Indian Retail Vision Startups - #Infilect #Wesense #Tangoeye

 

Keep Exploring!!!

CAP - #Consistency, #Availability, #PartitionTolerance

 #Consistency,

• When True, I remember exactly everything, I will recall the same details every time
• When False, I remember approximately before the latest info, I may have a poor memory

#Availability,

• When True, I will attend your call everything, You will get a reply (When you impress your partner)
• When False, I may not respond when I sleep (Post marriage no reply on every call)

#PartitionTolerance,

• When True, You have two networks/numbers to reach me, When one number is not available another number you can reach me, You have an alternate network
• When False, When one number is not available, Network issue, You cannot reach me

Keep Exploring!!!

Redshift Stored Proc Examples

Code Snippet from slides

Keep Exploring!!!

Transformers - Lesson 2

One-liners - Summary from link

• Lesson #1 - Sentences numerically represented a 1 and 0 on occurence of work, This is one hot encoding
• Lesson #2 - Sequences, when we setup, every next word would be possible combinations/ words divided by count, is the probability, First-word sequence
• Lesson #3 - If we know multiple words in sequence it would be easier / more confident to nail down on possible words
• Lesson #4 - When you consider word by word its is next word, When you keep sequences it becomes easier, Instead of one word consider it like triplets, bi-gram, tri-gram remember sequences in multiple combinations
• Lesson #5 - Create sequences skipping words
• Lesson #6 - Embeddings to leverage similar words
• Lesson #7 - Positional encoding to consider positions/location in the embedding space

I still need few more iterations but this this is first cut understanding.

Keep Exploring!!!

Attention - Lessons

From Application Developer vs Knowing how it works, Still trying to figure out, Be it backpropagation or Network design or Attention.

Wonderful thread 

"Attention is all you need" implementation from scratch in PyTorch. A Twitter thread:
1/

— abhishek (@abhi1thakur) December 13, 2021

Summarizing my lessons

• Lesson #1 - Encoder takes embeddings and source masks
• Lesson #2 - Decoder takes target embedding and target masks
• Lesson #3 - Encoder has sequence of encoders one connected to each other, Encoder1 -> Output -> Encoder2 -> ... Encoder N
• Lesson #4 - Encoder N will be connected to Decoder 1, Decoder has several Decoder Layers
• Lesson #5 - Encoder contains one sequential layer + attention + feedforward layer
• Lesson #6 - In RNN when we read we remember input gate, foreget gate, history gate, output gate. Something here you have the connection to self called self attention, Something like keeping the sequence history
• Lesson #7 - Multihead attention = Multiple self attention layers
• Lesson #8 - Self attention = attention to remember the same sequences, 1-2-3,1-2-3,Again a percentage of sequence might be picked up as historical info which may influnce the next token prediction
• Lesson #9/#10/#11 - Forward function = Softmax + matrix multiplcation
• Lesson #12/#13/#14 - Decoder has similar attention layer, multihead self attention
• Lesson #15/#16 - Padding, Positional encoding with embedding layer
• Lesson #17/#18 -  linear + softmax to decoder output

Keep Exploring!!!

Solopreneur vs Manager vs Entrepreneur

• Solopreneur - I learn everything and do it by myself, I jump quickly than trying to address/fix the gaps
• Manager - I need to get it done from a planning perspective, I may not need to be technically into it
• Entrepreneur - I grow myself plus I grow my team as well, I enable them to do achieve their goals as well as organizational goals

It again depends on personal interests/products/domain. With or without titles do what fills your passion :)

Keep Going!!!

Forecasting usecases in different domains #Forecasting #Features #DataScience

Keep Exploring!!!

Research Paper Reads - Forecasting

Paper #1 - Product age based demand forecast model for fashion retail

Key Notes

• 300 stores, 35k items and around 40 categories.
• Accurate demand forecast 6-12 months in advance
• Age-based prediction model
• Time series models are based on forecasts obtained from previous year’s sales of similar items
• Clustering, classification and prediction
• Determine the appropriate cluster for a new fashion item
• Attributes such as Sleeve length, Color, Pattern, Fastening type and Neck shape
• color and sleeve length are some of the crucial attributes for demand forecasting
• The average selling price at which people buy Dresses is 25$ and is 36$ for Kids wear

Paper #2 - Demand Forecasting in the Presence of Systematic Events: Cases in Capturing Sales Promotions

Key Notes

• Demand uplift by analyzing historical sales data and different combinations of promotions
• Promotion frequency and magnitude of demand uplift
• Promotions, holidays and special events
• contextual information include: changes in promotional plans, competitor activities, market intelligence, sudden climate changes and dynamic influencers
• Type of promotion (e.g., single-buy, buy one get one free, multi-buy)
• Advertisement type (e.g., in-store, online, catalogue)
• Baseline + Uplift and Predicted value
• Major and minor promotions are advertised in retailers’ weekly catalogues and are typically associated with discounts of approximately 50% and 30% off regular price, respectively
• Single buy, Multiple buy transactions per week

Paper #3 - Elasticity Based Demand Forecasting and Price Optimization for Online Retail

Key Notes

• Price elasticity demand value
• Relative change of demand and retail price in percentage, Compute it and add it

• Data Pre-processing module integrates data aggregation, missing data processing, data transformation, data
• normalization and outlier detection, Additional binary features: is_holiday and is_weekend, is_festiveweekednd, is_festiveweekday

• Optimal pricing formulation

Forecasting: theory and practice

Notes from Supply chain section

• Forecasting has always been at the forefront of decision making and planning
• A supply chain is ‘a network of stakeholders (e.g., retailers, manufacturers, suppliers) who collaborate to satisfy customer demand’
• Sales and Operations Planning (S&OP)
• The ‘bullwhip effect’ occurs whenever there is amplification of demand variability through the supply chain (Lee et al., 2004), leading to excess inventories
• Zero sales due to stock-outs or low demand occur very often at the SKU × store level, both at weekly and daily granularity
• Product level (PL) information consists of the time series of sales and returns, alongside
• information on the time each product spends with a customer
• Average custom spend per month
• Average sales per month per brand per category
• Unsold inventory count

More Reads

• An Experimental Review on Deep Learning Architectures for Time Series Forecasting*
• Sub-Seasonal Climate Forecasting via Machine Learning: Challenges, Analysis, and Advances
• Retail forecasting: research and practice
• CHALLENGES AND APPROACHES TO TIME-SERIES FORECASTING IN DATA CENTER TELEMETRY: A SURVEY
• FEW-SHOT LEARNING FOR TIME-SERIES FORECASTING
• Forecasting: theory and practice
• The challenges and realities of retailing in a COVID-19 world: Identifying trending and Vital During Crisis keywords during Covid-19 using Machine Learning
• Learnings from Kaggle’s Forecasting Competitions
• Towards forecast techniques for business analysts of large commercial data sets using matrix factorization methods
• Methods for Intermittent Demand Forecasting

Keep Exploring!!!

Happy New Year 2022

Wish you all a very happy, positive, successful, and prosperous 2022. May all your dreams come true. Keep Learning, Stay Kind, Keep Going!!!

Indian AI Startups #Offerings #AIProducts

 Indian AI Startups #Offerings #AIProducts

Keep Exploring!!!