Tensorflow Developer Certification - Resources and tips

Tensorflow Developer Certification is backed by Google Developers & Tensorflow itself. If you have come to this page I would assume you have a brief idea about the 'what & why of Tensorflow' and hence wouldn't get into those details here.

The basic information about the 5-hr hands-on practical exam and its expectations are covered in their official handbook.

Attention to input format/shapes & instructions

Carefully evaluate the format and shape of data as it flows from the input layer to all subsequent layers. Read carefully the problem-specific instructions on what should not be changed or should not be done, e.g., not normalizing the data, retaining the dimensions etc.,

Colab Pro

Though completely optional, having Colab Pro helped me in shortening the training times with GPU + High RAM. Shortened training times mean you could have more attempts with tuning the hyperparameters within the given timeframe.

Resources

Though my preparation for this specific certification started a few weeks ago, my foundation in machine learning was, a couple of years ago, through Andrew Ng's amazing course which is available for free at Coursera as well as on Youtube. Doing Udemy courses such as this from Frank Kane in my earlier stages also helped to get a simplified view of ML concepts.

Some of the resources that helped me clear this certification, and I am very thankful for are:

• https://machinelearningmastery.com/what-are-word-embeddings/
• https://machinelearningmastery.com/how-to-develop-a-convolutional-neural-network-from-scratch-for-mnist-handwritten-digit-classification/
• https://towardsdatascience.com/hands-on-nlp-deep-learning-model-preparation-in-tensorflow-2-x-2e8c9f3c7633
• https://towardsdatascience.com/time-series-analysis-on-multivariate-data-in-tensorflow-2f0591088502
• https://towardsdatascience.com/how-not-to-use-machine-learning-for-time-series-forecasting-avoiding-the-pitfalls-19f9d7adf424
• https://www.kaggle.com/code/nicholasjhana/multi-variate-time-series-forecasting-tensorflow/notebook
• https://www.relataly.com/stock-market-prediction-using-multivariate-time-series-in-python/1815/
• https://amitness.com/2020/02/tensorflow-hub-for-transfer-learning/
• http://introtodeeplearning.com/
• https://www.tensorflow.org/tutorials/keras/keras_tuner
• https://neptune.ai/blog/keras-tuner-tuning-hyperparameters-deep-learning-model
• https://www.kdnuggets.com/2020/03/tensorflow-keras-tokenization-text-data-prep.html
• https://github.com/nicholasjhana/tensorflow-certification-study-guide

Practice, practice, and practice :)

Nothing can prepare you better than solving as many ML/NN problems as possible, more importantly, typing the code instead of copy-pasting. You would be amazed at how many questions pop up when you try to code from scratch. Those questions would help get a lot of clarity which is much needed in the exam.

There are a variety of problems that you can pick and choose from Kaggle - try to solve at least 1-2 problems in each of the 5 categories that the exam expects you to master (NLP, Timeseries, Image classification, etc.)

I hope this helps!