Nazanin 🐹

@nazaninn

2495 Following

1208 Followers

Nazanin 🐹 pfp

Day 33 #100daysofMachineLearning Dived deeper into transfer learning, although just on a conceptual level. Also learnt about multi-task learning which is a pretty cool technique used in autonomous driving (self-driving cars) to learn multiple tasks with a single neural network.

0 reply

0 recast

0 reaction

Nazanin 🐹 pfp

Day 32 #100daysofMachineLearning I finally decided to start working on kaggle projects. I should have started this a while back, but anyways we're here now. I really hope to stay consistent with it. It's a great way to practice what I've learnt and build projects alongside.

0 reply

0 recast

122 reactions

Nazanin 🐹 pfp

Day 31 #100daysofMachineLearning Amongst other things, I learnt about error analysis and how it can be used in improving the performance of your model. It's a manual way of checking and analyzing the cause of wrong predictions and using the insights to improve your model.

1 reply

0 recast

155 reactions

Nazanin 🐹 pfp

Day 30 #100daysofMachineLearning Already started the third course of the DL specialization which is on "Structuring Machine Learning Projects". This course moves beyond all the technicalities and focuses on the strategy and best approach to apply when working on ML projects.

0 reply

1 recast

317 reactions

Nazanin 🐹 pfp

Day 29 #100daysofMachineLearning Completed the second course of the Deep Learning Specialization 🚀.

0 reply

0 recast

315 reactions

Nazanin 🐹 pfp

Day 28 #100daysofMachineLearning Learnt about the two major school of thoughts on hyperparamter tuning (pandas vs cavier approach). You can either choose babysit one model and monitor its performance or train multiple models in parallel and pick the one that works best.

2 replies

0 recast

315 reactions

Nazanin 🐹 pfp

🌀 Cast and get 3 free spins to earn points. Points will be exchanged for tokens at the end of the 1st season. 🫂 Get 3 more spins for each friend you invite. Read the «Rules» section to learn more. ☀️ Enter Onchain Summer with /pill

0 reply

0 recast

1 reaction

Nazanin 🐹 pfp

Day 27 #100daysofMachineLearning Rewatched the tuning process video to gain a better intuition. Noted that the learning rate is the most important hyperparameter to tune during training. Also learnt about the methods used in selecting the best values for the hyperparameters.

1 reply

0 recast

270 reactions

Nazanin 🐹 pfp

⛓ All trenches lead onchain

0 reply

0 recast

3 reactions

Nazanin 🐹 pfp

Day 26 #100daysofMachineLearning Completed the programming assignment for the optimization algorithms. Over to the last week of the course 🚀.

0 reply

0 recast

246 reactions

Nazanin 🐹 pfp

Day 25 #100daysofMachineLearning Started the programming exercise for optimization algorithm. I was at least able to do something before network started acting up again. Hoping to finish up with it tomorrow.

0 reply

0 recast

238 reactions

Nazanin 🐹 pfp

Day 24 #100daysofMachineLearning Still didn't do much today, but at least was able to watch videos on batch normalization. Batch norm is basically used to normalize the data (Z) in the hidden layers before passing it to the activation function (A) for faster training.

0 reply

0 recast

218 reactions

Nazanin 🐹 pfp

Day 23 #100daysofMachineLearning Wasn't able to take the programming assignment on the optimisation algorithms due to the network issue. But I was at least able to watch a video on hyperparamter tuning process in week 3. I honestly didn't pick up the idea in the video.

0 reply

0 recast

211 reactions

Nazanin 🐹 pfp

Day 22 #100daysofMachineLearning Wrapped up with the remaining optimization algorithms today (GD with momentum, RMSprop, and Adam optimization algorithm). Took my time to understand their implementations. Also learnt about learning rate decay and why it is used.

2 replies

1 recast

216 reactions

Nazanin 🐹 pfp

Day 21 #100daysofMachineLearning So today was packed with a lot of learning. Got to drill down into Gradient Descent (GD) and its types. I looked at: - Batch GD - Mini-batch GD - Stochastic GD

1 reply

0 recast

204 reactions

Nazanin 🐹 pfp

Day 20 #100daysofMachineLearning Got introduced to gradient checking, which is a technique used to ensure that the backpropagation implementation is correct. Also finished the quiz and last programming assignment for week 1.

1 reply

0 recast

115 reactions

Nazanin 🐹 pfp

Day 19 #100daysofMachineLearning Today was all about regularization. Learnt to implement L2 reg and dropout reg (inverted dropout) and got to understand how to use these techniques in reducing overfitting. Also completed the programming exercise on regularization.

1 reply

0 recast

106 reactions

Nazanin 🐹 pfp

Day 18 #100daysofMachineLearning Finished the first programming exercise for week 1 of the second course. Learnt how to implement zero, random and He initializations (designed for layers with ReLu activation functions) in code.

0 reply

0 recast

80 reactions

Nazanin 🐹 pfp

Day 17 #100daysofMachineLearning Picked up the second course in the DL specialization on improving deep neural networks. In this video, I learnt that DL is a highly iterative process and it's almost impossible to accurately get the right set of hyperparameters at first guess.

0 reply

0 recast

56 reactions

Nazanin 🐹 pfp

Day 16 #100daysofMachineLearning Finished the first course in the Deep Learning specialization. What a course! 1 down, 4 more to go 🚀. #deeplearning #coursera

1 reply

1 recast

60 reactions