Thread by @mparmar47

D1 of #50daysofudacity
I finished up to Lesson 2.19
My notes can be found here for quick refernce
docs.google.com/document/u/1/d…

D2 of #50daysofUdacity
I finished up to Lesson 2.25
Also completed lab assignment for a linear regression model to predict the price of taxi in new york city
My notes can be found here for quick reference

docs.google.com/document/u/1/d…

D3 of #50daysofudacity
I finished Lesson 2
Also completed lab assignment for linear regression model to predict the price of taxi in new york city
My notes can be found here for quick reference
docs.google.com/document/u/1/d…

D4 of #50daysofUdacity
I finished up to Lesson 3.10
Completed 2 assignment on data preparation (transformation, versioning)

My notes can be found here for quick reference
docs.google.com/document/u/1/d…

D4 of #50daysofUdacity
Key Learning
- Data quality is paramount for ML model performance
- Drift in Data can hamper model accuracy and its essential to monitor it
- Azure dataflow access is easy to use and powerful for data preparation

D5 of #50daysofUdacity
I finished up to Lesson 3.16
Completed a lab on feature engineering and selection with trained model on bike rent prediction using 2 types of model

docs.google.com/document/u/1/d…

D5 of #50daysofUdacity
Key Learning
- Feature engineering is crucial to develop the good ML models
- Feature selection is also important to improve accuracy of #ML models to avoid curse of dimensionality

D6 of #50daysofUdacity
I finished up to Lesson 3.20
#Ai #ML #Azure
docs.google.com/document/u/1/d…

D6 of #50daysofUdacity
#Ai #ML #Azure
Key Learning
- Data Drift needs to be monitored to sustain performance of model
- AZURE ML is machine learning managed services and have many component to support it

D6 of #50daysofUdacity
#Ai #ML #Azure
Key Learning
- Classification is important type of ML task which has 3 sub classes based on output: binary out, one among many output class, multiple among many output class

D7 & D8 of #50DaysofUdacity
#Ai #ML #Azure
I finished up to Lesson 3.32
Also completed 3 labs

docs.google.com/document/u/1/d…

D7 & D8 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Regression evaluation matrix is different then of classifier. RMSE, MAE, Rsquared and spearman correlation
- Strength in number is a concept to leverage wisdom of many models to reduce biases and improve accuracy

D7 & D8 of #50DaysofUdacity
#Ai #ML #Azure
Key learning
- Ensemble mechanism has 3 types boosting (reduce bias), bagging (reduce variance) and stacking (multiple different models)

D7 & D8 of #50DaysofUdacity
#Ai #ML #Azure
Key learning
- Automated machine learning, also referred to as automated ML or AutoML, is the process of automating the time consuming, iterative tasks of machine learning model development.

D9 of #50DaysofUdacity
#Ai #ML #Azure
I completed lesson 3.
My notes can be found here bit.ly/azmlnote

D9 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Data preparation and #management tasks
- #Feature Engineering
- Monitoring #data drift
- Model #training paradigms and flow of it
- Evaluation Processes for #ML and relevant metrics
- #Ensemble learning
- #AutoML

D9 of #50DaysofUdacity
#Ai #ML #Azure

All key learnings were practiced in 6 lab sessions on AZURE ML Studio

3 Lesson completed 4 to go

D10 of #50DaysofUdacity
#Ai #ML #Azure
I completed lesson 4.9
1 Lab is pending (spent 30 minutes in debugging, the issue with back end)

My notes can be found here
bit.ly/azmlnote

D10 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Supervised learning classification generally preformed on tabular data, image & audio and text data (data need to be converted to numerical format)

D10 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Supervised learning classification has 3 types (binary, multi class single label, multi class multilabel)
- 3 main algorithms in AzureML for multi class algorithms (logistic regression, neural network, decision tree forest)

D11 of #50DaysofUdacity
#Ai #ML #Azure
I completed lesson 4.16 and also finished 4 labs in total.
Created PseudoCode to understand #AutoML process in @Azure #ML Studio in easier way.

Here are my notes
bit.ly/azmlnote

D11 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Supervised learning regression generally preformed on tabular data, image & audio and text data (data need to be converted to numerical format)

D11 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- 3 main algorithms in AzureML for regression algorithms (linear regression, neural network, decision tree forest)

D12 of #50DaysofUdacity
#Ai #ML #Azure
I completed lesson 4 and also finished all labs
Here are my notes
bit.ly/azmlnote
Key Learning
-AutoML has many options and its worth exploring it
-Supervised regression has 3 algorithms- regression, decision trees, neural networks

D12 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Unsupervised learning learns relationship of data without any labeled dataset
- Semi supervised learning combines supervised and unsupervised approaches

D12 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Clustering algorithms are unsupervised in nature and has 4 types (centroid based, density based, distribution based, hierarchy based)

D13 of #50DaysofUdacity
#Ai #ML #Azure
I completed up to lesson 5.8 and also finished 1 labs
Here are my notes
bit.ly/azmlnote
Key Learning
- DL is subset of ML which is a subset of AI. All DL is ML but other way around is not true

D13 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- Artificial neural network are inspired from human brain but does not resemble them
- DL is capable of handling large data and learning complex arbitrary function on its own including automatic feature extraction

D13 of #50DaysofUdacity
#Ai #ML #Azure
Key Learning
- DL has many applications across domains like image classification, text translation, speech recognition, autonomous driving, etc.

D14 of #50daysofudacity
#AI #ml #Azure
I completed up to lesson 5.14 and also finished 1 labs.

Here are my notes
bit.ly/azmlnote
Posting here for people to refer notes. Happy to answer any questions.

D14 of #50daysofudacity
#AI #ml #Azure
Key Learning
- Apart from 3 main approaches (supervised, unsupervised, reinforcement) for ML, there are specialized cases
- Similarity learning, Text classification, Feature learning, Anomaly detection, forecasting are specialized cases

D14 of #50daysofudacity
#AI #ml #Azure
Key Learning
- #Recommendation system is a special case of similarity learning
- Recommendation system has 2 approaches content based and collaborative filtering

D15 and D16 of #50daysofUdacity
#AI #ML #AZURE
I completed up to lesson 5 and all labs.
Here are my notes
bit.ly/azmlnote

D16 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- #Anomaly detection is a special class of ML algorithm which is difficult to model due to high imbalance of data
- Anomaly works with supervised (#classification) and unsupervised (#clustering) approaches

D16 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- Anomaly detection has many applications (condition monitoring, fraud detection, intrusion detection, outlier detection)
-Forecasting is another special class of ML algorithm which works with orderable datasets (time or events)

D16 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- #Forecasting predicts the next data points based on time series data
- #ARIMA, #RNN, #LSTM, #GRU, Multi-variate regression, Prophet (#Facebook), ForecastTCN(#Microsoft) are set of algorithm used for forecasting

D17 & 18 of #50daysofUdacity
#AI #ML #AZURE
I completed up to lesson 6.16 and 3 labs
(forgot to update status yesterday but notes are updated)
Here are my notes
bit.ly/azmlnote

D17 & 18 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- #MachineLearning Managed Services makes ML development process easier by managing underlying environment, compute and other services
- Compute #clusters to be chosen based on need for training or deployment purposes

D17 & 18 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- Managed #notebook environments are amazing where all 5 steps from #data to #deployment of #model can be achieved

D17 & 18 of #50daysofUdacity
#AI #ML #AZURE
Key Learning
- To use MLMS effectively, there is a need to understand modeling of work flow and automate it with #MLOps and pipelines (#devops)

D19 of #50daysofUdacity
#AI #ML #AZURE
I completed up to lesson 6, 7 and 8; also completed all labs.
I have finished the course also. 😀🎊🎊
Here are my notes
bit.ly/azmlnote

D20, D21 and D22 #50daysofUdacity
#AI #ML #AZURE
I have done revision of all my notes
bit.ly/azmlnote (forgot to update it as i was feeling unwell)

D23-29 #50daysofUdacity
#AI #ML #AZURE
I have done revision of all notes and completed specific lab exercises to check impact of other options and models on final results. Also started preparing for Study Jam Webinar on topic of Future of #Devops and #Mlops ; #innovation

D30 #50daysofUdacity
#AI #ML #AZURE
I have prepared webinar and relevant flyers for upcoming #studyjam #pyJamming on 2 topics
- Future of #DevOps and #MLOps
- #Innovation & #Entrepreneurship

Here are flyers

D31 of #50daysofudacity learnt about #agile and #DevOps in preparation of my #webinar on Future of #DevOps and #mlops.
#pyjammimg #PoweredbyMicrosoft #PoweredbyUdacity

D32 of #50daysofudacity
Read multiple blogs on aspects of #AI #strategy, #ML, #NLP for #SEO, and #DevSecOps
sloanreview-mit-edu.cdn.ampproject.org/c/s/sloanrevie…
towardsdatascience.com/amazon-wants-t…
github.blog/2020-08-13-sec…
searchenginejournal.com/natural-langua…

D33 of #50daysofudacity
Actively participated on slack channel to answer many questions on #AI, #ML, #security and other aspects (close to 4 hr & 60 conversation)
Presented webinar on future of #DevOps & #mlops.

(slide deck link in description of video)

D33 of #50daysofudacity
Presented another webinar on topic of #innovation and #entrepreneurship

(slide deck in description of video)

Really enjoyed both session and participants also enjoyed session based on feedback.

D34 of #50daysofudacity
Attended a #hc32 conference tutorial on Scaling #DeepLearning models for training and inference.
The majority of large models are coming from #NLP domain and driving need for larger and Distributed Computing.

D35 of #50daysofudacity
attended #hc32 conference. Lot of interesting work is going on in space of hardware to catch up with #AI & #ML compute demands. Here is an interesting snapshot from @intel keynote "No Transistor Left behind" on topic of generality and performance debate

D35 of #50daysofudacity
attended #hc32 conference. Many groups are working on exciting hardware accelerators to support compute hungry #AI/#ML #algorithms

D36-40 of #50daysofudacity
Revised my notes and redone labs from lesson 3-4- and 5

D41 of #50daysofudacity
Read interesting blog post in #AI and #ml
Data science portfolio towardsdatascience.com/a-complete-dat…

Gartner #hypecycle 2020
forbes.com/sites/louiscol…

#ai #Intelligence explosion
forbes.com/sites/lanceeli…

Visualising #value
paulminors.com/blog/visualisi…