What is Pattern Recognition? How is it useful?

Pattern Recognition is a technique of sensing a repetitive occurrence in any situation. It can be very useful in problem solving.
Say we want to count the number of x in the following diagram
x x x x x x x x x x
x x x x x x x x x x
x x x x x x x x x x
x x x x x x x x x x
x x x x x x x x x x
We see there is a pattern in the way the ‘x’ is placed – uniformly across all columns in every row. What would an intelligent person do? Would he count all the singleton ‘x’ beginning from the first row and ending with the last? The answer, as you have rightly guessed, is NO. The person would count the number of ‘x’ in any row and multiply it with the number of rows. In our above figure, it should be 10 (number of x in each row) multiplied by 5 (number of rows) which gives the number of x to be equal to 50.
For problems of much larger sizes and complexity, detecting patterns can help in finding solution faster making it an efficient one. That is why Pattern Recognition is so useful in Computational problem solving.

Advertisements

Leave a comment

Filed under artificial intelligence, Data Science and Analytics

Tech and Society 3: Fake news

Tech and Society 3: Fake news
Recently a BBC commissioned Research published a report that alleged that the largest contributor to fake news emanating from India was churned out from the IT cell that owed allegiance to the party running the Govt. Quite predictably, seemingly credible news report surfaced that countered the BBC Research claim highlighting the shortcomings of the published report including technical points such as sample size and sampling bias. We have seen a similar situation a month or two back with the outbreak of disclosure about the Rafeal deal.

With the advent of Internet, it has become easy to publish reports and their rebuttal with such conviction that it becomes a mind-boggling affair for the average reader to fish out the real truth in the stories.Political parties with an axe to grind are mostly instrumental in carrying out such misdeeds. A recent report about a newly constructed bridge in the National Capital actually carried a picture of a bridge from Singapore. Increasingly, the net is emerging as a platform for political slugfest with a mix of genuine and fake news with the factory involved in producing the latter category getting increasingly better at their jobs. In that sense, Fake news have a huge potential to sway public opinions and if they continue to prosper in this unregulated manner, one wonders whether the fate of a genuine article is similarly destined like the shepherd boy prone to playing pranks on the unsuspecting villagers till one day when they got tired …

Leave a comment

Filed under Articles, Technology: Boon or bane?

When your child commits a crime?

Recently when wifey returned one day from work, she told me about an incident that left her fuming. She said that during the day she had received a whatsapp message from the class teacher of our son. It showed him in a picture as one of the three students lined up for punishment. The picture was similar to the ones we see in newspapers, where criminals are made to line up and photographs are taken of them where humiliation is writ large on their faces. The crime the three high school standard boys have committed was of not doing their homework, which emboldened the subject teacher to take a snap and forward it to the class teacher, who whatsapped that same picture to my wife. When my wife strongly objected to this way of punishing a child, reasoning that such issues should be resolved at school, the class teacher informed us that they keep sending such pictures of insincere students being punished similarly, and hardly any parents have criticized this method. I don’t know how far this is true about ‘accepting parents’ but IMHO this practice is a blatant misuse of technology, where groups of classes are formed on whatsapp, activities and information and notes sharing happen therein. Imagine if the teacher posts and the child sees his picture flashed in his class group in whatsapp!!! It may damage his morale beyond measure. Thankfully, this didn’t happen in the case of our child.

Not without reasons, France has banned mobile usage for school work recently. I strongly believe that school kids should be kept away from gadgets and no matter how much the world is changing, foundational concepts are built only by reading school text books the way conventional education was imparted in earlier days …

Leave a comment

Filed under Articles, Technology: Boon or bane?

Constraint Satisfaction problem with example

 

In real life problem solving and applications, we are asked to provide the best solution keeping certain constraints in mind. Let us consider an example. Each year the National Film Awards jury has to pick a certain film as the best film of that year. How would the assessing Jury arrive at a reasonably good judgement? Do they have to watch each and every single film that year? Simple logic says that if the Jury members decide to watch three films per day it will take them the whole year (India makes over a thousand films each year) to view all of them. How does one handle such a solution demanding situation?

As with all complex problems, we need to use good heuristics in this case because evaluation of all possible entity is simply infeasible. Let us say that at the outset the National Jury need to be provided a list of best fifty films that year which can possible make the final grade. Depending on the size of various regional film industries, the selection number from that industry can be fixed and local film critics/Local Jury member can allocate scores/grades and select the best from that industry. A criteria could be set that LFSR (Local Film Score Rating) should be 0.6 and above on a scale of 1.

Similarly the apex film jury can allocate NFSR (National Film Score Rating) and based on summative score of LFSR & NFSR the best film can be chosen. We may further set additional criteria that NFSR >0.7 and LFSR+NFSR>0.8. These criteria becomes essential to filter and set standards so that the winning film is worthy of the prestigious honor.

Let us assume that the decision is arrived at through the opinion of ten jury members, five at local level represented by LJ1, LJ2, LJ3, LJ4, LJ5 and five at National levels are represented by NJ1, NJ2, NJ3, NJ4 & NJ5 …

So, now we basically have to compute the following:

Maximize (LFSR+NFSR)   where   LJ1+LJ2+LJ3+LJ4+LJ5= LFSR

&      NJ1+NJ2+NJ3+NJ4+NJ5= NFSR

Subject to       (LFSR/5)> 0.6

(NFSR/5)> 0.7

((LFSR+NFSR)/10)>0.8

Leave a comment

Filed under Articles, artificial intelligence

Basics about Search

  1. Why are Searching and Searching algorithms important in AI?

The operation of an AI system can be characterised as a search process in which rules are tried until some sequence is found that produces the desired solution. While selecting a sequence of operations to determine an exact solution of a problem, an AI system has to face a large number of alternatives all of which are not at all equally promising for achieving the solution. Individual evaluation of all these alternatives even with the fastest available computer requires very long time which may exceed the entire lifetime of a human being depending on the complexity of the problem.

2. What is a Search Space?

A node is a discrete point and possible goal in the solution path. The set of all nodes is the search space

3. What are Blind Searches? Why do we need to know about blind searches?

A blind search (also called an uninformed search) is a search that has no information about its domain. The only capability of blind search is its ability to distinguish a non-goal state from the goal state. Blind searches have no preferences as to which nodes it should explore first. Examples of uninformed searches are Breadth First Search and Depth First Search.

One needs to know about uninformed search as they form the basis for some of the intelligent searches.

4. What is a Heuristic?

A heuristic is the approximation done to reduce the search space.  A heuristic function defines a state in terms of a number and this number is used for decision making in the search problems.

Heuristics are used when

  • The solution path for the problem is not known
  • The perfect solution is not computationally feasible

In Best First Search algorithm, the heuristic used is the distance of the node from the goal node

5. List some applications of Heuristic search

The application of Heuristic Search fall into three general classes

  • Path finding problems
  • Two player games
  • Constraint Satisfaction problem                                                                                                                                                                                                                              6.  What is an Informed Search?

Informed Search is a heuristic search in which there is a domain specific knowledge of the search. The heuristics are used in informed search to reduce the search space. Some of the informed searches are Best First Search and Hill Climbing.

 

Leave a comment

Filed under artificial intelligence

Challenges before Indian private universities- 2

ASSESSING FACULTY MERIT:

It has been over a decade now that I am in the Technical Higher Education space in the private sector. All these institutes are governed by Govt. regulatory agencies. When it comes to ranking the institutes, the current practice being followed by the accrediting agencies are to accord substantial weight-age to Research and Publications. Even faculty career progression is mostly linked to the number of publications the faculty is credited with.

I have seen on quite a few instances some faculties who have great difficulties in expressing themselves flawlessly showcasing their long publication list on social media. On occasions, students have given feedback about faculties who are not too sincere about teaching in the class, but they are engaged in publishing so-called academic paper every alternate week or month. What is the casualty in all of this? No prizes for guessing the correct answer – Teaching.

It is no surprise therefore that the students who graduate from such institutes are often found lacking to fit readily into the Industry and companies need to invest substantial resources on them.

I feel if institutes are keen to survive as Centers of Excellence in the long run they need to segregate faculties into two categories: Teaching and Researchers, the expectation from each of these categories ought to be entirely different.

If we are really serious about stemming sub-quality research, which can become detrimental to the cause of real education, there is an urgent need to stop patronage of it. A report that has been published (https://www.ndtv.com/education/88-per-cent-ugc-approved-list-of-journals-are-low-quality-study-1828429) recently has exposed the fallacy of evaluating merit of faculties through such low quality publication. Teachers should primarily be evaluated for their teaching skills and their acceptability and utility in the classroom. Ignoring this could have grave consequences for the future of quality technical education in the country…

 

Leave a comment

Filed under Pedagogical: Teaching-Learning related

What is an AI technique?

AI problem solving involves complexity. To handle complex problems, the techniques need to have the capability to deliver acceptable solutions. Some of the common AI techniques in use are the following:

  1. Learning – Learning enables to perform correctly and efficiently an already encountered problem in the past that have been solved. In a game of chess between a human and a computer, if the computer encounters a similar board position and a move by the player which led to its defeat, the computer would have learnt from its mistake and try to circumvent the situation by playing a different game so as not to repeat his failure.
  2. Heuristics – Heuristics are intelligent guesses. When an AI chess program is your opponent in a game of chess, the intelligent program must be able to infer correctly (intelligent guess) the move that you would make …
  3. Decision making under uncertainty – Driving requires reasoning/decision making under uncertain situations by the driver. A driverless car should be equipped with similar capabilities to move on the roads. AI techniques such as Fuzzy logic are often used for arriving at decisions under uncertain situations.
  4. Natural Language Processing (NLP) – Text Mining and Analytics techniques, Sentiment Analysis and NLP techniques at word, sentence, paragraph and document level, the interpretation of unstructured text data at syntax and semantic level, effective speech recognition and synthesis methodology are essential for AI systems.
  5. Embed Common Sense– Humans have an innate ability of logical reasoning. If an individual is a Ph. D. it is common sense that the person possesses an under-graduate and post-graduate degree. Embedding machines with common sense reasoning is a non-trivial task.
  6. Knowledge Based (Logic) Reasoning – Deriving knowledge / inference from a given set of facts. Feeding parametric values of a patient into medical Expert Systems may result in proper diagnosis of the ailment and enable right prescription of drugs.
  7. Swarm intelligence – For combinatorial problems i.e., problems having a large number of possible alternatives, collective intelligence in the bird and animal kingdom such as ants (ant colony optimization) and birds (particle swarm optimization) for arriving at optimized values need be incorporated for problem solving by AI systems.
  8. Case Based Reasoning (CBR) – CBR systems make use of knowledge/experience gained in the past for current problem solving.
  9. Predictive ability – Neural Networks, Statistical techniques like Regression and K-Nearest Neighbor, Bayesian Classification enable good prediction and has wide applications …
  10. Soft Computing – The ability to perform computation with imprecise data using Genetic Algorithms, Fuzzy Logic and Neural Networks or a combination of them forming hybrid techniques are widely used in several AI problems …
  11. Fault Tolerance – In many situations, humans are able to interpret intelligently even from wrong answers. As an example, India has many linguistic groups with similar surnames. There have been instances when people kept referring to ‘Mr. Das’ and ‘Mr. Bose’ and I had to infer that they were actually referring to me – Ghosh. The three surnames are mostly seen in the Bengali community and others tend to err while addressing… such capabilities need to be incorporated into AI systems…

The above non-exhaustive listing is a pointer to the widely used AI methods and techniques in recent times…

Leave a comment

Filed under Articles, artificial intelligence