Keras: Embrace Randomness - Revision history

Onnowpurbo: /* Taktik untuk Melaporkan Ketidakpastian */

2019-09-12T06:46:13Z

Taktik untuk Melaporkan Ketidakpastian

← Older revision		Revision as of 06:46, 12 September 2019
Line 126:		Line 126:
	Anda dapat mengumpulkan populasi ukuran kinerja. Gunakan statistik pada populasi ini.		Anda dapat mengumpulkan populasi ukuran kinerja. Gunakan statistik pada populasi ini.

	~~This tactic is called report summary statistics~~.		Taktik ini dikenal sebagai Report Summary Statistics.

	The distribution of results is most likely a Gaussian, so a great start would be to report the mean and standard deviation of performance. Include the highest and lowest performance observed.		The distribution of results is most likely a Gaussian, so a great start would be to report the mean and standard deviation of performance. Include the highest and lowest performance observed.
Line 141:		Line 141:
	Lean on statistical significance tests. Statistical tests can determine if the difference between one population of result measures is significantly different from a second population of results.		Lean on statistical significance tests. Statistical tests can determine if the difference between one population of result measures is significantly different from a second population of results.

	~~Report the~~ significance ~~as well~~.		Laporkan significance-nya juga.

	~~This too is a~~ best practice, ~~that sadly does not have enough adoption~~.		Ini juga merupakan best practice, yang sayangnya tidak banyak yang mengadopsi.
	Wait, What About Final Model Selection
			==Wait, What About Final Model Selection==

	The final model is the one prepared on the entire training dataset, once we have chosen an algorithm and configuration.		The final model is the one prepared on the entire training dataset, once we have chosen an algorithm and configuration.

Onnowpurbo: /* Tactics to Report the Uncertainty */

2019-09-12T06:37:53Z

Tactics to Report the Uncertainty

← Older revision		Revision as of 06:37, 12 September 2019
Line 118:		Line 118:
	Ini lebih lazim dengan stochastic optimization dan neural networks, tetapi sama relevan secara umum. Cobalah.		Ini lebih lazim dengan stochastic optimization dan neural networks, tetapi sama relevan secara umum. Cobalah.

	===~~Tactics to Report the Uncertainty~~===		===Taktik untuk Melaporkan Ketidakpastian===

	~~Never report the performance of your machine learning algorithm with a single number~~.		Jangan pernah melaporkan kinerja algoritma pembelajaran mesin anda dengan sebuah angka.

	~~If you do~~, ~~you’ve most likely made an error~~.		Jika ya, kemungkinan besar anda membuat kesalahan.

	~~You have gathered a population of performance measures~~. ~~Use statistics on this population~~.		Anda dapat mengumpulkan populasi ukuran kinerja. Gunakan statistik pada populasi ini.

	This tactic is called report summary statistics.		This tactic is called report summary statistics.

Onnowpurbo: /* Taktik untuk Mengurani Ketidakpastian */

2019-09-12T06:35:51Z

Taktik untuk Mengurani Ketidakpastian

← Older revision		Revision as of 06:35, 12 September 2019
Line 102:		Line 102:
	* Laporkan Ketidakpastian (Report the Uncertainty).		* Laporkan Ketidakpastian (Report the Uncertainty).

	===Taktik untuk ~~Mengurani~~ Ketidakpastian===		===Taktik untuk Mengurangi Ketidakpastian===

	Jika kita mendapatkan model yang berbeda pada setiap kali kita menjalankan algoritma, apa yang bisa kita lakukan?		Jika kita mendapatkan model yang berbeda pada setiap kali kita menjalankan algoritma, apa yang bisa kita lakukan?

Onnowpurbo: /* Taktik Untuk Mengatasi Ketidakpastian Algoritma Stochastic */

2019-09-12T06:34:49Z

Taktik Untuk Mengatasi Ketidakpastian Algoritma Stochastic

← Older revision		Revision as of 06:34, 12 September 2019
Line 99:		Line 99:
	Ada 2 strategi sederhana yang dapat kita gunakan:		Ada 2 strategi sederhana yang dapat kita gunakan:

	* Reduce the Uncertainty.		* Kurangi Ketidakpastian (Reduce the Uncertainty).
	* Report the Uncertainty.		* Laporkan Ketidakpastian (Report the Uncertainty).

	===~~Tactics to Reduce the Uncertainty~~===		===Taktik untuk Mengurani Ketidakpastian===

	~~If we get different models essentially every time we run an algorithm~~, ~~what can we do~~?		Jika kita mendapatkan model yang berbeda pada setiap kali kita menjalankan algoritma, apa yang bisa kita lakukan?

	~~How about we try running the algorithm many times and gather a population of performance measures~~.		Bagaimana kalau kita mencoba menjalankan algoritma berkali-kali dan mengumpulkan populasi ukuran kinerja.

	~~We already do this if we use~~ k-fold cross validation. ~~We build~~ k ~~different models~~.		Kita sudah melakukan ini jika kita menggunakan k-fold cross validation. Kita membuat k model yang berbeda.

	~~We can increase~~ k ~~and build even more models~~, ~~as long as the~~ data ~~within each fold remains representative of the problem~~.		Kita dapat meningkatkan k dan membangun lebih banyak model, selama data dalam setiap lipatan tetap mewakili masalah.

	~~We can also repeat our evaluation process n times to get even more numbers in our population of performance measures~~.		Kita juga dapat mengulangi proses evaluasi kita berkali-kali untuk mendapatkan jumlah pengukuran kinerja yang lebih banyak lagi.

	~~This tactic is called~~ random ~~repeats or~~ random ~~restarts~~.		Taktik ini di kenal sebagai random repeat atau random restart.

	~~It is more prevalent with~~ stochastic optimization ~~and~~ neural networks, ~~but is just as relevant generally~~. ~~Try it~~.		Ini lebih lazim dengan stochastic optimization dan neural networks, tetapi sama relevan secara umum. Cobalah.

	===Tactics to Report the Uncertainty===		===Tactics to Report the Uncertainty===

Onnowpurbo: /* Tactics To Address The Uncertainty of Stochastic Algorithms */

2019-09-12T06:26:28Z

Tactics To Address The Uncertainty of Stochastic Algorithms

← Older revision		Revision as of 06:26, 12 September 2019
Line 93:		Line 93:
	Ini adalah expektasi yang sangat nyata yang harus kita tangani dalam praktik.		Ini adalah expektasi yang sangat nyata yang harus kita tangani dalam praktik.

	==~~Tactics To Address The Uncertainty of~~ Stochastic ~~Algorithms~~==		==Taktik Untuk Mengatasi Ketidakpastian Algoritma Stochastic==

	~~Thankfully~~, ~~academics have been struggling with this challenge for a long time~~.		Untungnya, para akademisi telah berjuang dengan tantangan ini sejak lama.

	~~There are~~ 2 ~~simple strategies that you can use~~:		Ada 2 strategi sederhana yang dapat kita gunakan:

	Reduce the Uncertainty.		* Reduce the Uncertainty.
	Report the Uncertainty.		* Report the Uncertainty.

	===Tactics to Reduce the Uncertainty===		===Tactics to Reduce the Uncertainty===

Onnowpurbo: /* Machine Learning Algorithms are Stochastic */

2019-09-12T06:24:44Z

Machine Learning Algorithms are Stochastic

← Older revision		Revision as of 06:24, 12 September 2019
Line 68:		Line 68:
	Itu harus menjadi bagian default dari setiap percobaan yang kita jalankan.		Itu harus menjadi bagian default dari setiap percobaan yang kita jalankan.

	==Machine Learning ~~Algorithms are~~ Stochastic==		==Algoritma Machine Learning adalah Stochastic==

	~~If a machine learning algorithm gives a different~~ model ~~with a different sequence of~~ random ~~numbers~~, ~~then which~~ model ~~do we pick~~?		Jika algoritma pembelajaran mesin memberikan model yang berbeda dengan urutan random number yang berbeda, maka model mana yang kita pilih?

	~~Ouch~~. ~~There’s the rub~~.		Ups ...

	~~I get asked this question from time to time and I love it~~.		Ini pertanyaan dasar machine learning, seperti ..

	~~It’s a sign that someone really gets to the meat~~ of ~~all this applied machine learning stuff – or is about to.~~		* Different runs of an algorithm with…
			* Different random numbers give…
			* Different models with…
			* Different performance characteristics…

	~~Different runs of an algorithm with…~~		Akan tetapi perbedaan yang ada masih dalam sebuah cakupan / range.
	~~Different random numbers give…~~
	~~Different models with…~~
	~~Different performance characteristics…~~

	~~But the differences are within a range~~.		Nama fancy untuk perbedaan ini atau perilaku random dalam suatu rentang adalah stochastic.

	~~A fancy name for this difference or random behavior within a range is~~ stochastic.		Algoritma Machine learning dalam prakteknya adalah stochastic.

	~~Machine learning algorithms are~~ stochastic ~~in practice~~.		* Expect them to be stochastic.
			* Expect there to be a range of models to choose from and not a single model.
			* Expect the performance to be a range and not a single value.

	~~Expect them to be stochastic.~~		Ini adalah expektasi yang sangat nyata yang harus kita tangani dalam praktik.
	~~Expect there to be a range of models to choose from and not a single model.~~
	~~Expect the performance to be a range and not a single value.~~

	~~These are very real expectations that you MUST address in practice~~.

	==Tactics To Address The Uncertainty of Stochastic Algorithms==		==Tactics To Address The Uncertainty of Stochastic Algorithms==

Onnowpurbo: /* Tactics to Report the Uncertainty */

2019-09-12T06:16:33Z

Tactics to Report the Uncertainty

← Older revision		Revision as of 06:16, 12 September 2019
Line 120:		Line 120:
	It is more prevalent with stochastic optimization and neural networks, but is just as relevant generally. Try it.		It is more prevalent with stochastic optimization and neural networks, but is just as relevant generally. Try it.

	==Tactics to Report the Uncertainty==		===Tactics to Report the Uncertainty===

	Never report the performance of your machine learning algorithm with a single number.		Never report the performance of your machine learning algorithm with a single number.
Line 190:		Line 190:

	Ask your question in the comments and I will do my best to answer.		Ask your question in the comments and I will do my best to answer.


	==Referensi==		==Referensi==

Onnowpurbo: /* Tactics to Reduce the Uncertainty */

2019-09-12T06:16:23Z

Tactics to Reduce the Uncertainty

← Older revision		Revision as of 06:16, 12 September 2019
Line 104:		Line 104:
	Report the Uncertainty.		Report the Uncertainty.

	==Tactics to Reduce the Uncertainty==		===Tactics to Reduce the Uncertainty===

	If we get different models essentially every time we run an algorithm, what can we do?		If we get different models essentially every time we run an algorithm, what can we do?

Onnowpurbo: /* Random Seed dan Hasil yang Reproducible */

2019-09-12T06:15:36Z

Random Seed dan Hasil yang Reproducible

← Older revision		Revision as of 06:15, 12 September 2019
Line 60:		Line 60:
	Kita bisa mendapatkan hasil yang dapat direproduksi dengan mem-fix-kan seed random number generator sebelum setiap model yang kita buat.		Kita bisa mendapatkan hasil yang dapat direproduksi dengan mem-fix-kan seed random number generator sebelum setiap model yang kita buat.

	~~In fact~~, ~~this is a~~ best practice.		Sebenarnya, ini adalah best practice.

	~~We should be doing this if not already~~.		Kita harus melakukan ini jika belum.

	~~In fact~~, ~~we should be giving the same sequence of~~ random ~~numbers to each algorithm we compare and each technique we try~~.		Faktanya, kita harus memberikan urutan random number yang sama untuk setiap algoritma yang kita bandingkan dan setiap teknik yang kita coba.

	~~It should be a~~ default ~~part of each experiment we run~~.		Itu harus menjadi bagian default dari setiap percobaan yang kita jalankan.

	==Machine Learning Algorithms are Stochastic==		==Machine Learning Algorithms are Stochastic==

Onnowpurbo: /* Random Seeds and Reproducible Results */

2019-09-12T06:13:26Z

Random Seeds and Reproducible Results

← Older revision		Revision as of 06:13, 12 September 2019
Line 42:		Line 42:
	Tanpa keraguan. Tidak ada keraguan, keacakan (randomness) memainkan peran besar dalam penerapan machine learning.		Tanpa keraguan. Tidak ada keraguan, keacakan (randomness) memainkan peran besar dalam penerapan machine learning.

	==Random ~~Seeds and~~ Reproducible ~~Results~~==		==Random Seed dan Hasil yang Reproducible==

	~~Run an algorithm on a~~ dataset ~~and get a~~ model.		Jalankan algoritma pada dataset dan dapatkan model.

	~~Can you get the same~~ model ~~again given the same~~ data?		Bisakah kita mendapatkan model yang sama dengan data yang sama?

	~~You should be able to~~. ~~It should be a requirement that is high on the list for your modeling project~~.		Kita harusnya bisa. Ini harus menjadi persyaratan yang di lekatakn cukup tinggi dalam daftar di proyek pemodelan kita.

	~~We achieve reproducibility in applied~~ machine learning ~~by using the exact same code~~, data ~~and sequence of~~ random ~~numbers~~.		Kita bisa mencapai reproduktifitas dalam penerapan machine learning dengan menggunakan kode, data, dan urutan angka random yang sama persis.

	~~Random numbers are generated in software using a~~ pretend random number generator. ~~It’s a simple math function that generates a sequence of numbers that are random enough for most applications~~.		Nomor random dihasilkan dalam perangkat lunak menggunakan pretend random (pseudorandom) number generator. Ini adalah fungsi matematika sederhana yang menghasilkan urutan angka yang cukup acak untuk sebagian besar aplikasi.

	~~This math function is deterministic~~. ~~If it uses the same starting point called a~~ seed number, ~~it will give the same sequence of~~ random ~~numbers~~.		Fungsi matematika ini bersifat deterministik. Jika menggunakan titik awal yang sama yang disebut seed number, maka fungsi ini akan memberikan urutan angka random yang sama.

	Problem solved.		Problem solved.
	~~Mostly.~~

	~~We can get reproducible results by fixing the~~ random number ~~generator’s seed before each~~ model ~~we construct~~.		Kita bisa mendapatkan hasil yang dapat direproduksi dengan mem-fix-kan seed random number generator sebelum setiap model yang kita buat.

	In fact, this is a best practice.		In fact, this is a best practice.