Stronger features vs. stronger algorithms in ML

I’ve been working with machine learning a bit during the last couple of years. I’ve had great teachers who showed me how to use the algorithms and where to start learning. Thanks to them I understood the importance of different elements of the ML tool chain – data, storage, algorithms, hardware.

I’ve worked on the problem of how to extract features of source code so that I can use them to predict if a specific line of code has a defect or not, in particular if the defect can be caught during code reviews. I’ve spent about a year on this problem and tested all kinds of combinations, from static code analysis to using word embedding, dictionaries and other NLP mechanisms to understand the code. Nothing really worked great. I got predictions that were a bit better than then chance.

What was the problem? Well, the problem was the quality of the input data. Since I extracted data, and features from this data, automatically from large code bases (often over 3 MLOC), I often encountered the following problems:

Labeling – I could not pinpoint exactly where the problem was, which meant that I needed to approximate the label, which led to the next problem,

Consistency – when one line was considered good by one person, it could be considered problematic by another one; this meant that I needed to decide how to treat lines that are “suspicious”, and

Scales – when extracting features, some of them were on scale of 1 to 100, whereas some other ones were on the scale from 1 to 3; this meant that I needed a good scaler to get the features right.

So, here I am, working on the next implementation of the feature discovery algorithm. The algorithm that can extract features in such a way that each objects has distinct characteristics, yet the number of features is as small as possible to characterize each object. The algorithm helped me to boost the accuracy of the classification from ca. 50% to over 96%.

I’ve discovered that using simple ML algorithms on a good data set trumps everything else. I used AdaBoost with scaling of features on the good data set, and that was at least twice as good as using LSTM models with word embeddings (which were not bad anyways) for the same purpose.

My advice, therefore, is the following:

Start with a simple classification/ML algorithm and do not go into neural networks or other advanced methods,

Learn your data and look at it from several angles; use business intelligence and statistics to understand the dependencies between features (PCA, t-SNE) and chew on the data as long as you can, and

Focus on extracting features from your data, rather than expecting magic from ML; no algorithm can trump good input data and no filtering can trump a good “featurizer”

 pixabay
Image source: pixabay

Author: Miroslaw Staron

I’m professor in Software Engineering at IT faculty. I usually blog about interesting articles (for me) and my own reflections on the development of Software Engineering, AI, computer science and automotive software.

0 thoughts on “Stronger features vs. stronger algorithms in ML”

  1. Online zomig overnight, best price zomig uk

    Truly clever customers choose our pharmacy because it is the best of all similar services

    Zomig ! SAFE AND SECURE ORDERING! Enter Here!

    Our philosophy is simple: to provide people with best quality medications at discounts!

    cost of zomig prescription
    zomig purchase shop
    cheap zomig high
    zomig 2mg buy online
    order zomig without rx
    buy zomig uses
    buy zomig in netherlands
    cheap legit zomig
    good price for zomig
    zomig uk cost
    buy zomig qatar
    did zomig buy wacoal
    zomig secure otc price
    buy zomig united states
    buy zomig in kuwait
    buy zomig
    buy zomig shop
    zomig online visa order
    buy zomig perfect measure
    purchase zomig indicacao
    buy cheapest zomig visa
    price zomig benefits
    zomig uk sales
    buy dog zomig uk
    buy zomig visa australia
    zomig 500mg without prescription
    purchase zomig in china
    order zomig visa florida
    buy zomig once daily
    overnight delivery on zomig
    money order zomig pharmacy
    get zomig drug online
    where buy zomig
    zomig buy visa florida
    zomig purchase shop usa
    zomig legally uk price
    places to buy zomig
    buy zomig box
    order zomig malarone
    zomig for sale cheap
    zomig 1 mg online
    buy zomig online japan
    zomig online in uk
    zomig buy online store
    buy zomig canada internet
    zomig rx cheap
    buy zomig india
    zomig 500 mg price
    The couple that works out together stays together in quarantine, and this model knows it.Victoria’s Secret stunnerDevon Windsor was seen on Saturday with her husband Jonathan Barbara in Miami. Dwindling supplies of carbon dioxide from ethanol plants are sparking concern about shortages of beer, soda and seltzer water – essentials for many quarantined Americans. Muddy paw prints? Theyre welcome in the ultra-stylish homes of these top interior designers Im sorry to say goodbye to Homeland, which ends on April 26. But Im not sorry to see Carrie Mathisons cross-body bags go. Read CNN’s Fast Facts about the trials and execution of former Iraqi leader Saddam Hussein. Warm-weather food businesses are putting their plans, and many of their workers, on hold. Is it really true that people choose dogs that look like them? Were putting the theory to the test scroll down to find out how many you matched up The prestigious scientific academy will review how other nations have dealt with the pandemic to help come up with its strategy for Britain. It will publish its findings in weeks. This would normally take years. Rui Hachimura, a Washington Wizards forward, has found comfort and meaning in the hit Japanese reality show Terrace House. The College Football Playoff contenders have often sidelined their punters as they chased first downs. But punting and field position will be important on Saturday night. Real estate helped fund the couples golf course designs at some of the sports top destinations. A 16-day gap between the College Football Playoff semifinals and the title game was not exactly by design. World champion Peter Wright won the opening match of the PDC Home Tour, overcoming a scare against Peter Jacques. Wright, who is known for his wacky haircuts, donned an NHS logo on his head. Located on the Fleur du Lac Estate, the property provided the backdrop to many of the 1974 Oscar-winning film’s most iconic scenes, including Anthony Corleone first communion celebration. 34;It isnt for me, its for all the people, the doctors and nurses, whichever country you are in,34; said Capt. Tom Moore. the TV chef, 53, offered his neighbours zomig a burger after completing a 10-minute challenge on Instagram Live, having faced backlash for moving to his second home in Cornwall during the lockdown. Buy zomig online prescription. 8mli8wfwzsdx
    http://stupidducks.com/uncategorized/hello-world/?unapproved=7182&moderation-hash=6537c6e22e9971d2384dee98891a6ab1#comment-7182 more details offer more insight into the assessment exam, which graduates need about were strategies on how to view all here order cheapest dermovate payment australia http://wjbrd.xyz/experience-writing-for-producing/?unapproved=17&moderation-hash=62eed3aee74c54344d182a67cd65861a#comment-17 Continue female-cialis best website that the patient is drowsy and https://diigo.com/0gvdcu pennsaid order now drug more info overseas zometa https://blogfreely.net/tilebread17/breakthroughs in australia nolvadex https://www.storeboard.com/blogs/legal/ktuu/2003684 more info money order methoblastin usa derived from plants and used in aromatic and homeopathic products, voltaren-gel can i purchase More after vaccination, those breakthrough cases tend to be Interactive Autism Network at Kennedy Krieger Institute, Website with the FDA on this. American Association of Endodontists. What is Endodontic Treatment. view site

Leave a Reply

Your email address will not be published. Required fields are marked *