Confusion Matrix…

I completed a Machine Learning course by Andrew NG on Coursera nearly 2 months ago, a course with an enrollment of 1.7mn people apparently. Not sure what the ‘course completion %’ is, though. Thanks to the course, I was able to revise some basics, primarily the math and intuition behind Machine Learning models. After you build a model, one of the important metrics by which people judge your model is the model accuracy. The problem is that there are so many of these evaluation metrics that quite often someone asks me to give a feedback on my model using a different accuracy metric.

In a typical classification problem solved using Machine learning techniques, the outputs can be represented in a 2x2 matrix form which is called the ‘Confusion matrix’, shown below:-

1) Sensitivity = TP/(TP+FN) = TP/P
A useful metric when you are interested in predicting the positives right. Tests in Medical diagnosis are a good example. We wish to have as less false negatives as possible, i.e., no room for non-identification of a disease when the test is conducted.

2) Specificity = TN/(TN+FP) = TN/N
A useful metric when you wish to have the negatives right. Follow-up tests in Medical diagnosis are the best example. We don’t want to start treatment on people who don’t have the disease in the first place.

3) Precision = TP/(TP+FP)
A metric that tells you how precise is your prediction on the positives.

4) Recall = Sensitivity
Another fancy name for Sensitivity.

5) True Negative% = Specificity*100
Just a different name for Specificity.

6) True Positive% = Recall*100
Again, just a different name for sensitivity but more intuitive.

7) Accuracy = (TP+TN)/(TP+FN+FP+TN)
Helps to know what percentage of the predictions are correct. The limitation of this metric is in the cases of rare events. Say for example, 1% of the population has heart related diseases and the model predicts that none of them have heart diseases, the accuracy of the model will still be 99%.

8) F1-score = (2*Precision*Sensitivity)/(Precision+Sensitivity)
A good metric because it penalizes if you have got either high False positives or False Negatives.

Of course, we would like to have all these metrics as high as possible (close to 1 / near 100%) and that simply means our model is very good. The thing to wonder about is that if this 2x2 matrix can create so many evaluation metrics (many not shared above), how sophisticated/complicated is this evaluation business!

Saturday Translations…

We accidently come across a nice Sher and then we go, reading many more. This post is an attempt to translate one of those Ghazals.

अब के हम बिछड़े तो शायद कभी ख़्वाबों में मिलें
जिस तरह सूखे हुए फूल किताबों में मिलें।

Ab ke bichhde to shaayad kabhi khwaabon mein milen
Jis tarah sookhe huye phool kitaabon mein milen.

If we part this time maybe we'll meet in the dreams
The way dried old flowers are found in the books.

ढूँढ उजड़े हुए लोगों में वफ़ा के मोती
ये ख़ज़ाने तुझे मुमकिन है ख़राबों में मिलें।

Dhoond ujde huye logon mein vafa ke moti
Ye khazaane tujhe mumkin hai kharaabon mein milen.

Search for pearls of loyalty among the devastated
This treasure is most likely found in the wrecks.

ग़म-ए-दुनिया भी ग़म-ए-यार में शामिल कर लो
नशा बढ़ता है शराबें जो शराबों में मिलें।

Gam-e-duniya bhi gam-e-yaar mein shaamil kar lo
Nasha badhta hai sharaaben jo sharaabon mein milen.

Include the sorrow of my beloved in the sorrow of this world
The intoxication rises, after all, when the wines mix in wines.

तू ख़ुदा है न मिरा इश्क़ फ़रिश्तों जैसा
दोनों इंसाँ हैं तो क्यूँ इतने हिजाबों में मिलें।

Tu khuda hai na mira ishq farishton jaisa
Donon insaan hain to kyun itne hizaabon mein milen.

You are neither a God nor is my love angel-like
We are both humans so why do we meet in veils.

आज हम दार पे खींचे गए जिन बातों पर
क्या अजब कल वो ज़माने को निसाबों में मिलें।

Aaj hum daar pe kheenche gaye jin baaton par
Kya ajab kal vo zamaane ko nisaabon mein milen.

Those ideas for which I was tortured in the gallows today
How strange, tomorrow the world found those in the texts.

अब न वो मैं न वो तू है न वो माज़ी है 'फ़राज़'
जैसे दो साए तमन्ना के सराबों में मिलें।

Ab na vo main na vo tu hai na vo maazi hai ‘Faraz’
Jaise do saaye tamanna ke saraabon mein milen.

Neither the old me or you nor are those past times here 'Faraz'
It is just like two shadows cast across the mirage of desires.

– अहमद फ़राज़ (Ahmad Faraz)

To take our spirits to yet another level, here comes the nice Sher that sparked the idea for this post.

हम जिसके हो गये वो हमारा ना हो सका
यूँ भी हुआ हिसाब बराबर कभी कभी।

Ham jiske ho gaye vo humara na ho saka
Yun bhi hua hisaab baraabar kabhi kabhi.

I became theirs who could never become mine
Even like this the account balances sometimes.

– आल-ए-अहमद सुरूर (Aal-e-Ahmad Suroor)

That’s all for now. Happy Sunday! 😀

Forgiveness…

Today is an important day in the Jain calendar. People start their conversations on this day by saying ‘Khamat Khaamna’ to each other, which means ‘Forgive me if I have said or done anything that hurt you (un)intentionally’. Yesterday was ‘Samvatsari’ – the holiest day and end of an eight day festival [also called as Paryushan] celebrated by Shwetamber Jains. The word Samvatsari has its origin in Sanskrit, the original word being ‘Samvatsara’ which means ‘a year’ and in this context, means ‘occuring once a year’. Unlike other festivals which are about gatherings, food, decorations, gifts, etc. this one is about contemplation, prayer and repentance. Many Jains fast for around 36 hours on the 8th day and at the end of the fast, they seek forgiveness for their words and deeds from all the living beings of the world.

On this important day, ‘Khamat Khaamna’ to our readers.🙏

Park Visits…

We regularly visit a nearby park and many fun things happen while we are there. Most of the times, I’m still thinking about the fun stuff when we walk back home while Sakura tries to get hold of my hair but her hands already have a few strands. There is this 4 year old boy who always enters the park after we do. It can’t be a coincidence every time, it should be intentional that he enters the park a few minutes later. He speaks to me everyday and gives me some information about the park and nothing else. Yesterday he said ‘there is rain water clogged under that swing’. The day before he said ‘the grass in this park has grown so much’. The day before that he said ‘the grass in this park has snakes. I saw one’. So everyday I wait for him to come, give me some information after which he goes away. There are days when I have tried to strike a conversation with him – ‘so are you done with school today?’ and he just ‘nods’.

Sakura and I occupy the same swing. It’s so kind of the caretakers they never ask me to not use a swing designed for children. I guess they know that Sakura can’t ride on her own as yet, so they must be fine with me giving her company. The first time we were on the swing Sakura sat all cuddled in my lap. She was looking at the soil underneath and I observed that her hands moved slowly onto the sides as if trying to locate something. A few minutes later, I saw her holding on to the swing. A few days passed and I knew she had become very comfortable with the swing as she knew instinctively how to hold on to the swing. This is the difference I feel between training a neural network and teaching a child. The former only does what you have trained it for but a child can do things with no training or coaching, it in fact can apply its own principles. She moves her legs just the way I do just to add to the momentum of the swing. Last month, I was teaching her how to hit the mattress using her palms and observe the kind of sound produced. A few days later I saw her hitting some of the toys the same way and trying to see what kind of sounds they produced. To my astonishment, she is hitting my arms and face the same way as she hits those toys and is stunned that no interesting sound is being produced. What a child may be thinking is pretty interesting to imagine and people say there is so much not known in the area of Child psychology.

For now, I’m taking Sakura to the verandah where she notices all kinds of vehicles passing by. She is either smiling at people who are on their bicycles or is calling those who are speeding up in their cars.

Have a nice weekend.

Sketchbook Volume 13

After a hiatus of almost two years, a new volume of Sketchbook is out. And guess what? I have got it as ‘usual’. But unusually, this time from Amazon.co.jp!

It got delivered by DHL within 4 days. Though the packaging was less sturdier than packages of yester-years.

Anyway, opening the ‘envelope’ gave way directly to the long-awaited volume:

Finally, Ooba Tsukiyo on the cover! I have been waiting for this to happen since volume 10. And she gets the whole page to herself too. That is awesome!

This volume has a lot of commentary about bugs, insects, birds, etc. from Nagisa but there are a few enjoyable ‘silent’ panels too nonetheless. So will share such pages from this volume some other time. Till then

Enjoy this!

First Two Months…

This post is just going to be a list of updates. Sorry but if I elaborate, these will be a few posts at least.

1) I’m learning how to take care of Sakura Jain on my own. Sometimes it is overwhelming, sometimes tiring but mostly just fine and heart warming.

2) We (me and my Superpartner) are watching ‘Black Mirror’ on Netflix. Sharing my favorite episode, feels like we have this system with Ola & Uber drivers already.

3) I recently completed a few specializations on Coursera – a good brush-up after University. These were specializations on hypothesis driven testing, confidence intervals, sample constructions, and the like.

4) I now speak in the voice of Shin-chan [the Hindi dubbed one that we see on Indian channels] 50% of the times with Sakura-chan.

5) Lastly, I’m hearing a lot of sad news of people falling ill or falling through life itself.

A point outside the points above – Sakura has gone from saying ‘uh-goo’ to ‘dee-mu’ [I thought it would be ‘muh-mma’ or something similar but thanks to my luck, Greek letters come first].

Many Things…

Many things have happened since the last time We/I posted on this blog.

Like the end of scanlation of Detective Academy Q.

Like the appearance of a paper of mine on arXiv after quite a delay.

Like the publication of another paper of mine in NPB after quite a lot of delay.

Like the Birth of Our Daughter Sakura.

On that note, let us read some great Shayari…

क़हक़हों का तिलिस्म टूट गया
दर्द ने जब हंसी उड़ाई है।

Kahakahon ka tilism toot gaya
Dard ne jab hansi udaayi hai.

The charm of cackles faded away
When the pain laughed out loud.

अनसुना करने से पहले सोच लो तुम एक बार
खामोशी का शोर है ये, बेजुबानी का नहीं।

Ansuna karne se pahle soch lo tum ek baar
Khaamoshi ka shor hai ye, bezubaani ka nahin.

You should consider it once, before ignoring
It is the scream of silence, not of muteness.

– मदन मोहन दानिश (Madan Mohan Daanish)

मंज़िल समझके बैठ गए जिनको चंद लोग
मैं ऐसे रास्तों से गुजरता चला गया।

Manzil samajhke baith gaye jinko chand log
Main aise raaston se gujarta chala gaya.

Some people stopped, thinking it’s the destination
While I, on such roads, kept walking continuously.

दुनिया समझ में आई, मगर आई देर से
कच्चा बहुत था रंग उतरता चला गया।

Duniya samajh mein aayi, magar aayi der se
Kachcha bahut tha rang utarta chala gaya.

I understood the world, albeit a little late
The fake dye kept fading continuously.

– वसीम बरेलवी (Waseem Barelvi)

More Shayari from Dr. Waseem Barelvi

New Manga

I have started reading Dr. Stone… which can be labelled as being a “Science-based” manga. Chemistry for most of the time, in my ‘opinion’. One ‘Arc’ worth 42 chapters is over and another is to start next week with more action I guess! Have a look at it if you can find it somewhere.

This blog is going to get infrequent updates from now on for reasons known and unknown.

See you some other time and place in some future post.

End-of-the-Post-Link

Final Year of the Poll…

If any of you ‘readers’ has noticed the poll on this blog, good for you and those who have actually voted on it: Thanks & Happy New Year.

There has been some glitch on Blogger’s server and 7 voters’ data disappeared sometime this year; only 2 remained as seen below:

The poll is a few years (3?) old and won’t be renewed for 2018 so here I publish the results as they should have been:

As can be clearly inferred, “Readers” want us to write more about Sher-o-Shayari! Followed by Travelogues, Photography and Humour. My personal favourites Manga/Anime and Programming are the least-favoured it seems. Though, they are not going anywhere, let me tell you… going nowhere.

Anyway, let me end this post by giving a detailed breakdown of the votes:

	Humour	Manga/Anime	Photography	Programming	Sher-o-Shayari	Travelogues
Us	√	√	√	√	√	√
3(N)	√	√	√	√	√	√
4(S)	×	×	×	×	√	×
5(R)	×	×	√	×	√	√
6	×	×	×	×	√	×
7(P)	×	×	×	×	×	√
8	√	×	√	×	√	√
9	×	×	×	×	√	×

That's all for this year.

See you next year!