the Reactions 00:33:21 Tony Qin (TA): It’s the ‘raise hand’ one you see when you click on participants 00:33:26 Anon. Loss Function: Gotcha 00:33:29 Reshmi Ghosh (TA): Raised hand feature is in the participators window 00:33:47 Anon. Kirchhoff: Q: The form 11-785 Group Forms is no longer accepting responses. Will we still be automatically assigned a group? 00:33:54 Anon. Eigenvector: Hi I have filled the form for study groups but I haven't received any communication regarding how to form one.. 00:34:06 Reshmi Ghosh (TA): You will be notified about groups shortly 00:34:26 Reshmi Ghosh (TA): @nikita email Sean P., the TA. He is responsible for student groups 00:34:35 Anon. Kirchhoff: Thanks 00:35:03 Anon. Eigenvector: Thanks 00:35:09 Anon. Jacobian: will the poll answers carry points? 00:35:20 Reshmi Ghosh (TA): yes 00:35:22 Anon. Voltage-gate: The link for poll is on the website? 00:35:30 Reshmi Ghosh (TA): Nope, the poll will be on zoom 00:35:34 Reshmi Ghosh (TA): Stay AWAKE!! :D 00:36:23 Anon. Max Pool: will the polls be graded for completion or correctness? 00:38:01 Anon. Instance: I thought this course has no partication score? How many percentage of scores the polls will carry? 00:39:01 Anon. Instance: cool, thanks 00:39:36 Miao Wang (TA): They will be used for attendance tracking and help you find out how you performed during the class. 00:39:59 Anon. Momentum: Will we be notified about polls in advance? Due to time difference, I may not be able to attend every lecture and have to watch the recordings instead. Thanks! 00:40:36 Reshmi Ghosh (TA): If you are in class only then you respond to polls. They appear anywhere doing the lecture time 00:41:19 N Miya Sylvester (TA): We are able to see whether you have watched lecture videos, and in the case that you cannot attend in person we will be checking to see if you have reviewed the lecture recordings online. 00:41:27 Reshmi Ghosh (TA): If you cannot attend lectures, and instead choose to watch recording, we track how many lectures you watch, and the length of your watching time 00:43:26 Anon. Backprop: When do we need to watch the recordings? Do we need to watch it on the same day as the lecture so attendance could be counted? 00:43:40 N Miya Sylvester (TA): No you can watch it at your convenience. 00:43:51 N Miya Sylvester (TA): Weekly quizzes are based on weekly lectures 00:44:01 Anon. Backprop: Thanks 00:45:14 Anon. Args: will quiz zero be based on this weeks lectures or the topics of HW 0 00:45:26 Reshmi Ghosh (TA): HW0 00:45:45 Reshmi Ghosh (TA): It’s basic Python and AWS 00:46:16 Anon. Optimizer: What is the zoom link for today’s TA office hour? 00:46:54 Reshmi Ghosh (TA): Should be updated shortly. 00:47:08 Anon. ResNet34: IBM 00:47:08 Anon. Center Loss: CMU 00:47:09 Anon. Kirchhoff: Google 00:47:09 Anon. Whyami: cmu 00:47:10 Anon. Quadratic: Wein Hall 00:47:11 Anon. print(‘Hello world!’): ibm? 00:47:11 Anon. Sum-Product: IBM 00:47:11 Anon. Sequence: ibm 00:47:12 Anon. Sanger’s Rule: ibm 00:47:13 Anon. Sparse Matrix: cmu 00:47:15 Anon. Eigenvector: ibm 00:47:16 Anon. Momentum: IBM 00:47:16 Anon. Kernel Trick: IBM 00:47:16 Anon. PaddedSequence: Google Deep Mind? 00:47:16 Anon. Gradient Flow: google 00:47:17 Anon. Thalamus: cmu 00:47:18 Anon. Loss Function: Google 00:47:21 Anon. Convolution: google 00:47:22 Anon. Inception: CMU? 00:47:23 Anon. Dot Product: deepmind? 00:47:23 Anon. Sodium Ion: cmu 00:47:25 Anon. RNN: CMU 00:47:25 Anon. Jacobian: cmu 00:47:34 Anon. Multivariate Gaussian: IBM 00:47:42 Anon. Train: cmu 00:47:52 Anon. Matrix: IBM 00:47:53 Anon. Backprop: cmu 00:47:58 Anon. Gabor transforms: cmu 00:49:05 Anon. Autoencoder: is it possible for you to publish the script online? 00:49:14 N Miya Sylvester (TA): The office hour zoom should be the same as the lecture zoom link 00:50:06 N Miya Sylvester (TA): Oh wait, it seems we changed that. We’ll let you know what the zoom office hour link is on Piazza 00:51:52 Miao Wang (TA): yeah, we will update the OH zoom link on piazza and website shortly 00:52:07 Anon. Center Loss: looool 00:52:14 Anon. Instance: ..... 00:53:21 Anon. Baseline: ? 00:55:12 Anon. K-Fold: Socrates? 00:55:19 Anon. Attractor: 1 00:55:20 Anon. Eigenvector: Plato? 00:55:21 Anon. Thalamus: 1 00:55:22 Anon. hello_world.py: yes 00:55:22 Anon. ResNet34: +1 00:55:22 Anon. Loss Function: 1 00:55:23 Anon. Deep Dream: yeah 00:55:23 Anon. EC2: 1 00:55:24 Anon. Hessian: +1 00:55:24 Anon. Convolution: Yep seen before 00:55:24 Anon. Eigenvector: me 00:55:24 Anon. Neurotransmitter: +1 00:55:24 Anon. YOLOv2: +1 00:55:24 Anon. PaddedSequence: 1 00:55:24 Anon. Lipid Bilayer: +1 00:55:24 Anon. print(‘Hello world!’): yes 00:55:25 Anon. Inception: 1+ 00:55:25 Anon. Bidirectional: +! 00:55:27 Anon. Autoencoder: +1 00:55:27 Anon. Bidirectional: +! 00:55:28 Anon. Sum-Product: +1 00:55:28 Anon. RCNN: +1 00:55:31 Anon. Loss Function: Seen it in my art history class 00:55:31 Anon. Matrix: +1 00:55:31 Anon. Args: +1 00:55:31 Anon. Multivariate Gaussian: +1 00:55:31 Anon. Dot Product: yes 00:55:32 Anon. Alpha: +1 00:55:33 Anon. ALBERT: +1 00:55:33 Anon. Recall Capacity: yes 00:55:33 Anon. Kalman Filter: 1 00:55:34 Anon. is_leaf: +1 00:55:35 Anon. Scalar: +1 00:55:36 Anon. Train: yes 00:55:49 N Miya Sylvester (TA): There’s a yes/no button as well as the raise hand 00:57:02 Anon. Fourier Transform: you are muted 01:02:35 Anon. ResNet34: I can see the entire screen 01:02:36 Anon. RNN: yes 01:02:41 Anon. ALBERT: yes 01:02:47 Anon. Node of Ranvier: Professor's voice sometimes became very blurry. I wonder if professor could use a better microphone in later courses? Thanks a lot! 01:03:04 Anon. Baseline: I suggest a blue snowball 01:03:09 Anon. Drop Connection: I second 01:03:12 Anon. Args: I third 01:03:17 Anon. Autoencoder: i fourth 01:03:32 Anon. Lipid Bilayer: i fifth 01:03:38 Anon. Connectionist: I sixth 01:04:04 Anon. Deep Learner: I seventh 01:04:10 Anon. Sum-Product: I eighth 01:04:19 N Miya Sylvester (TA): We’ll suggest it to him for you all after class 01:04:50 Anon. Lipid Bilayer: i second 01:05:40 Anon. vim: +1 01:05:41 Rita Singh (Prof): Ok, folks, we have a choice we can try a much better mic now or fix it properly for the next lecture, I think with 200+ listeners the audio isn't what is was when we tested it out with a few people 01:08:49 Anon. Oja’s Rule: dont think it’s bad enough to warrant stopping 01:09:07 Anon. Refractory Period: ^ 01:09:17 Anon. ResNet18: ^ 01:09:35 Anon. Axon: Will the online videos have subtitles? 01:09:44 Anon. Instance: .... 01:09:57 Anon. Instance: lmao 01:10:09 Tony Qin (TA): There should be a transcript 01:10:10 N Miya Sylvester (TA): No I don’t believe so, we would need automatic transcription 01:10:53 Reshmi Ghosh (TA): Media tech has a transcript. So when the recorded video is uploaded, you can go check the parts you had trouble with 01:10:58 Anon. Axon: It's not so bad now.. just in case some parts need to be reviewed after. 01:11:10 Reshmi Ghosh (TA): ANSWER everyone!!! 01:11:11 Miao Wang (TA): poll is up, tic toc, folks! 01:11:17 Anon. Faster-RCNN: Tik tok? 01:11:32 Reshmi Ghosh (TA): No that is a different avenue XD 01:11:38 Anon. Baseline: So, did we agree this was based on completion (though I do know the answer) 01:11:38 Anon. Args: courses on tik tok when? 01:11:56 Anon. RBM: I forgot to submit 01:11:57 Anon. Matrix: some chats are so many and disturbing 01:12:30 Anon. Test: ^ Agreed 01:12:37 Anon. Sum: Well actually I have to read Professor’s lip to figure out what he is saying… 01:12:49 Anon. RBM: Can we get informed by closing the poll? 01:12:59 Anon. Sum-Product: Modified Harvard 01:13:00 Anon. AlphaGo: X86-64 01:13:01 Anon. Attractor: x86 01:13:03 Anon. Dropout: x86/64 01:13:04 Anon. Undirected Edge: X86? 01:13:06 Anon. Asynchronous Update: Von Neumann 01:13:07 Anon. Center Loss: Havard sth? 01:13:08 Anon. YOLOv4: RISC 01:13:08 Anon. Adam: von newmann 01:13:15 Anon. Eigenvector: von neumann 01:13:15 Anon. batch_first: Von neumann 01:13:16 Anon. Thalamus: Von newmann 01:13:17 Anon. Center Loss: Looooool 01:13:19 Anon. RCNN: Von neumann 01:13:22 Anon. Matrix: on neumann 01:13:23 Anon. AlphaGo: von neumann 01:13:29 Tony Qin (TA): Von Neumann/Princeton Machine 01:13:34 Anon. Train: von neumann 01:13:34 Anon. ALBERT: von neumann 01:13:36 Anon. Recall Capacity: Von Neumann 01:13:37 Anon. Hessian: Von neumann 01:13:38 Anon. PaddedSequence: von neumann 01:13:42 Anon. hello_world.py: von neumann 01:13:42 Anon. Decoder: von neumann 01:13:45 Anon. Python: von neumann 01:13:49 Anon. Dot Product: von neumann 01:13:49 Anon. Lipid Bilayer: von neumann 01:13:49 Anon. Node of Ranvier: von neumann 01:13:51 Anon. Multivariate Gaussian: Von Neumann 01:13:53 Anon. Neurotransmitter: Vom Neumann 01:14:05 Anon. XOR Gate: Von Neumann 01:14:08 Anon. Derivative: Von Neumann 01:14:09 Anon. Momentum: Von Neumann 01:14:17 Anon. Args: no one's mentioned it but I think it might be Von Neumann 01:15:49 Anon. Jacobian: cannot hear the audio 01:16:18 Anon. GAN: I think it is fine for others? I can hear it right now 01:16:23 Reshmi Ghosh (TA): Check connection? I am able to 01:16:28 Anon. Args: i can hear 01:16:29 Anon. Jacobian: ok 01:19:14 Anon. Autoencoder: lol 01:19:17 Anon. Transformer: loool 01:19:18 Anon. K-Fold: lol 01:19:20 Anon. Center Loss: LOL! 01:19:32 Anon. Args: fat brains rise up 01:19:35 Anon. Recall Capacity: Fat is helpful it seems :P 01:20:01 Anon. Center Loss: Revenge of the fatheads 01:20:01 Anon. vim: Fat head broys 01:27:19 Anon. Sum-Product: You can’t decrease weights 01:27:23 Anon. GAN: The weights only increase 01:27:24 Anon. Convolution: when they gets too close? 01:28:20 Reshmi Ghosh (TA): Answer now 01:28:23 Miao Wang (TA): POLL is up :) 01:29:01 Anon. Validate: (writes a neural network to predict when the polls are 01:30:20 Anon. Backprop: I may forget to use an Andrew email to sign in Zoom, but I did register. Does that affect my attendance? 01:31:57 Anon. Loss Function: Is η a constant? 01:32:14 Tony Qin (TA): Yes. It’s the learning rate 01:32:24 Anon. ResNet34: I don't think the learning rate has to be constant in practice? 01:32:28 Anon. Perceptron: learning rates are not necessarily constant 01:33:03 Tony Qin (TA): Correct. The learning rate can be changed over time 01:33:15 Anon. EC2: you can use dynamic learning rate 01:33:29 Anon. vim: [Poggers] 01:34:05 N Miya Sylvester (TA): Xiaoqing - Your name is recorded for attendance when you enter the zoom meeting 01:35:44 Reshmi Ghosh (TA): Poll time! 01:38:12 Anon. Loss Function: What is the motivation for adding a bias term? 01:38:48 Anon. ALBERT: I think we need timer for quiz. 01:38:59 Anon. Dot Product: so that the decision boundary doesn't have to cross the origin 01:39:06 Anon. Instance: bias describes how easily a neroun will fire 01:39:47 Miao Wang (TA): we planned 30 seconds for the polls, it worked well in the test run. 01:41:06 N Miya Sylvester (TA): Weights decide how fast activation functions are triggered, and the bias is used to delay triggering — it depends on what’s best for your data 01:41:29 Anon. K-Fold: Will the chat be saved so that people watching the video on media tech can see any questions and answers? (or you could compile the questions into a Piazza Post for the lesson) 01:42:01 Anon. EC2: agree, will the answer of questions be posted later? 01:42:53 Anon. Momentum: Can we have a countdown for the poll? 01:43:19 Anon. Scheduler: how do we interpret the poll result? in the previous question the 6 was highlighted red, but the solution was 7? 01:43:40 Tony Qin (TA): Let me find out, Haofan. We may just post something on Piazza 01:43:57 Anon. Sum: Professor mentioned it was 7 01:44:01 Anon. Soma: Red is majority vote 01:44:41 Miao Wang (TA): Yes, maybe we could also give a concrete timing for polls 01:44:59 Anon. Node of Ranvier: can someone give a brief explain for why the answer is 7? 01:45:09 Anon. Scheduler: would also be helpful if either 1) poll shows correct result 2) professor/TA verbally says answer after every poll 01:45:10 Anon. Asynchronous Update: Same question here 01:45:20 Anon. GAN: 6 for each side 1 for combining the results 01:45:23 Anon. Baseline: You need one extra neuron as the output neuron 01:45:30 Anon. Keras: 6 for each edge of the hexagon and 1 to AND the 6 inequalities 01:45:40 Anon. Node of Ranvier: cool, thanks 01:45:41 Anon. EC2: 6 neurons for boundary, 1 neuron for voting 01:45:44 Anon. Hessian: I believe the professor says 6for each point and 1 for output 01:45:49 Reshmi Ghosh (TA): Hexagon so 6 sides and then there is one final neuron required for the output 01:46:00 Anon. Asynchronous Update: Thank you 01:46:30 Reshmi Ghosh (TA): POLL PEOPLE! 01:46:34 Miao Wang (TA): POLL is up guys 01:47:03 Anon. Train: can we have a timer for poll? 01:47:43 Miao Wang (TA): Yeah, we do have a timer for poll 01:48:18 Anon. Instance: I also think we need a timer for the poll such that we could see how much time left. For now, I tend to randomly select an answer in case I will miss it 01:48:38 Anon. Kirchhoff: +1 01:48:41 Anon. Train: +1 01:48:45 Anon. Autoencoder: +1