We’re used to hearing that AI and machine learning is hopelessly complex, impossible to implement quickly, and that if you want to get on board the machine learning bandwagon you’ll need to invest heavily in PhDs, specialists and expensive experts.
This way of thinking is simplistic and behind the times: machine learning is a broad set of technologies, and over the past few months and years there have been huge strides in making machine learning’s benefits much more accessible to startups, scale ups and lone developers alike.
Over the past few months I’ve spent a great deal of time investigating, learning about and iterating on a number of different machine learning technologies to take advantage of the vast quantities of time series data we have about infrastructure performance from my company’s product.
We’re collecting billions of metrics every data from hundreds of thousands of systems, all of which can be used to understand patterns and make future predictions. Read on for some easy, actionable advice on how to get started from scratch with machine learning — it’s easier than you think!
Avoid TensorFlow — for now
Google made headlines in 2015 by open-sourcing TensorFlow, their internal AI and machine learning framework. Released as an open source project, TensorFlow is following the same strategy as Kubernetes — provide such a good product that it becomes the industry standard, and offer a hosted, managed cloud version for those who don’t want to maintain it themselves.
You can run TensorFlow workloads yourself but Google’s Cloud Machine Learning Platform offers a much more optimised version, running on proprietary TensorFlow Processing Unit chipsets. The strategy is all about making Google Cloud the best choice for these jobs.
However, popularity can be deceptive and based on my personal experience TensorFlow is often not the best solution for startups and small companies. TensorFlow is great in that you get a high degree of control over your project but that control comes at a cost. TensorFlow is a framework, and we’ve found it requires significant data science knowledge and a lot of trial and error in building, iterating and improving your models.
It’s not a toolset you should pick up if you’re after easy results or plug-and-play functionality. Unless you’re a big corporation (which we’re not) or have the budgets to hire data scientists to get into model development, it might be tricky to secure enough budget to invest in TensorFlow from the start, so you’d be much better trying more simplistic managed solutions first.
The rise of ‘machine-learning-as-a-service’
For companies just starting out, the best place to begin is looking at the managed service solutions from the likes of Amazon, Microsoft and Google. These solutions are much more accessible to generalist teams, and companies that use them get the benefit of vendors updating them and improving service over time. Indeed, your own datasets help to improve the models!
This is because the larger the training data set, the more accurate the models can be. Anyone can play with theoretical models but the truly interesting work comes out of having real data, and this is an advantage the big players have even before they add your data into the mix.
We’ve found that Amazon Machine Learning is a great place to start. AML differs from TensorFlow in a number of ways: with TensorFlow, you build your own models and can then execute them against your datasets wherever you like whereas AML requires you upload your dataset to Amazon then use their API to execute queries. The downside is you don’t get to control the models and can’t see into the workings of the system – you rely on Amazon to get it right. This “plug and play” type approach but is less customised and flexible, so you may end up needing replacing it with something more specialist in the future.
If you need a very particular type of functionality — detecting items in a video, speech to text or translation, then there are specialist services from all the cloud providers. These services use machine learning behind the scenes, but you don’t need to think about it — send over the item for analysis and get the results through an API. These APIs are quite specific and so if they do a good job, you can just leave them to get on with it. It’s unlikely you’ll want to customise them enough to make it worth starting from scratch.
Outside of the big three cloud providers, there are a host of technology startups including Algorthmia, BigML and MLJar aiming to offer machine learning through an API or SaaS application.
Put your use cases first
I’ve seen many companies make the mistake of rushing into machine learning without having a clear use case in mind, and this is a significant error. There are robust ecosystems around each of the above ‘MLaaS’ platforms, and so you’ll need to have awareness of the APIs available to you. Tools like Amazon Polly (text to speech) or the Google Cloud Video Intelligence API deliver specialist functionality without requiring a high degree of knowledge as a prerequisite.
Since they are offered as an API, you can mix and match across providers and even test which does a better job where the service is the same. Most people will probably stick with the cloud platform the rest of their infrastructure is hosted on, but that’s not always necessary (data transfer cost and latency may become an issue once you hit scale though).
At my company, we’ve been migrating from IBM Softlayer to Google Cloud and the data transfer fees of (encrypted) traffic across the internet is part of the total cost consideration, and an incentive to complete the move quickly! Once it’s all within Google’s network then the lower (or zero) data fees apply when using their services, and Google is widely considered to have well designed machine learning capabilities.
I’ve found the advantage of using machine learning as a service APIs is that any developer can pick them up and start playing. Serious machine learning with TensorFlow requires a lot of time and real data science knowledge, which may be worth investing in over the long term. However, to get something up and running quickly and test the value proposition to your users, there are a variety of options.
I’ve had a lot of fun testing out the different machine learning APIs and solutions out there, and this element of fun and discovery makes it much easier to lead a team on a small exploratory project. I’ve also found that implementing something like Google’s 20 percent time, or even an internal hackathon could also be a good opportunity to get everyone focused on building an initial prototype.
Machine learning is a very over-hyped set of technologies — it’s currently ranked by Gartner as a buzzword, at the very top of their peak of inflated expectations. However there’s a vibrant set of technologies under this umbrella term, and you don’t necessarily need to have a highly-specialised workforce to take advantage of them. Start small, use the managed services provided by the big tech firms, and you’ll be surprised by how far you can go.
Shaneeshwaruni Divya Charitra Episode 112
Unable to get the powerful weapon for himself, Raahu plans to make sure the Devas don’t get it either. He instigates Indradev to get the powerful weapon before it is handed over to Suryadev. He informs Shani about his plot to make the Devas fight among themselves. Will Shani reach in time to stop this from happening?
Watch Shaneeshwaruni Divya Charitra Daily Serial online
Mon – Fri @ 8:30PM on Gemini TV
Can Shani put a stop to Raahu’s vicious plan?
Shaneeshwaruni Divya Charithra Telugu Dubbing Version of Shani Dev
Featured Image by viacom18
Don’t Eat These Foods on Diwali
నరకాసురుడు అనే రాక్షసుణ్ణి చంపినా సందర్భంగా జరుపుకుంటున్న పండుగ దీపావళి భారతదేశం లోనే హిందువులు చేసుకుంటున్న పండగలలో ఇది ఒక్కటిగా చెప్పుకోవొచ్చు.దీపావళి నాడు లక్ష్మి దేవి ని పూజిస్తే ఐశ్వర్యాలు సిద్ధిస్తాయని,విజయం సాధిస్తామని చెడుని ఎదిరించి దైర్యం వస్తుందని అంటారు పెద్దలు.దీపావళి మహా లక్ష్మి కి చాలా ప్రీతిపాత్రమైనది గ చెప్పుకుంటారు.ఆ రోజు భక్తి శ్రద్దలతో తల్లిని పూజిస్తే ఎలాంటి కోరికలు అయిన తీరుతాయని నమ్మకం.ఇక దీపావళికి భక్తిగా నోముకుంటే ఆ సంవత్సరం అంతా శుభం కలుగుతుంది.అయితే ఇప్పుడు దీపావళికి అర్ధం మార్చేశారు దేఎపావలి అంటే పటాసులు కాల్చడమే పండగ అనుకుంటున్నారు
ఇక మరికొందరు దీపావళి పండగ అనగానే పొద్దున్న లేవటం మటన్ లేదా చికెన్ వంటి మాంసాహారం తెచ్చి వండుకొని తింటుంటారు.దీపావళి మహా లక్ష్మి కి ప్రీతిపాత్రమైన రోజు అందుకే ఆ రోజు మాంసాహారం తీసుకోవడం మంచిది కాదు.దీపావళి నాడు పూర్తిగా ఉపవాసం చేసి కేవలం శాఖాహారం అది కూడ ఆకు కూరలను తీసుకున్నట్లైతే మంచి జరుగుతుందని చెబుతున్నారు పండితులు.ఎంత పూజ చేసి ఎన్ని కొబ్బరి కాయలు కొట్టిన కూడ ఆ రోజు మాంసాహారం తింటే ఫలితం ఉండదు.లక్ష్మి దేవి పూజ చేసిన రోజు మాసాహారం తీసుకోవడం మంచిది కాదని అలా చేయడం వల్ల అష్ట దరిద్రాలు చుట్టుకుంటాయని అంటారు.అలా చేస్తే ముఖ్యంగా ఆర్ధిక ఇబ్బందులు ఎదురు అవుతాయి.డబ్బులు వచ్చినట్లు వచ్చి ఖర్చు అవుతుంటాయి అందుకే ఆ రోజు ఎత్తి పరిస్థితుల్లో మాంసాహారం తీసుకోవోద్దని చెబుతున్నారు
7 Haunted Hotels That’ll Scare
The idea of ‘ghosts’ has always been interesting but experiencing paranormal activity can lead to some indelible scars on one’s mind. People all around the world have felt spooky feeling in some hotels. Some fled while some stayed back to narrate stories. From icy sensation to weird sounds to haunting apparitions, some hotels of the world have really seen “too much”. Apart from celebrities, royals and common people, some hotels have also been visited by ghosts and spirits.
Here’s a list of spooky hotels from around the world:
#1 Stanley Hotel, Colorado
A hotel of paranormal activity has witnessed the ghosts of F.O. Stanley, the founder, and his wife passing through the lobby quite frequently. Also, the sound of children running and giggling down the halls is a common phenomenon in this spooky hotel. Opened in 1907, this hotel lost one of his housemaids, Elizabeth Wilson, in 1911. It is said her ghost once returned to assist the guest with their belongings in room 217. Chilling, isn’t?
#2 Hotel Roosevelt, California
This hotel is world famous for the celebrities it houses and also the lavish parties. Sadly, this lovely hotel has also witnessed some spooky happenings like the ghost of Marilyn Monroe viewing herself in the mirror placed in her former suite and also the ghost of the late actor Montgomery Cliff playing his trombone from room 928. How spooky is this!
#3 Langham Hotel, London
A very notable historical hotel in London, Langham hotel was established in 1865 and since then it has seen the coming and going of famous celebrities and some lavish parties. Other than celebrities and royals, ghosts also find respite in this hotel which has often heard guests complaining about seeing the apparition of Napoleon the third, a German soldier and a doctor who murdered his wife. If willing you can book room 333, the most haunted of all rooms! Willing, anyone?
#4 Hotel Raj Kiran, Lonavala
The authorities got so disturbed from constant complaints from residents of the room at the corner of reception area about bed sheets being pulled off forcefully, seeing a blue ray of light in midnight, that they have now stopped renting that room. It’s been vacant since many years. So scary!
#5 Ballygally Castle, North Ireland
The castle was built by Lord Shaw and his wife Lady Isabella in 1625. Sadly, Lady Isabella fell off from the top of the tower. Some say she jumped while some say she was pushed. Her ghost is seen wandering in the hotels, knocking at the doors and then disappearing. In her honor, the hotel has reserved a room as the “The Ghost Room”.
#6 The Savoy, Mussoorie
One of the most famous haunted hotels in India, The Savoy was built in 1902 but after the murder of Lady Garnet Orme in 1910, began witnessing weird paranormal activities in the hotel. Famous for its haunted corridors and walls, the hotel was visited by the Indian Paranormal Society who recorded a lady’s whispers. Wonder what she was trying to say?
#7 Hotel Burchianti, Italy
Built in Florence, the hotel is famous for its impeccable styled rooms and the opulent décor. But apart from these what makes this hotel a bit more special is a list of paranormal activities going on at a regular basis. Guests have often seen the ghosts of children skipping down the halls, a lady sitting in a chair and knitting and a maid doing her cleaning jobs early morning. Visitors of Fresco Room where once Mussolini stayed, complain of being watched continuously and feeling an icy breath on their faces. Haunting!
Featured Image Source: http://www.rebelsmarket.com
Subscribe to our Newsletter
Serials3 days ago
Shaneeshwaruni Divya Charitra Episode 118
Serials1 day ago
Shaneeshwaruni Divya Charitra Episode 119
Entertainment2 days ago
Watch: Guru Randhawa’s song ‘Lahore’ is all about being smitten in love
Business2 days ago
Andrew Ng’s Landing.ai wants to bring artificial intelligence to the manufacturing industry, starting with Foxconn
Entertainment3 days ago
Bigg Boss 11, Episode 72, 12 December 2017: Arshi, Akash want to use Shilpa’s friendship to become finalists
Entertainment2 days ago
Woofer: Snoop Dogg, Dr Zeus, Nargis Fakhri unite their creative energies for this stylised Punjabi number
Entertainment2 days ago
Soha Ali Khan on the perils of being moderately famous, turning author, and lessons her family taught her
Entertainment3 days ago
Kolkata’s Jazz Fest 2017, featuring Erik Truffaz, Amyt Dutta, was a showcase of the genre’s experimental potential